<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.1 20151215//EN" "http://jats.nlm.nih.gov/publishing/1.1/JATS-journalpublishing1.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="research-article" dtd-version="1.1" specific-use="sps-1.9" xml:lang="pt">
  <front>
    <journal-meta>
      <journal-id journal-id-type="nlm-ta">Arq Bras Cardiol</journal-id>
      <journal-id journal-id-type="publisher-id">abc</journal-id>
      <journal-title-group>
        <journal-title>Arquivos Brasileiros de Cardiologia</journal-title>
        <abbrev-journal-title abbrev-type="publisher">Arq. Bras. Cardiol.</abbrev-journal-title>
      </journal-title-group>
      <issn pub-type="ppub">0066-782X</issn>
      <issn pub-type="epub">1678-4170</issn>
      <publisher>
        <publisher-name>Sociedade Brasileira de Cardiologia - SBC</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id specific-use="scielo-v3" pub-id-type="publisher-id">6Jyy5jHtZVmY6Sm6DXVrdNw</article-id>
      <article-id specific-use="scielo-v2" pub-id-type="publisher-id">S0066-782X2026000200604</article-id>
      <article-id pub-id-type="other">00604</article-id>
      <article-id pub-id-type="doi">10.36660/abc.20250498</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Artigo Original</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Avalia&#231;&#227;o do ChatGPT-4.0 Versus ChatGPT-Mini na Gera&#231;&#227;o de Conte&#250;do sobre Hipertens&#227;o Baseado em Diretrizes</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-4570-869X</contrib-id>
          <name>
            <surname>Ata&#237;des</surname>
            <given-names>R&#244;mullo Jos&#233; Costa</given-names>
          </name>
          <role>Concep&#231;&#227;o e desenho da pesquisa</role>
          <role>Obten&#231;&#227;o de dados</role>
          <role>An&#225;lise e interpreta&#231;&#227;o dos dados e An&#225;lise estat&#237;stica</role>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Campos</surname>
            <given-names>Marcos Adriano Garcia</given-names>
          </name>
          <role>Concep&#231;&#227;o e desenho da pesquisa</role>
          <role>Obten&#231;&#227;o de dados</role>
          <role>An&#225;lise e interpreta&#231;&#227;o dos dados e An&#225;lise estat&#237;stica</role>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Souza</surname>
            <given-names>Jo&#227;o V&#237;tor Perez de</given-names>
          </name>
          <role>An&#225;lise e interpreta&#231;&#227;o dos dados e An&#225;lise estat&#237;stica</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Rocha</surname>
            <given-names>Rafael Cardoso</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff3">
            <sup>3</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0009-0008-7777-6727</contrib-id>
          <name>
            <surname>Lacalle</surname>
            <given-names>Almir Alamino</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff3">
            <sup>3</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Vieira</surname>
            <given-names>Ciro Bezerra</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-6242-0885</contrib-id>
          <name>
            <surname>Artioli</surname>
            <given-names>Thiago</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff4">
            <sup>4</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-5610-7921</contrib-id>
          <name>
            <surname>Medeiros</surname>
            <given-names>Tiago Cordeiro</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff5">
            <sup>5</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0381-3344</contrib-id>
          <name>
            <surname>Souza</surname>
            <given-names>Erito Marques de</given-names>
            <suffix>Filho</suffix>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff6">
            <sup>6</sup>
          </xref>
          <xref ref-type="aff" rid="aff7">
            <sup>7</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0884-4190</contrib-id>
          <name>
            <surname>Gismondi</surname>
            <given-names>Ronaldo</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff7">
            <sup>7</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6352-3330</contrib-id>
          <name>
            <surname>Campana</surname>
            <given-names>&#201;rika Maria Gon&#231;alves</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff8">
            <sup>8</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-1261-5785</contrib-id>
          <name>
            <surname>Romeo</surname>
            <given-names>Francisco Jos&#233;</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff9">
            <sup>9</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Razuk</surname>
            <given-names>Victor</given-names>
          </name>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff9">
            <sup>9</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Vissoci</surname>
            <given-names>Jo&#227;o Ricardo Nickenig</given-names>
          </name>
          <role>Concep&#231;&#227;o e desenho da pesquisa</role>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Lopes</surname>
            <given-names>Renato Delascio</given-names>
          </name>
          <role>Concep&#231;&#227;o e desenho da pesquisa</role>
          <role>Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do</role>
          <xref ref-type="aff" rid="aff10">
            <sup>10</sup>
          </xref>
        </contrib>
      </contrib-group>
      <aff id="aff1">
        <label>1</label>
        <institution content-type="orgdiv1">Faculdade de Medicina</institution>
        <institution content-type="orgname">Universidade de S&#227;o Paulo</institution>
        <addr-line>
          <named-content content-type="city">S&#227;o Paulo</named-content>
          <named-content content-type="state">SP</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Faculdade de Medicina da Universidade de S&#227;o Paulo, S&#227;o Paulo, SP &#8211; Brasil</institution>
      </aff>
      <aff id="aff2">
        <label>2</label>
        <institution content-type="orgdiv1">Departamento de Medicina de Emerg&#234;ncia</institution>
        <institution content-type="orgname">Duke University</institution>
        <addr-line>
          <named-content content-type="city">Durham</named-content>
        </addr-line>
        <country country="US">EUA</country>
        <institution content-type="original">Departamento de Medicina de Emerg&#234;ncia &#8211; Duke University, Durham &#8211; EUA</institution>
      </aff>
      <aff id="aff3">
        <label>3</label>
        <institution content-type="orgdiv2">Divis&#227;o de Cardiologia</institution>
        <institution content-type="orgdiv1">Hospital das Cl&#237;nicas de Ribeir&#227;o Preto</institution>
        <institution content-type="orgname">Universidade de S&#227;o Paulo</institution>
        <addr-line>
          <named-content content-type="city">Ribeir&#227;o Preto</named-content>
          <named-content content-type="state">SP</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Divis&#227;o de Cardiologia do Hospital das Cl&#237;nicas de Ribeir&#227;o Preto da Universidade de S&#227;o Paulo, Ribeir&#227;o Preto, SP &#8211; Brasil</institution>
      </aff>
      <aff id="aff4">
        <label>4</label>
        <institution content-type="orgname">Instituto Dante Pazzanese de Cardiologia</institution>
        <addr-line>
          <named-content content-type="city">S&#227;o Paulo</named-content>
          <named-content content-type="state">SP</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Instituto Dante Pazzanese de Cardiologia, S&#227;o Paulo, SP &#8211; Brasil</institution>
      </aff>
      <aff id="aff5">
        <label>5</label>
        <institution content-type="orgdiv1">Hospital Universit&#225;rio</institution>
        <institution content-type="orgname">Universidade Federal do Maranh&#227;o</institution>
        <addr-line>
          <named-content content-type="city">S&#227;o Lu&#237;s</named-content>
          <named-content content-type="state">MA</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Hospital Universit&#225;rio &#8211; Universidade Federal do Maranh&#227;o, S&#227;o Lu&#237;s, MA &#8211; Brasil</institution>
      </aff>
      <aff id="aff6">
        <label>6</label>
        <institution content-type="orgname">Universidade Federal Rural do Rio de Janeiro</institution>
        <addr-line>
          <named-content content-type="city">Serop&#233;dica</named-content>
          <named-content content-type="state">RJ</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Universidade Federal Rural do Rio de Janeiro, Serop&#233;dica, RJ &#8211; Brasil</institution>
      </aff>
      <aff id="aff7">
        <label>7</label>
        <institution content-type="orgname">Universidade Federal Fluminense</institution>
        <addr-line>
          <named-content content-type="city">Niter&#243;i</named-content>
          <named-content content-type="state">RJ</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Universidade Federal Fluminense, Niter&#243;i, RJ &#8211; Brasil</institution>
      </aff>
      <aff id="aff8">
        <label>8</label>
        <institution content-type="orgname">Universidade do Estado do Rio de Janeiro</institution>
        <addr-line>
          <named-content content-type="city">Rio de Janeiro</named-content>
          <named-content content-type="state">RJ</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original">Universidade do Estado do Rio de Janeiro, Rio de Janeiro, RJ &#8211; Brasil</institution>
      </aff>
      <aff id="aff9">
        <label>9</label>
        <institution content-type="orgdiv2">Departamento de Medicina Cardiovascular</institution>
        <institution content-type="orgname">University of Miami Miller School of Medicine</institution>
        <institution content-type="orgdiv1">Jackson Memorial Hospital</institution>
        <addr-line>
          <named-content content-type="city">Miami</named-content>
        </addr-line>
        <country country="US">EUA </country>
        <institution content-type="original">Departamento de Medicina Cardiovascular &#8211; University of Miami Miller School of Medicine/Jackson Memorial Hospital, Miami &#8211; EUA </institution>
      </aff>
      <aff id="aff10">
        <label>10</label>
        <institution content-type="orgname">Duke Clinical Research Institute Cardiology</institution>
        <addr-line>
          <named-content content-type="city">Durham</named-content>
        </addr-line>
        <country country="US">EUA</country>
        <institution content-type="original">Duke Clinical Research Institute Cardiology, Durham &#8211; EUA</institution>
      </aff>
      <author-notes>
        <corresp id="c01"><label>Correspond&#234;ncia:</label>R&#244;mullo Jos&#233; Costa Ata&#237;des Av. Dr. Arnaldo, 455. CEP 01246-903, Cerqueira C&#233;sar, Pacaembu, SP - Brasil E-mail: <email>romulloataides2@hotmail.com</email> </corresp>
        <fn fn-type="edited-by">
          <label>Editor respons&#225;vel pela revis&#227;o:</label>
          <p>Gl&#225;ucia Maria Moraes de Oliveira</p>
        </fn>
        <fn fn-type="coi-statement">
          <label>Potencial conflito de interesse:</label>
          <p>N&#227;o h&#225; conflito com o presente artigo</p>
        </fn>
      </author-notes>
      <pub-date date-type="pub" publication-format="electronic">
        <day>16</day>
        <month>03</month>
        <year>2026</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>02</month>
        <year>2026</year>
      </pub-date>
      <volume>123</volume>
      <issue>2</issue>
      <elocation-id>e20250498</elocation-id>
      <history>
        <date date-type="received">
          <day>23</day>
          <month>07</month>
          <year>2025</year>
        </date>
        <date date-type="rev-recd">
          <day>16</day>
          <month>12</month>
          <year>2025</year>
        </date>
        <date date-type="accepted">
          <day>13</day>
          <month>01</month>
          <year>2026</year>
        </date>
      </history>
      <permissions>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/" xml:lang="pt">
          <license-p>Este &#233; um artigo de acesso aberto distribu&#237;do sob os termos da licen&#231;a de atribui&#231;&#227;o pelo Creative Commons</license-p>
        </license>
      </permissions>
      <abstract>
        <title>Resumo</title>
        <sec>
          <title>Fundamento</title>
          <p> Modelos de linguagem baseados em intelig&#234;ncia artificial est&#227;o sendo cada vez mais utilizados para gerar materiais de educa&#231;&#227;o em sa&#250;de para pacientes. No entanto, sua acur&#225;cia, completude e ader&#234;ncia &#224;s diretrizes cl&#237;nicas permanecem incertas.</p>
        </sec>
        <sec>
          <title>Objetivos</title>
          <p> Comparar o ChatGPT-Mini e o ChatGPT-4.0 na gera&#231;&#227;o de conte&#250;do educativo sobre hipertens&#227;o arterial sist&#234;mica (HAS) quanto &#224; acur&#225;cia, completude, qualidade estrutural utilizando o Ensuring Quality Information for Patients (EQIP), consist&#234;ncia das respostas e alinhamento com diretrizes estabelecidas.</p>
        </sec>
        <sec>
          <title>M&#233;todos</title>
          <p> Um conjunto padronizado de 31 perguntas relacionadas &#224; HAS foi submetido a ambos os modelos. As respostas foram avaliadas de forma independente por 10 cl&#237;nicos cegos, utilizando um escore EQIP modificado, uma escala de acur&#225;cia de 5 pontos e uma escala de completude de 3 pontos. A consist&#234;ncia das respostas foi avaliada por meio do BERTScore. As compara&#231;&#245;es entre os modelos foram realizadas utilizando o teste de Wilcoxon para amostras independentes, bicaudal (p &lt; 0,05). Os tamanhos de efeito foram apresentados como diferen&#231;as medianas de Hodges&#8211;Lehmann (HL) e delta de Cliff (&#948;), ambos com intervalos de confian&#231;a de 95% (IC 95%). A confiabilidade interavaliadores foi estimada por meio do coeficiente de correla&#231;&#227;o intraclasse (ICC; modelo de efeitos aleat&#243;rios em duas vias, concord&#226;ncia absoluta).</p>
        </sec>
        <sec>
          <title>Resultados</title>
          <p> As medidas de tend&#234;ncia central favoreceram o ChatGPT-4.0, embora as diferen&#231;as tenham sido pequenas. As medianas foram as seguintes: acur&#225;cia, 4,10 (3,70-4,20) versus 3,73 (3,60-4,05); completude, 1,26 (1,17-1,41) versus 1,10 (0,96-1,23); e escore EQIP total, 19,5 (18,0-25,0) versus 18,5 (16,0-23,0) para o ChatGPT-4.0 e o ChatGPT-Mini, respectivamente. As diferen&#231;as medianas de HL foram pequenas, com ICs de 95% cruzando zero (acur&#225;cia: +0,37, &#8722;0,25 a +0,50; completude: +0,16, &#8722;0,06 a +0,36; EQIP: +1,0, &#8722;1,0 a +6,0). Os valores de &#948; de Cliff foram consistentemente pequenos e positivos nos desfechos prim&#225;rios, indicando apenas modesta domin&#226;ncia estoc&#225;stica do ChatGPT-4.0. A clareza de identifica&#231;&#227;o tendeu a ser maior com o ChatGPT-4.0, enquanto a consist&#234;ncia das respostas, medida pelo BERTScore F1, foi geralmente maior para o ChatGPT-Mini (&gt; 0,92 versus 0,885-0,932). A confiabilidade interavaliadores foi de boa a excelente em todas as medidas (ICC &gt; 0,80).</p>
        </sec>
        <sec>
          <title>Conclus&#245;es</title>
          <p> O ChatGPT-4.0 demonstrou melhorias pequenas e n&#227;o significativas em acur&#225;cia, completude e qualidade estrutural em compara&#231;&#227;o com o ChatGPT-Mini. Os tamanhos de efeito foram modestos, e todos os ICs de 95% inclu&#237;ram zero. O ChatGPT-Mini produziu respostas mais consistentes. Esses achados ressaltam a import&#226;ncia de relatar rotineiramente tamanhos de efeito com ICs de 95% e apoiam o uso de m&#233;todos padronizados de avalia&#231;&#227;o e de estruturas de valida&#231;&#227;o em tempo real para conte&#250;do m&#233;dico educativo gerado por IA.</p>
        </sec>
      </abstract>
      <abstract abstract-type="graphical">
        <p>
          <fig id="f01">
            <label>Figura Central:</label>
            <caption>
              <title>Avalia&#231;&#227;o do ChatGPT-4.0 Versus ChatGPT-Mini na Gera&#231;&#227;o de Conte&#250;do sobre Hipertens&#227;o Baseado em Diretrizes</title>
            </caption>
            <attrib>ACC: Col&#233;gio Americano de Cardiologia; AHA: Associa&#231;&#227;o Americana do Cora&#231;&#227;o; EQIP: Ensuring Quality Information for Patients; ESC: Sociedade Europeia de Cardiologia; ESH: Sociedade Europeia de Hipertens&#227;o; HAS: hipertens&#227;o arterial sist&#234;mica; IA: intelig&#234;ncia artificial; ICC: coeficiente de correla&#231;&#227;o intraclasse.</attrib>
            <alternatives>
              <graphic xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/3439b79d7fbf6ae6d3a7b56f8115be8fec820a3f.tif"/>
              <graphic xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/605fb7bfbf8ce332f04a5a46871a2060aa9f0653.png" specific-use="scielo-web"/>
              <graphic xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/04d95c2e31c7fe0faa677b703c26d46755634701.jpg" specific-use="scielo-web" content-type="scielo-267x140"/>
            </alternatives>
          </fig>
        </p>
      </abstract>
      <kwd-group xml:lang="pt">
        <title>Palavras-chave</title>
        <kwd>Intelig&#234;ncia Artificial</kwd>
        <kwd>Hipertens&#227;o</kwd>
        <kwd>Educa&#231;&#227;o de Pacientes como Assunto</kwd>
      </kwd-group>
      <funding-group>
        <funding-statement><bold>Fontes de financiamento:</bold> O presente estudo n&#227;o teve fontes de financiamento externas.</funding-statement>
      </funding-group>
      <counts>
        <fig-count count="8"/>
        <table-count count="0"/>
        <equation-count count="0"/>
        <ref-count count="37"/>
      </counts>
    </article-meta>
  </front>
  <body>
    <sec sec-type="intro">
      <title>Introdu&#231;&#227;o</title>
      <p>O uso da intelig&#234;ncia artificial (IA) no manejo da hipertens&#227;o arterial sist&#234;mica (HAS) oferece um potencial substancial.<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref></sup> No entanto, as evid&#234;ncias atuais indicam que sua utiliza&#231;&#227;o ainda permanece insuficientemente sistematizada, particularmente no que se refere aos crit&#233;rios de avalia&#231;&#227;o e &#224;s abordagens centradas no paciente.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref></sup> Por exemplo, ferramentas baseadas em IA, como chatbots, demonstraram utilidade no apoio ao autocuidado da HAS; contudo, os estudos existentes destacam de forma consistente a necessidade de estruturas de avalia&#231;&#227;o mais robustas e de orienta&#231;&#245;es mais claras e consistentes para os pacientes.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
      <p>Almagazzachi et al. investigaram o uso de IA generativa na educa&#231;&#227;o de pacientes com HAS e relataram que, apesar de fornecerem apoio significativo, essas ferramentas apresentam inconsist&#234;ncias no alinhamento com diretrizes cl&#237;nicas estabelecidas, incluindo as da Sociedade Europeia de Hipertens&#227;o (do ingl&#234;s ESH).<sup><xref ref-type="bibr" rid="B6">6</xref></sup> Al&#233;m disso, foi documentada variabilidade no desempenho dos chatbots entre diferentes idiomas, o que refor&#231;a ainda mais a necessidade de metodologias padronizadas para garantir acur&#225;cia, confiabilidade e ader&#234;ncia &#224;s diretrizes.<sup><xref ref-type="bibr" rid="B7">7</xref></sup> A ado&#231;&#227;o de uma abordagem centrada no usu&#225;rio no desenvolvimento de solu&#231;&#245;es de IA para o manejo da HAS pode aprimorar a sistematiza&#231;&#227;o e melhorar os desfechos dos pacientes.<sup><xref ref-type="bibr" rid="B5">5</xref></sup> Assim, embora a IA demonstre grande potencial nessa &#225;rea, &#233; necess&#225;rio um arcabou&#231;o mais estruturado e focado no paciente para que seu potencial seja plenamente concretizado no cuidado da HAS.<sup><xref ref-type="bibr" rid="B8">8</xref></sup></p>
      <p>Diversos estudos t&#234;m buscado sistematizar a avalia&#231;&#227;o de informa&#231;&#245;es educativas geradas por IA para pacientes.<sup><xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref></sup> No entanto, esses esfor&#231;os n&#227;o incorporaram um processo sistem&#225;tico de desenvolvimento de perguntas de avalia&#231;&#227;o baseado em consenso de especialistas.<sup><xref ref-type="bibr" rid="B11">11</xref></sup></p>
      <p>O presente estudo tem como objetivo avaliar o desempenho de duas vers&#245;es do modelo ChatGPT, ChatGPT-Mini e ChatGPT-4.0, na gera&#231;&#227;o de conte&#250;do m&#233;dico relacionado &#224; HAS. Instrumentos validados foram utilizados para avaliar a qualidade das informa&#231;&#245;es destinadas aos pacientes e para examinar os pontos fortes e as limita&#231;&#245;es da IA na educa&#231;&#227;o m&#233;dica e no cuidado ao paciente.<sup><xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref></sup> Al&#233;m disso, ao aplicar uma avalia&#231;&#227;o estruturada e padronizada das respostas geradas por IA, este estudo busca identificar &#225;reas em que o conte&#250;do produzido por IA pode ser deficiente, fornecendo, assim, subs&#237;dios relevantes para orientar o desenvolvimento futuro da IA na &#225;rea da sa&#250;de.<sup><xref ref-type="bibr" rid="B13">13</xref></sup> Essa abordagem sistem&#225;tica permite avaliar o qu&#227;o eficazmente os modelos de IA fornecem materiais educativos precisos e centrados no paciente e contribui para o aprimoramento de estrat&#233;gias baseadas em IA para a dissemina&#231;&#227;o de informa&#231;&#245;es m&#233;dicas.</p>
    </sec>
    <sec sec-type="methods">
      <title>M&#233;todos</title>
      <sec>
        <title>Desenho do estudo</title>
        <p>Este estudo empregou um desenho baseado em question&#225;rio (<italic>survey</italic>) (<xref ref-type="fig" rid="f02">Figura 1</xref>). Perguntas abordando o diagn&#243;stico, o tratamento e o progn&#243;stico da HAS foram submetidas ao ChatGPT-Mini e ao ChatGPT-4.0 (<xref ref-type="sec" rid="suppl01">Tabela Suplementar 1</xref>). As respostas geradas foram comparadas com as recomenda&#231;&#245;es das diretrizes de HAS da ESH de 2023 e do Col&#233;gio Americano de Cardiologia/Associa&#231;&#227;o Americana do Cora&#231;&#227;o (ACC/AHA) de 2017. Todas as respostas foram avaliadas por cardiologistas e m&#233;dicos em forma&#231;&#227;o em cardiologia, de forma independente.</p>
        <p>
          <fig id="f02">
            <label>Figura 1</label>
            <caption>
              <title>&#8211; Fluxograma do estudo para a avalia&#231;&#227;o de conte&#250;do sobre hipertens&#227;o gerado por intelig&#234;ncia artificial. ACC: Col&#233;gio Americano de Cardiologia; AHA: Associa&#231;&#227;o Americana do Cora&#231;&#227;o; EQIP: Ensuring Quality Information for Patients; ESH: Sociedade Europeia de Hipertens&#227;o; HAS: hipertens&#227;o arterial sist&#234;mica.</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf01.tif"/></fig>
        </p>
      </sec>
      <sec>
        <title>Coleta de dados</title>
        <p>A equipe de pesquisa desenvolveu 31 perguntas relacionadas &#224; HAS (Tabela 1; consultar o <xref ref-type="sec" rid="suppl01">Material Suplementar 1</xref> para detalhes adicionais). O conte&#250;do e a estrutura de cada pergunta foram orientados pelo escore <italic>Ensuring Quality Information for Patients</italic> (EQIP).<sup><xref ref-type="bibr" rid="B14">14</xref></sup> O escore EQIP &#233; uma ferramenta validada, desenvolvida para avaliar a qualidade das informa&#231;&#245;es fornecidas aos pacientes em contextos de cuidado em sa&#250;de, com foco na acur&#225;cia, confiabilidade e utilidade no apoio &#224; tomada de decis&#245;es informadas relacionadas &#224; sa&#250;de.</p>
        <p>Cada pergunta foi submetida 10 vezes, em ingl&#234;s, a ambas as vers&#245;es do ChatGPT, entre 25 e 30 de julho de 2024. Todas as respostas foram registradas de forma sistem&#225;tica, iniciando-se novas conversas com o uso de <italic>prompts</italic> incondicionais em modo de chat tempor&#225;rio. Nos casos em que o formato da resposta n&#227;o abordava adequadamente a pergunta original, o <italic>prompt</italic> &#8220;<italic>Be specific and incorporate medical knowledge about hypertension from the guideline of hypertension from AHA/ACC 2017 and ESH 2023</italic>&#8221; foi acrescentado &#224; conversa conforme necess&#225;rio.</p>
      </sec>
      <sec>
        <title>Avalia&#231;&#227;o especializada das respostas do GPT</title>
        <p>Um total de 10 m&#233;dicos (cinco cardiologistas e cinco m&#233;dicos em forma&#231;&#227;o em cardiologia), provenientes de m&#250;ltiplas institui&#231;&#245;es no Brasil e nos Estados Unidos, avaliou de forma independente as respostas geradas pelo ChatGPT (<xref ref-type="fig" rid="f03">Figura 2</xref>). Os avaliadores receberam instru&#231;&#245;es escritas detalhando como interpretar as respostas, aplicar os crit&#233;rios de avalia&#231;&#227;o e comparar o conte&#250;do com as diretrizes de HAS ACC/AHA de 2017 e ESH de 2023. Embora os avaliadores pudessem consultar as diretrizes, foram incentivados a basear suas avalia&#231;&#245;es principalmente em sua experi&#234;ncia cl&#237;nica.</p>
        <p>
          <fig id="f03">
            <label>Figura 2</label>
            <caption>
              <title>&#8211; Exemplo da avalia&#231;&#227;o de respostas sobre HAS geradas por intelig&#234;ncia artificial. EQIP: Ensuring Quality Information for Patients; HAS: hipertens&#227;o arterial sist&#234;mica.</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf02.tif"/></fig>
        </p>
        <p>Todos os avaliadores estavam cegos quanto &#224; origem das respostas e n&#227;o sabiam se o conte&#250;do havia sido gerado pelo ChatGPT-Mini ou pelo ChatGPT-4.0. As avalia&#231;&#245;es foram coletadas por meio de question&#225;rios criados na plataforma REDCap (<xref ref-type="sec" rid="suppl01">Material Suplementar 1</xref>).</p>
      </sec>
    </sec>
    <sec>
      <title>Medidas</title>
      <p>Com base em informa&#231;&#245;es derivadas das diretrizes e no conhecimento cl&#237;nico, os avaliadores classificaram cada resposta utilizando tr&#234;s componentes: o escore EQIP para avaliar a qualidade global da informa&#231;&#227;o e duas escalas do tipo Likert para avaliar, de forma independente, a acur&#225;cia e a completude (<xref ref-type="sec" rid="suppl01">Material Suplementar 1</xref>).</p>
      <p>O escore EQIP foi adaptado para a avalia&#231;&#227;o das respostas geradas pelo ChatGPT. Ele consiste em crit&#233;rios que abrangem tr&#234;s dom&#237;nios: conte&#250;do (avaliando acur&#225;cia, relev&#226;ncia e abrang&#234;ncia), identifica&#231;&#227;o (avaliando clareza e facilidade de identifica&#231;&#227;o das informa&#231;&#245;es) e estrutura dos dados (examinando organiza&#231;&#227;o, legibilidade e apresenta&#231;&#227;o das informa&#231;&#245;es). O instrumento EQIP modificado permite uma pontua&#231;&#227;o m&#225;xima de 31 pontos, distribu&#237;dos entre os dom&#237;nios de conte&#250;do (m&#225;ximo de 16 pontos), identifica&#231;&#227;o (m&#225;ximo de 5 pontos) e estrutura dos dados (m&#225;ximo de 10 pontos).</p>
      <p>Cada crit&#233;rio do EQIP foi pontuado utilizando um sistema de tr&#234;s pontos: 1 (resposta correta), 0,5 (resposta parcialmente correta) ou 0 (resposta incorreta). O escore EQIP total foi calculado como a soma dos pontos em todos os crit&#233;rios, com escores mais elevados indicando informa&#231;&#245;es de maior qualidade. Para todas as respostas, um ponto foi automaticamente atribu&#237;do ao item &#8220;<italic>Can you date the above-named information?</italic>&#8221;, uma vez que esse elemento poderia potencialmente revelar a vers&#227;o do ChatGPT avaliada.</p>
      <p>A acur&#225;cia foi avaliada por meio de uma escala de cinco pontos: completamente incorreta (nenhuma informa&#231;&#227;o correta e contradit&#243;ria &#224;s diretrizes de HAS); mais incorreta do que correta (predominantemente incorreta e amplamente inconsistente com as diretrizes); aproximadamente correta e incorreta (propor&#231;&#227;o semelhante de informa&#231;&#245;es corretas e incorretas, parcialmente alinhada &#224;s diretrizes); mais correta do que incorreta (predominantemente correta e geralmente alinhada &#224;s diretrizes); e completamente correta (totalmente precisa, abrangente e inteiramente consistente com as diretrizes de HAS).</p>
      <p>A completude foi avaliada por meio de uma escala de tr&#234;s pontos, refletindo o grau em que cada resposta abordou a pergunta: incompleta (aspectos-chave ausentes ou abordados de forma insuficiente); adequada (todos os aspectos solicitados abordados com a informa&#231;&#227;o m&#237;nima necess&#225;ria); e abrangente (todos os aspectos abordados com contexto adicional relevante al&#233;m do esperado).</p>
      <p>A consist&#234;ncia das respostas ao longo das 10 repeti&#231;&#245;es de cada pergunta foi avaliada utilizando o BERTScore, que quantifica a similaridade sem&#226;ntica com base em escores de precis&#227;o, <italic>recall</italic> e F1. Os resultados foram apresentados graficamente para comparar a similaridade de conte&#250;do entre as respostas.<sup><xref ref-type="bibr" rid="B15">15</xref></sup></p>
      <sec>
        <title>Amostragem do ChatGPT</title>
        <p>O n&#250;mero m&#237;nimo de avaliadores necess&#225;rio para alcan&#231;ar um coeficiente de correla&#231;&#227;o intraclasse (ICC) de 0,8, com precis&#227;o de &#177; 0,1 e intervalo de confian&#231;a de 95% (IC 95%), ao avaliar os 31 itens gerados por cada vers&#227;o do ChatGPT, foi estimado em 10.</p>
      </sec>
      <sec>
        <title>An&#225;lise estat&#237;stica</title>
        <p>A distribui&#231;&#227;o das vari&#225;veis cont&#237;nuas foi avaliada por meio do teste de Shapiro&#8211;Wilk. Como todas as vari&#225;veis cont&#237;nuas apresentaram distribui&#231;&#227;o n&#227;o normal, os dados s&#227;o apresentados como medianas e intervalos interquartis (IIQ). As compara&#231;&#245;es entre o ChatGPT-Mini e o ChatGPT-4.0 foram realizadas utilizando o teste de Wilcoxon para amostras independentes (bicaudal, p &lt; 0,05).</p>
        <p>Os tamanhos de efeito foram apresentados com ICs 95%, incluindo as diferen&#231;as medianas de Hodges&#8211;Lehmann para estimar deslocamentos de localiza&#231;&#227;o e o delta de Cliff (&#948;) para avaliar domin&#226;ncia estoc&#225;stica. Os ICs 95% foram obtidos por meio de m&#233;dias de Walsh ou m&#233;todos de bootstrap n&#227;o param&#233;trico, conforme apropriado. A concord&#226;ncia interavaliadores foi quantificada utilizando o ICC (modelo de efeitos aleat&#243;rios em duas vias, concord&#226;ncia absoluta), com ICs 95%.</p>
        <p>As compara&#231;&#245;es entre o ChatGPT-4.0 e o ChatGPT-Mini nos dom&#237;nios de acur&#225;cia, completude e EQIP foram conduzidas utilizando o delta de Cliff, com valores de &#948; maiores que 0 indicando superioridade do ChatGPT-4.0. Todas as estimativas s&#227;o apresentadas com ICs de 95%, e a signific&#226;ncia estat&#237;stica foi definida como p &lt; 0,05. As an&#225;lises estat&#237;sticas foram realizadas no software R (vers&#227;o 4.2.2). As an&#225;lises de BERTScore foram conduzidas no Google Colab utilizando o pacote <italic>bert-score</italic>.</p>
      </sec>
    </sec>
    <sec sec-type="results">
      <title>Resultados</title>
      <sec>
        <title>Amostra do estudo e inclus&#227;o dos dados</title>
        <p>Todas as respostas geradas pelo ChatGPT-Mini e pelo ChatGPT-4.0 foram inclu&#237;das na an&#225;lise, sem exclus&#245;es ou dados ausentes. O conjunto padronizado de 31 perguntas foi aplicado de forma uniforme a ambos os modelos. Ao longo das 10 repeti&#231;&#245;es por pergunta, n&#227;o foram identificados epis&#243;dios de alucina&#231;&#227;o por IA. Al&#233;m disso, n&#227;o foram necess&#225;rios <italic>prompts</italic> de seguimento para esclarecimento das respostas, uma vez que todos os resultados abordaram adequadamente as perguntas originais.</p>
      </sec>
      <sec>
        <title>Acur&#225;cia e completude</title>
        <p>Conforme apresentado na Tabela 2, ambos os modelos alcan&#231;aram escores elevados de acur&#225;cia e completude. Os valores medianos foram numericamente mais altos para o ChatGPT-4.0; entretanto, n&#227;o foram observadas diferen&#231;as estatisticamente significativas entre os dois modelos para nenhum dos desfechos.</p>
      </sec>
      <sec>
        <title>global (escores EQIP)</title>
        <p>O escore EQIP, que integra os dom&#237;nios de conte&#250;do, identifica&#231;&#227;o e estrutura, indicou qualidade global amplamente compar&#225;vel entre os dois modelos. Conforme apresentado na Tabela 2, o ChatGPT-4.0 tendeu a alcan&#231;ar escores totais de EQIP numericamente mais elevados, bem como escores mais altos nos subdom&#237;nios de conte&#250;do e identifica&#231;&#227;o, em compara&#231;&#227;o com o ChatGPT-Mini. Os escores estruturais foram muito semelhantes entre os modelos, resultando em totais globais de EQIP compar&#225;veis.</p>
        <p>As an&#225;lises de tamanho de efeito utilizando o &#948; de Cliff favoreceram o ChatGPT-4.0 em todos os dom&#237;nios do EQIP; no entanto, todos os ICs 95% correspondentes inclu&#237;ram zero (Tabela 3), indicando que essas diferen&#231;as n&#227;o foram estatisticamente robustas. A <xref ref-type="fig" rid="f04">Figura 3</xref> ilustra a distribui&#231;&#227;o dos escores EQIP entre os dom&#237;nios para cada modelo, demonstrando avalia&#231;&#245;es geralmente favor&#225;veis dos avaliadores quanto ao conte&#250;do, &#224; identifica&#231;&#227;o e &#224; estrutura. Nessa figura, cada linha horizontal corresponde a um item individual do EQIP: os itens 1-16 representam o conte&#250;do, os itens 17-20 representam a identifica&#231;&#227;o e os itens 21-30 representam a estrutura. Em todos os itens, ambos os modelos receberam avalia&#231;&#245;es consistentemente favor&#225;veis e, em grande parte, compar&#225;veis.</p>
        <p>
          <fig id="f04">
            <label>Figura 3</label>
            <caption>
              <title>&#8211; Compara&#231;&#227;o das respostas do ChatGPT com base no escore EQIP modificado ao longo das quest&#245;es sobre HAS. As respostas geradas pelo ChatGPT-Mini e pelo ChatGPT-4.0 ao longo de 30 perguntas relacionadas &#224; HAS foram avaliadas utilizando um escore EQIP modificado. As respostas foram classificadas como 1 (correta), 0,5 (parcialmente correta) ou 0 (incorreta). O eixo x representa a propor&#231;&#227;o de respostas em cada categoria, e o eixo y lista as perguntas avaliadas. EQIP: Ensuring Quality Information for Patients; HAS: hipertens&#227;o arterial sist&#234;mica.</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf03.tif"/></fig>
        </p>
      </sec>
      <sec>
        <title>Avalia&#231;&#245;es por BERTScore</title>
        <p>As an&#225;lises por BERTScore revelaram diferen&#231;as entre os modelos quanto &#224; precis&#227;o, ao <italic>recall</italic> e &#224; consist&#234;ncia das respostas. O ChatGPT-Mini demonstrou consistentemente escores F1 mais elevados, frequentemente superiores a 0,92, enquanto o ChatGPT-4.0 apresentou escores F1 variando de 0,8851 a 0,9316 (consultar o <xref ref-type="sec" rid="suppl02">Material Suplementar 2</xref>).</p>
        <p>Em rela&#231;&#227;o &#224; precis&#227;o, o ChatGPT-Mini obteve um escore de 0,9356, em compara&#231;&#227;o com 0,8643 para o ChatGPT-4.0, em resposta a uma pergunta sobre a defini&#231;&#227;o de HAS. De forma semelhante, para o <italic>recall</italic>, o ChatGPT-Mini alcan&#231;ou um escore de 0,9558, enquanto o ChatGPT-4.0 obteve 0,9330 em uma pergunta de seguimento relacionada.</p>
        <p>De modo geral, o ChatGPT-Mini apresentou maior consist&#234;ncia na estabilidade das respostas, com escores F1 raramente inferiores a 0,90. Em contraste, o ChatGPT-4.0 demonstrou maior variabilidade, incluindo um escore F1 de 0,8757 em resposta a uma pergunta sobre tratamentos n&#227;o farmacol&#243;gicos para a HAS.</p>
      </sec>
      <sec>
        <title>Confiabilidade interavaliadores</title>
        <p>A confiabilidade interavaliadores foi consistentemente elevada em todas as m&#233;tricas avaliadas, com valores de ICC superiores a 0,80 para ambos os modelos.</p>
        <p>Para os escores EQIP, os ICCs foram de 0,844 (IC 95%, 0,677-0,952) para o ChatGPT-Mini e de 0,840 (IC 95%, 0,665-0,951) para o ChatGPT-4.0. Para os escores de acur&#225;cia, os ICCs foram de 0,742 (IC 95%, 0,456-0,922) para o ChatGPT-Mini e de 0,688 (IC 95%, 0,339-0,906) para o ChatGPT-4.0. Para os escores de completude, os ICCs foram id&#234;nticos para ambos os modelos, de 0,824 (IC 95%, 0,627-0,947 para o ChatGPT-Mini e 0,628-0,947 para o ChatGPT-4.0).</p>
        <p>Esses valores de ICC indicam concord&#226;ncia interavaliadores de moderada a forte, sustentando a confiabilidade da metodologia de avalia&#231;&#227;o utilizada neste estudo. De modo geral, os achados sugerem desempenho superior do ChatGPT-4.0 nos principais desfechos de qualidade (Figura Central).</p>
      </sec>
    </sec>
    <sec sec-type="discussion">
      <title>Discuss&#227;o</title>
      <p>&#192; medida que a IA se torna cada vez mais integrada &#224; educa&#231;&#227;o do paciente, persistem preocupa&#231;&#245;es quanto &#224; acur&#225;cia, &#224; completude e &#224; consist&#234;ncia das informa&#231;&#245;es m&#233;dicas geradas por IA. Embora a IA apresente potencial substancial para ampliar a acessibilidade aos cuidados de sa&#250;de, sua capacidade de fornecer conte&#250;do confi&#225;vel e aderente &#224;s diretrizes permanece um desafio cr&#237;tico, particularmente no manejo de condi&#231;&#245;es cr&#244;nicas como a HAS. Neste estudo, avaliamos de forma sistem&#225;tica o desempenho do ChatGPT-Mini e do ChatGPT-4.0 na gera&#231;&#227;o de educa&#231;&#227;o em HAS centrada no paciente, com foco no alinhamento &#224;s diretrizes, na qualidade informacional e na consist&#234;ncia das respostas.</p>
      <p>Nossos achados indicam que o ChatGPT-4.0 apresentou melhorias incrementais em rela&#231;&#227;o ao ChatGPT-Mini quanto &#224; acur&#225;cia, &#224; completude e &#224; estrutura das respostas. Contudo, essas diferen&#231;as foram modestas e n&#227;o atingiram signific&#226;ncia estat&#237;stica, provavelmente refletindo tamanhos de efeito pequenos. O ChatGPT-4.0 alcan&#231;ou uma m&#233;dia de acur&#225;cia ligeiramente superior &#224; do ChatGPT-Mini (4,02 &#177; 0,35 vs 3,86 &#177; 0,45), al&#233;m de uma mediana de completude mais elevada (1,26 vs 1,10). Apesar desses ganhos num&#233;ricos, ambos os modelos apresentaram limita&#231;&#245;es relevantes na completude do conte&#250;do. Essas observa&#231;&#245;es s&#227;o consistentes com estudos pr&#233;vios que relatam que modelos mais recentes de IA melhoram a coer&#234;ncia lingu&#237;stica e a organiza&#231;&#227;o estrutural,<sup><xref ref-type="bibr" rid="B11">11</xref></sup> mas ainda podem carecer de profundidade e especificidade suficientes ao abordar recomenda&#231;&#245;es cl&#237;nicas complexas.<sup><xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref></sup></p>
      <p>Para avaliar de forma sistem&#225;tica a qualidade do conte&#250;do m&#233;dico gerado por IA, aplicamos instrumentos de avalia&#231;&#227;o estabelecidos, incluindo o escore EQIP e escalas do tipo Likert. O escore EQIP, originalmente desenvolvido para avaliar materiais escritos de educa&#231;&#227;o do paciente,<sup><xref ref-type="bibr" rid="B18">18</xref></sup> forneceu um arcabou&#231;o estruturado para avaliar clareza, relev&#226;ncia e confiabilidade nas respostas geradas por IA.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> Embora o EQIP n&#227;o tenha sido formalmente validado para conte&#250;do gerado por IA, seu uso neste estudo destaca a crescente necessidade de estruturas padronizadas de avalia&#231;&#227;o adaptadas &#224; comunica&#231;&#227;o em sa&#250;de mediada por IA. As escalas de Likert complementaram a avalia&#231;&#227;o da acur&#225;cia e da completude percebidas, capturando julgamentos profissionais quanto &#224; confiabilidade das informa&#231;&#245;es geradas por IA.<sup><xref ref-type="bibr" rid="B12">12</xref></sup></p>
      <p>O ChatGPT-4.0 alcan&#231;ou um escore EQIP global ligeiramente superior ao do ChatGPT-Mini (21,4 &#177; 3,7 vs 19,5 &#177; 4,0), sugerindo melhorias modestas na qualidade do conte&#250;do educativo relacionado &#224; HAS. As diferen&#231;as mais pronunciadas foram observadas no dom&#237;nio de conte&#250;do (9,15 &#177; 2,32 vs 8,35 &#177; 2,32) e no dom&#237;nio de identifica&#231;&#227;o (3,80 &#177; 1,14 vs 2,85 &#177; 1,08), refletindo maior abrang&#234;ncia e transpar&#234;ncia. Ainda assim, essas diferen&#231;as n&#227;o foram estatisticamente significativas, indicando que aprimoramentos na apresenta&#231;&#227;o e na estrutura nem sempre se traduzem de forma consistente em conte&#250;do mais confi&#225;vel ou aderente &#224;s diretrizes.<sup><xref ref-type="bibr" rid="B19">19</xref></sup></p>
      <p>Ao examinar a propor&#231;&#227;o de respostas completamente corretas, o ChatGPT-4.0 superou o ChatGPT-Mini, produzindo respostas corretas (escore = 1) em aproximadamente 70%-80% das perguntas avaliadas, em compara&#231;&#227;o com menos de 50% no ChatGPT-Mini. Apesar dessa melhora, a ader&#234;ncia &#224;s diretrizes estabelecidas de HAS, incluindo as recomenda&#231;&#245;es ACC/AHA de 2017 e ESH de 2023, permaneceu inconsistente. Embora ambos os modelos geralmente se alinhassem aos princ&#237;pios centrais das diretrizes, as respostas frequentemente careciam de especificidade cl&#237;nica ou divergiam de padr&#245;es baseados em evid&#234;ncias. Esse padr&#227;o reflete achados de avalia&#231;&#245;es mais amplas de educa&#231;&#227;o do paciente gerada por IA, nas quais as taxas de acur&#225;cia relatadas variam de 71,7% a 94,3%, mas d&#233;ficits de profundidade e precis&#227;o cl&#237;nica permanecem comuns.<sup><xref ref-type="bibr" rid="B20">20</xref></sup> O ChatGPT-Mini apresentou uma propor&#231;&#227;o maior de respostas incorretas ou parcialmente corretas, particularmente em perguntas que exigiam explica&#231;&#245;es cl&#237;nicas detalhadas. O ChatGPT-4.0 reduziu a frequ&#234;ncia de respostas incompletas para abaixo de 20% na maioria das vari&#225;veis; contudo, 10%-30% das respostas ainda se enquadraram em categorias de menor qualidade, ressaltando lacunas informacionais persistentes.</p>
      <p>Esses achados s&#227;o consistentes com um estudo pr&#233;vio que avaliou o ChatGPT-4.0 em cinco condi&#231;&#245;es hepato-pancreato-biliares, no qual foi relatada uma mediana de escore EQIP de 16 (IIQ, 14,5-18), indicando defici&#234;ncias substanciais na acur&#225;cia das cita&#231;&#245;es e na coer&#234;ncia estrutural.<sup><xref ref-type="bibr" rid="B10">10</xref></sup> Embora nosso estudo tenha observado escores EQIP mais elevados para conte&#250;do relacionado &#224; HAS, ambas as investiga&#231;&#245;es destacam as limita&#231;&#245;es cont&#237;nuas dos modelos de IA em se alinhar plenamente &#224;s recomenda&#231;&#245;es m&#233;dicas baseadas em evid&#234;ncias. De forma importante, embora modelos mais recentes possam melhorar a legibilidade e a organiza&#231;&#227;o, sua acur&#225;cia factual n&#227;o parece ter avan&#231;ado de maneira proporcional.<sup><xref ref-type="bibr" rid="B21">21</xref></sup> Isso refor&#231;a a necessidade de processos rigorosos de valida&#231;&#227;o e de supervis&#227;o humana cont&#237;nua para garantir que melhorias na clareza n&#227;o comprometam a ader&#234;ncia &#224;s diretrizes cl&#237;nicas.<sup><xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B22">22</xref></sup></p>
      <p>Apesar dessas limita&#231;&#245;es, o ChatGPT-4.0 demonstra potencial como ferramenta suplementar para a educa&#231;&#227;o do paciente no manejo da HAS. Sua capacidade de gerar rapidamente explica&#231;&#245;es informadas por diretrizes pode reduzir a carga informacional dos profissionais de sa&#250;de, especialmente em discuss&#245;es iniciais sobre estrat&#233;gias terap&#234;uticas e interven&#231;&#245;es no estilo de vida.<sup><xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref></sup> Pesquisas anteriores sugerem que pacientes mais bem informados tendem a aderir melhor &#224;s terapias prescritas, o que pode melhorar os desfechos em condi&#231;&#245;es cr&#244;nicas como a HAS.<sup><xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref></sup> No entanto, o conte&#250;do gerado por IA, isoladamente, ainda pode n&#227;o fornecer orienta&#231;&#245;es suficientemente abrangentes ou acion&#225;veis, sobretudo diante do ceticismo p&#250;blico persistente em rela&#231;&#227;o &#224;s aplica&#231;&#245;es da IA na &#225;rea da sa&#250;de.<sup><xref ref-type="bibr" rid="B25">25</xref>-<xref ref-type="bibr" rid="B27">27</xref></sup></p>
      <p>Uma preocupa&#231;&#227;o cr&#237;tica no conte&#250;do m&#233;dico gerado por IA &#233; o fen&#244;meno das &#8220;alucina&#231;&#245;es&#8221;, no qual os modelos produzem informa&#231;&#245;es plaus&#237;veis, por&#233;m incorretas. Notavelmente, n&#227;o foram identificadas alucina&#231;&#245;es em nenhum dos modelos ao longo das 10 intera&#231;&#245;es repetidas neste estudo, o que sugere desempenho confi&#225;vel dentro deste dom&#237;nio cl&#237;nico espec&#237;fico. Contudo, estudos anteriores indicam que as alucina&#231;&#245;es permanecem um risco mais amplo na IA aplicada &#224; sa&#250;de, particularmente em t&#243;picos complexos ou menos bem definidos.<sup><xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref></sup> Esse risco, combinado &#224; conhecida sensibilidade ao desenho dos <italic>prompts</italic>, refor&#231;a a import&#226;ncia cont&#237;nua da supervis&#227;o cl&#237;nica em contextos cl&#237;nicos e educacionais.<sup><xref ref-type="bibr" rid="B30">30</xref></sup></p>
      <p>A IA tamb&#233;m vem influenciando de forma crescente o desenvolvimento de diretrizes cl&#237;nicas e a comunica&#231;&#227;o com pacientes, ao facilitar a gera&#231;&#227;o de perguntas cl&#237;nicas e apoiar intera&#231;&#245;es em tempo real. Sistemas baseados em IA t&#234;m demonstrado potencial para priorizar recomenda&#231;&#245;es cl&#237;nicas e otimizar a entrega de informa&#231;&#245;es.<sup><xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref></sup> No entanto, a ader&#234;ncia inconsistente aos padr&#245;es cl&#237;nicos permanece uma limita&#231;&#227;o importante, exigindo o refinamento cont&#237;nuo dessas tecnologias.<sup><xref ref-type="bibr" rid="B33">33</xref></sup></p>
      <p>A consist&#234;ncia das respostas foi ainda avaliada por meio das m&#233;tricas do BERTScore, que quantificam a similaridade sem&#226;ntica entre respostas repetidas. O ChatGPT-Mini demonstrou maior estabilidade das respostas, alcan&#231;ando maior precis&#227;o (0,9356 vs 0,8643) e <italic>recall</italic> (0,9558 vs 0,9330) em compara&#231;&#227;o com o ChatGPT-4.0. Em contraste, o ChatGPT-4.0 exibiu maior variabilidade, com escores F1 variando de 0,8851 a 0,9316. Esses achados sugerem que o ChatGPT-Mini pode ser mais adequado para materiais educativos estruturados que exigem alta consist&#234;ncia, enquanto a variabilidade do ChatGPT-4.0 pode favorecer o engajamento em contextos mais din&#226;micos e conversacionais. Padr&#245;es semelhantes foram relatados em estudos anteriores, nos quais modelos mais recentes priorizam flexibilidade em detrimento da consist&#234;ncia estrita.<sup><xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref></sup></p>
      <p>Por fim, a confiabilidade interavaliadores foi elevada em todos os dom&#237;nios avaliados, com valores de ICC superiores a 0,80 para os escores de EQIP, acur&#225;cia e completude. Essa forte concord&#226;ncia entre avaliadores sustenta a robustez da metodologia de avalia&#231;&#227;o e indica que as diferen&#231;as observadas refletem varia&#231;&#245;es reais no desempenho dos modelos, e n&#227;o subjetividade dos avaliadores.</p>
      <p>Em s&#237;ntese, este estudo contribui para a literatura crescente sobre o papel da IA na comunica&#231;&#227;o em sa&#250;de e na educa&#231;&#227;o do paciente. Embora os avan&#231;os nos modelos de IA tenham melhorado a clareza e a organiza&#231;&#227;o estrutural, garantir acur&#225;cia, consist&#234;ncia e ader&#234;ncia &#224;s diretrizes permanece essencial para sua integra&#231;&#227;o segura &#224; pr&#225;tica cl&#237;nica.<sup><xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B36">36</xref></sup> Nossos achados ressaltam a necessidade urgente de estruturas padronizadas e rigorosas de avalia&#231;&#227;o para conte&#250;do m&#233;dico gerado por IA. Abordagens emergentes, como modelos de avalia&#231;&#227;o estruturada que incluem o <italic>Conversational Reasoning Assessment Framework for Testing in Medicine</italic>, podem oferecer ferramentas valiosas para avaliar de forma sistem&#225;tica o desempenho da IA em contextos de sa&#250;de.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> Pesquisas futuras devem se concentrar em mecanismos de valida&#231;&#227;o em tempo real que conectem dinamicamente as respostas geradas por IA &#224;s diretrizes cl&#237;nicas atualizadas e promovam a colabora&#231;&#227;o interdisciplinar entre cl&#237;nicos, desenvolvedores de IA e autoridades regulat&#243;rias, a fim de estabelecer padr&#245;es robustos de &#233;tica, seguran&#231;a e qualidade para o uso da IA na &#225;rea da sa&#250;de.<sup><xref ref-type="bibr" rid="B25">25</xref></sup></p>
      <sec>
        <title>Limita&#231;&#245;es do estudo</title>
        <p>Embora este estudo forne&#231;a subs&#237;dios relevantes sobre o desempenho do ChatGPT-Mini e do ChatGPT-4.0 na educa&#231;&#227;o do paciente, algumas limita&#231;&#245;es devem ser reconhecidas. Primeiro, a an&#225;lise foi restrita a um &#250;nico dom&#237;nio cl&#237;nico, a HAS, o que limita a generaliza&#231;&#227;o dos achados para outras condi&#231;&#245;es m&#233;dicas. Como o desempenho da IA pode variar entre especialidades, investiga&#231;&#245;es futuras devem avaliar conte&#250;dos gerados por IA em uma gama mais ampla de t&#243;picos cl&#237;nicos.</p>
        <p>Segundo, apesar do uso de ferramentas estruturadas de avalia&#231;&#227;o, como o EQIP e o BERTScore, a an&#225;lise baseou-se em julgamento humano especializado, o que &#233; inerentemente subjetivo, mesmo com crit&#233;rios padronizados de pontua&#231;&#227;o. A variabilidade na interpreta&#231;&#227;o dos especialistas pode ter influenciado os escores, ressaltando a necessidade de abordagens complementares de valida&#231;&#227;o automatizada e baseadas em consenso em pesquisas futuras.</p>
        <p>Terceiro, a avalia&#231;&#227;o foi limitada a respostas em l&#237;ngua inglesa, restringindo a aplicabilidade dos resultados a popula&#231;&#245;es n&#227;o falantes de ingl&#234;s. Como estudos pr&#233;vios demonstraram variabilidade no desempenho da IA entre diferentes idiomas, s&#227;o necess&#225;rias pesquisas adicionais para avaliar diferen&#231;as lingu&#237;sticas no conte&#250;do m&#233;dico gerado por IA e para garantir acesso equitativo &#224; educa&#231;&#227;o do paciente precisa em escala global.</p>
        <p>Al&#233;m disso, os modelos de IA foram avaliados em um ambiente de teste controlado e est&#225;tico e n&#227;o foram expostos a intera&#231;&#245;es reais com pacientes ou a <italic>feedback</italic> iterativo. O conte&#250;do m&#233;dico gerado por IA &#233; inerentemente din&#226;mico, sendo influenciado por atualiza&#231;&#245;es do modelo e pela evolu&#231;&#227;o das diretrizes cl&#237;nicas. Estudos futuros devem incorporar estrat&#233;gias de valida&#231;&#227;o em tempo real que comparem continuamente as sa&#237;das da IA com os padr&#245;es cl&#237;nicos atuais, a fim de manter acur&#225;cia e relev&#226;ncia.</p>
        <p>Por fim, este estudo concentrou-se exclusivamente no ChatGPT-Mini e no ChatGPT-4.0. Outros grandes modelos de linguagem, com arquiteturas e paradigmas de treinamento distintos (p.ex., Claude, Gemini e Med-PaLM), podem apresentar caracter&#237;sticas de desempenho diferentes. Avalia&#231;&#245;es comparativas envolvendo m&#250;ltiplos sistemas de IA s&#227;o, portanto, necess&#225;rias para definir de forma mais abrangente o papel da IA na educa&#231;&#227;o do paciente.</p>
      </sec>
    </sec>
    <sec sec-type="conclusions">
      <title>Conclus&#227;o</title>
      <p>Este estudo demonstra que, embora o ChatGPT-4.0 ofere&#231;a melhor coer&#234;ncia estrutural em compara&#231;&#227;o com o ChatGPT-Mini, ele n&#227;o produz ganhos estatisticamente significativos em completude factual ou consist&#234;ncia das respostas. Esses achados destacam a import&#226;ncia da implementa&#231;&#227;o de estruturas padronizadas de avalia&#231;&#227;o, mecanismos de valida&#231;&#227;o em tempo real e supervis&#227;o humana cont&#237;nua para apoiar o uso seguro e eficaz da IA na educa&#231;&#227;o do paciente.</p>
      <p>Pesquisas futuras devem enfatizar estrat&#233;gias din&#226;micas de valida&#231;&#227;o e o desenvolvimento de padr&#245;es de avalia&#231;&#227;o universalmente aceitos, a fim de garantir que os sistemas de IA m&#233;dica permane&#231;am clinicamente confi&#225;veis e eticamente respons&#225;veis. Embora as tecnologias de IA tenham potencial para ampliar a acessibilidade aos cuidados de sa&#250;de, a supervis&#227;o humana permanece essencial para prevenir desinforma&#231;&#227;o e preservar a confian&#231;a dos pacientes. Em &#250;ltima an&#225;lise, a integra&#231;&#227;o respons&#225;vel da IA na pr&#225;tica m&#233;dica depender&#225; de colabora&#231;&#227;o interdisciplinar cont&#237;nua para manter padr&#245;es rigorosos de acur&#225;cia, seguran&#231;a e integridade &#233;tica.</p>
    </sec>
    <sec id="suppl" sec-type="supplementary-material">
      <title>Material suplementar</title>
      <supplementary-material id="suppl01" mime-subtype="pdf" mimetype="application">
        <label>Supplemental Content 1</label>
        <media mime-subtype="pdf" mimetype="application" xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/cf0dad04c1dcc423600ea13700e5be97ebec7c4f.pdf"/>
      </supplementary-material>
      <supplementary-material id="suppl02" mime-subtype="pdf" mimetype="application">
        <label>Supplemental Content 2</label>
        <media mime-subtype="pdf" mimetype="application" xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/294918e6c92161ac3dbbf06bd155d73431bfe8a1.pdf"/>
      </supplementary-material>
    </sec>
  </body>
  <back>
    <ref-list>
      <title>Refer&#234;ncias</title>
      <ref id="B1">
        <label>1</label>
        <mixed-citation>1. Gala D, Behl H, Shah M, Makaryus AN. The Role of Artificial Intelligence in Improving Patient Outcomes and Future of Healthcare Delivery in Cardiology: A Narrative Review of the Literature. Healthcare. 2024;12(4):481. doi: 10.3390/healthcare12040481.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Gala</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Behl</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Shah</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Makaryus</surname>
              <given-names>AN</given-names>
            </name>
          </person-group>
          <article-title>The Role of Artificial Intelligence in Improving Patient Outcomes and Future of Healthcare Delivery in Cardiology: A Narrative Review of the Literature</article-title>
          <source>Healthcare</source>
          <year>2024</year>
          <volume>12</volume>
          <issue>4</issue>
          <size units="pages">481</size>
          <pub-id pub-id-type="doi">10.3390/healthcare12040481</pub-id>
        </element-citation>
      </ref>
      <ref id="B2">
        <label>2</label>
        <mixed-citation>2. Visco V, Izzo C, Mancusi C, Rispoli A, Tedeschi M, Virtuoso N, et al. Artificial Intelligence in Hypertension Management: An Ace up Your Sleeve. J Cardiovasc Dev Dis. 2023;10(2):74. doi: 10.3390/jcdd10020074.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Visco</surname>
              <given-names>V</given-names>
            </name>
            <name>
              <surname>Izzo</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Mancusi</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Rispoli</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Tedeschi</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Virtuoso</surname>
              <given-names>N</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Artificial Intelligence in Hypertension Management: An Ace up Your Sleeve</article-title>
          <source>J Cardiovasc Dev Dis</source>
          <year>2023</year>
          <volume>10</volume>
          <issue>2</issue>
          <size units="pages">74</size>
          <pub-id pub-id-type="doi">10.3390/jcdd10020074</pub-id>
        </element-citation>
      </ref>
      <ref id="B3">
        <label>3</label>
        <mixed-citation>3. Vinufrancis A, Al Hussein H, Patel HV, Nizami A, Singh A, Nunez B, et al. Assessing the Quality and Reliability of AI-Generated Responses to Common Hypertension Queries. Cureus. 2024;16(8):e66041. doi: 10.7759/cureus.66041.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Vinufrancis</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Al Hussein</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Patel</surname>
              <given-names>HV</given-names>
            </name>
            <name>
              <surname>Nizami</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Singh</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Nunez</surname>
              <given-names>B</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Assessing the Quality and Reliability of AI-Generated Responses to Common Hypertension Queries</article-title>
          <source>Cureus</source>
          <year>2024</year>
          <volume>16</volume>
          <issue>8</issue>
          <elocation-id>e66041</elocation-id>
          <pub-id pub-id-type="doi">10.7759/cureus.66041</pub-id>
        </element-citation>
      </ref>
      <ref id="B4">
        <label>4</label>
        <mixed-citation>4. Jo E, Song S, Kim JH, Lim S, Kim JH, Cha JJ, et al. Assessing GPT-4's Performance in Delivering Medical Advice: Comparative Analysis with Human Experts. JMIR Med Educ. 2024;10:e51282. doi: 10.2196/51282.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Jo</surname>
              <given-names>E</given-names>
            </name>
            <name>
              <surname>Song</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Kim</surname>
              <given-names>JH</given-names>
            </name>
            <name>
              <surname>Lim</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Kim</surname>
              <given-names>JH</given-names>
            </name>
            <name>
              <surname>Cha</surname>
              <given-names>JJ</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Assessing GPT-4's Performance in Delivering Medical Advice: Comparative Analysis with Human Experts</article-title>
          <source>JMIR Med Educ</source>
          <year>2024</year>
          <volume>10</volume>
          <elocation-id>e51282</elocation-id>
          <pub-id pub-id-type="doi">10.2196/51282</pub-id>
        </element-citation>
      </ref>
      <ref id="B5">
        <label>5</label>
        <mixed-citation>5. Griffin AC, Khairat S, Bailey SC, Chung AE. A Chatbot for Hypertension Self-Management Support: User-Centered Design, Development, and Usability Testing. JAMIA Open. 2023;6(3):ooad073. doi: 10.1093/jamiaopen/ooad073.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Griffin</surname>
              <given-names>AC</given-names>
            </name>
            <name>
              <surname>Khairat</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Bailey</surname>
              <given-names>SC</given-names>
            </name>
            <name>
              <surname>Chung</surname>
              <given-names>AE</given-names>
            </name>
          </person-group>
          <article-title>A Chatbot for Hypertension Self-Management Support: User-Centered Design, Development, and Usability Testing</article-title>
          <source>JAMIA Open</source>
          <year>2023</year>
          <volume>6</volume>
          <issue>3</issue>
          <size units="pages">ooad073</size>
          <pub-id pub-id-type="doi">10.1093/jamiaopen/ooad073</pub-id>
        </element-citation>
      </ref>
      <ref id="B6">
        <label>6</label>
        <mixed-citation>6. Almagazzachi A, Mustafa A, Sedeh AE, Gonzalez AEV, Polianovskaia A, Abood M, et al. Generative Artificial Intelligence in Patient Education: ChatGPT Takes on Hypertension Questions. Cureus. 2024;16(2):e53441. doi: 10.7759/cureus.53441.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Almagazzachi</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Mustafa</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Sedeh</surname>
              <given-names>AE</given-names>
            </name>
            <name>
              <surname>Gonzalez</surname>
              <given-names>AEV</given-names>
            </name>
            <name>
              <surname>Polianovskaia</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Abood</surname>
              <given-names>M</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Generative Artificial Intelligence in Patient Education: ChatGPT Takes on Hypertension Questions</article-title>
          <source>Cureus</source>
          <year>2024</year>
          <volume>16</volume>
          <issue>2</issue>
          <elocation-id>e53441</elocation-id>
          <pub-id pub-id-type="doi">10.7759/cureus.53441</pub-id>
        </element-citation>
      </ref>
      <ref id="B7">
        <label>7</label>
        <mixed-citation>7. Ioannidis A, Tsounis D, Bouras G, Komninou D, Pechlevanis A, Zalokosta E, et al. Accuracy of Artificial Intelligence Chatbots' Replies in Greek vs. English and in Accordance with the 2023 ESH Guidelines for the Management of Arterial Hypertension. J Hypertens. 2024;42(Suppl 1):e3. doi: 10.1093/eurjpc/zwae175.101.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ioannidis</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Tsounis</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Bouras</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Komninou</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Pechlevanis</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Zalokosta</surname>
              <given-names>E</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Accuracy of Artificial Intelligence Chatbots' Replies in Greek vs. English and in Accordance with the 2023 ESH Guidelines for the Management of Arterial Hypertension</article-title>
          <source>J Hypertens</source>
          <year>2024</year>
          <volume>42</volume>
          <supplement>Suppl 1</supplement>
          <elocation-id>e3</elocation-id>
          <pub-id pub-id-type="doi">10.1093/eurjpc/zwae175.101</pub-id>
        </element-citation>
      </ref>
      <ref id="B8">
        <label>8</label>
        <mixed-citation>8. Yelne S, Chaudhary M, Dod K, Sayyad A, Sharma R. Harnessing the Power of AI: A Comprehensive Review of Its Impact and Challenges in Nursing Science and Healthcare. Cureus. 2023;15(11):e49252. doi: 10.7759/cureus.49252.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Yelne</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Chaudhary</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Dod</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Sayyad</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Sharma</surname>
              <given-names>R</given-names>
            </name>
          </person-group>
          <article-title>Harnessing the Power of AI: A Comprehensive Review of Its Impact and Challenges in Nursing Science and Healthcare</article-title>
          <source>Cureus</source>
          <year>2023</year>
          <volume>15</volume>
          <issue>11</issue>
          <elocation-id>e49252</elocation-id>
          <pub-id pub-id-type="doi">10.7759/cureus.49252</pub-id>
        </element-citation>
      </ref>
      <ref id="B9">
        <label>9</label>
        <mixed-citation>9. Cappellani F, Card KR, Shields CL, Pulido JS, Haller JA. Reliability and Accuracy of Artificial Intelligence ChatGPT in Providing Information on Ophthalmic Diseases and Management to Patients. Eye. 2024;38(7):1368-1373. doi: 10.1038/s41433-023-02906-0.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Cappellani</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Card</surname>
              <given-names>KR</given-names>
            </name>
            <name>
              <surname>Shields</surname>
              <given-names>CL</given-names>
            </name>
            <name>
              <surname>Pulido</surname>
              <given-names>JS</given-names>
            </name>
            <name>
              <surname>Haller</surname>
              <given-names>JA</given-names>
            </name>
          </person-group>
          <article-title>Reliability and Accuracy of Artificial Intelligence ChatGPT in Providing Information on Ophthalmic Diseases and Management to Patients</article-title>
          <source>Eye</source>
          <year>2024</year>
          <volume>38</volume>
          <issue>7</issue>
          <fpage>1368</fpage>
          <lpage>1373</lpage>
          <pub-id pub-id-type="doi">10.1038/s41433-023-02906-0</pub-id>
        </element-citation>
      </ref>
      <ref id="B10">
        <label>10</label>
        <mixed-citation>10. Walker HL, Ghani S, Kuemmerli C, Nebiker CA, M&#252;ller BP, Raptis DA, et al. Reliability of Medical Information Provided by ChatGPT: Assessment Against Clinical Guidelines and Patient Information Quality Instrument. J Med Internet Res. 2023;25:e47479. doi: 10.2196/47479.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Walker</surname>
              <given-names>HL</given-names>
            </name>
            <name>
              <surname>Ghani</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Kuemmerli</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Nebiker</surname>
              <given-names>CA</given-names>
            </name>
            <name>
              <surname>M&#252;ller</surname>
              <given-names>BP</given-names>
            </name>
            <name>
              <surname>Raptis</surname>
              <given-names>DA</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Reliability of Medical Information Provided by ChatGPT: Assessment Against Clinical Guidelines and Patient Information Quality Instrument</article-title>
          <source>J Med Internet Res</source>
          <year>2023</year>
          <volume>25</volume>
          <elocation-id>e47479</elocation-id>
          <pub-id pub-id-type="doi">10.2196/47479</pub-id>
        </element-citation>
      </ref>
      <ref id="B11">
        <label>11</label>
        <mixed-citation>11. Goodman RS, Patrinely JR, Stone CA Jr, Zimmerman E, Donald RR, Chang SS, et al. Accuracy and Reliability of Chatbot Responses to Physician Questions. JAMA Netw Open. 2023;6(10):e2336483. doi: 10.1001/jamanetworkopen.2023.36483.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Goodman</surname>
              <given-names>RS</given-names>
            </name>
            <name>
              <surname>Patrinely</surname>
              <given-names>JR</given-names>
            </name>
            <name>
              <surname>Stone</surname>
              <given-names>CA</given-names>
              <suffix>Jr</suffix>
            </name>
            <name>
              <surname>Zimmerman</surname>
              <given-names>E</given-names>
            </name>
            <name>
              <surname>Donald</surname>
              <given-names>RR</given-names>
            </name>
            <name>
              <surname>Chang</surname>
              <given-names>SS</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Accuracy and Reliability of Chatbot Responses to Physician Questions</article-title>
          <source>JAMA Netw Open</source>
          <year>2023</year>
          <volume>6</volume>
          <issue>10</issue>
          <elocation-id>e2336483</elocation-id>
          <pub-id pub-id-type="doi">10.1001/jamanetworkopen.2023.36483</pub-id>
        </element-citation>
      </ref>
      <ref id="B12">
        <label>12</label>
        <mixed-citation>12. Xu X, Chen Y, Miao J. Opportunities, Challenges, and Future Directions of Large Language Models, Including ChatGPT in Medical Education: A Systematic Scoping Review. J Educ Eval Health Prof. 2024;21:6. doi: 10.3352/jeehp.2024.21.6.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Xu</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Chen</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Miao</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <article-title>Opportunities, Challenges, and Future Directions of Large Language Models, Including ChatGPT in Medical Education: A Systematic Scoping Review</article-title>
          <source>J Educ Eval Health Prof</source>
          <year>2024</year>
          <volume>21</volume>
          <size units="pages">6</size>
          <pub-id pub-id-type="doi">10.3352/jeehp.2024.21.6</pub-id>
        </element-citation>
      </ref>
      <ref id="B13">
        <label>13</label>
        <mixed-citation>13. Gritti MN, AlTurki H, Farid P, Morgan CT. Progression of an Artificial Intelligence Chatbot (ChatGPT) for Pediatric Cardiology Educational Knowledge Assessment. Pediatr Cardiol. 2024;45(2):309-13. doi: 10.1007/s00246-023-03385-6.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Gritti</surname>
              <given-names>MN</given-names>
            </name>
            <name>
              <surname>AlTurki</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Farid</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Morgan</surname>
              <given-names>CT</given-names>
            </name>
          </person-group>
          <article-title>Progression of an Artificial Intelligence Chatbot (ChatGPT) for Pediatric Cardiology Educational Knowledge Assessment</article-title>
          <source>Pediatr Cardiol</source>
          <year>2024</year>
          <volume>45</volume>
          <issue>2</issue>
          <fpage>309</fpage>
          <lpage>313</lpage>
          <pub-id pub-id-type="doi">10.1007/s00246-023-03385-6</pub-id>
        </element-citation>
      </ref>
      <ref id="B14">
        <label>14</label>
        <mixed-citation>14. Moult B, Franck LS, Brady H. Ensuring Quality Information for Patients: Development and Preliminary Validation of a New Instrument to Improve the Quality of Written Health Care Information. Health Expect. 2004;7(2):165-75. doi: 10.1111/j.1369-7625.2004.00273.x.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Moult</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Franck</surname>
              <given-names>LS</given-names>
            </name>
            <name>
              <surname>Brady</surname>
              <given-names>H</given-names>
            </name>
          </person-group>
          <article-title>Ensuring Quality Information for Patients: Development and Preliminary Validation of a New Instrument to Improve the Quality of Written Health Care Information</article-title>
          <source>Health Expect</source>
          <year>2004</year>
          <volume>7</volume>
          <issue>2</issue>
          <fpage>165</fpage>
          <lpage>175</lpage>
          <pub-id pub-id-type="doi">10.1111/j.1369-7625.2004.00273.x</pub-id>
        </element-citation>
      </ref>
      <ref id="B15">
        <label>15</label>
        <mixed-citation>15. Zhang T, Kishore V, Wu F, Weinberger KQ, Artzi Y. BERTScore: Evaluating Text Generation with BERT [Internet]. New york: arXiv; 2020 [cited 2025 Feb 08]. Available from: <ext-link ext-link-type="uri" xlink:href="http://arxiv.org/abs/1904.09675">http://arxiv.org/abs/1904.09675</ext-link>.</mixed-citation>
        <element-citation publication-type="webpage">
          <person-group person-group-type="author">
            <name>
              <surname>Zhang</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Kishore</surname>
              <given-names>V</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Weinberger</surname>
              <given-names>KQ</given-names>
            </name>
            <name>
              <surname>Artzi</surname>
              <given-names>Y</given-names>
            </name>
          </person-group>
          <source>BERTScore: Evaluating Text Generation with BERT</source>
          <comment>Internet</comment>
          <publisher-loc>New york</publisher-loc>
          <publisher-name>arXiv</publisher-name>
          <year>2020</year>
          <date-in-citation content-type="cited-date">cited 2025 Feb 08</date-in-citation>
          <ext-link ext-link-type="uri" xlink:href="http://arxiv.org/abs/1904.09675">http://arxiv.org/abs/1904.09675</ext-link>
        </element-citation>
      </ref>
      <ref id="B16">
        <label>16</label>
        <mixed-citation>16. Gibson D, Jackson S, Shanmugasundaram R, Seth I, Siu A, Ahmadi N, et al. Evaluating the Efficacy of ChatGPT as a Patient Education Tool in Prostate Cancer: Multimetric Assessment. J Med Internet Res. 2024;26:e55939. doi: 10.2196/55939.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Gibson</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Jackson</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Shanmugasundaram</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Seth</surname>
              <given-names>I</given-names>
            </name>
            <name>
              <surname>Siu</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Ahmadi</surname>
              <given-names>N</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Evaluating the Efficacy of ChatGPT as a Patient Education Tool in Prostate Cancer: Multimetric Assessment</article-title>
          <source>J Med Internet Res</source>
          <year>2024</year>
          <volume>26</volume>
          <elocation-id>e55939</elocation-id>
          <pub-id pub-id-type="doi">10.2196/55939</pub-id>
        </element-citation>
      </ref>
      <ref id="B17">
        <label>17</label>
        <mixed-citation>17. Yang R, Nair SV, Ke Y, D'Agostino D, Liu M, Ning Y, et al. Disparities in Clinical Studies of AI Enabled Applications from a Global Perspective. NPJ Digit Med. 2024;7(1):209. doi: 10.1038/s41746-024-01212-7.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Yang</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Nair</surname>
              <given-names>SV</given-names>
            </name>
            <name>
              <surname>Ke</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>D'Agostino</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Liu</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Ning</surname>
              <given-names>Y</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Disparities in Clinical Studies of AI Enabled Applications from a Global Perspective</article-title>
          <source>NPJ Digit Med</source>
          <year>2024</year>
          <volume>7</volume>
          <issue>1</issue>
          <size units="pages">209</size>
          <pub-id pub-id-type="doi">10.1038/s41746-024-01212-7</pub-id>
        </element-citation>
      </ref>
      <ref id="B18">
        <label>18</label>
        <mixed-citation>18. Wei K, Fritz C, Rajasekaran K. Answering Head and Neck Cancer Questions: An Assessment of ChatGPT Responses. Am J Otolaryngol. 2024;45(1):104085. doi: 10.1016/j.amjoto.2023.104085.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Wei</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Fritz</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Rajasekaran</surname>
              <given-names>K</given-names>
            </name>
          </person-group>
          <article-title>Answering Head and Neck Cancer Questions: An Assessment of ChatGPT Responses</article-title>
          <source>Am J Otolaryngol</source>
          <year>2024</year>
          <volume>45</volume>
          <issue>1</issue>
          <size units="pages">104085</size>
          <pub-id pub-id-type="doi">10.1016/j.amjoto.2023.104085</pub-id>
        </element-citation>
      </ref>
      <ref id="B19">
        <label>19</label>
        <mixed-citation>19. Lewin S, Chetty R, Ihdayhid AR, Dwivedi G. Ethical Challenges and Opportunities in Applying Artificial Intelligence to Cardiovascular Medicine. Can J Cardiol. 2024;40(10):1897-906. doi: 10.1016/j.cjca.2024.06.029.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Lewin</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Chetty</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Ihdayhid</surname>
              <given-names>AR</given-names>
            </name>
            <name>
              <surname>Dwivedi</surname>
              <given-names>G</given-names>
            </name>
          </person-group>
          <article-title>Ethical Challenges and Opportunities in Applying Artificial Intelligence to Cardiovascular Medicine</article-title>
          <source>Can J Cardiol</source>
          <year>2024</year>
          <volume>40</volume>
          <issue>10</issue>
          <fpage>1897</fpage>
          <lpage>1906</lpage>
          <pub-id pub-id-type="doi">10.1016/j.cjca.2024.06.029</pub-id>
        </element-citation>
      </ref>
      <ref id="B20">
        <label>20</label>
        <mixed-citation>20. Ayers JW, Poliak A, Dredze M, Leas EC, Zhu Z, Kelley JB, et al. Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum. JAMA Intern Med. 2023;183(6):589-96. doi: 10.1001/jamainternmed.2023.1838.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ayers</surname>
              <given-names>JW</given-names>
            </name>
            <name>
              <surname>Poliak</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Dredze</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Leas</surname>
              <given-names>EC</given-names>
            </name>
            <name>
              <surname>Zhu</surname>
              <given-names>Z</given-names>
            </name>
            <name>
              <surname>Kelley</surname>
              <given-names>JB</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum</article-title>
          <source>JAMA Intern Med</source>
          <year>2023</year>
          <volume>183</volume>
          <issue>6</issue>
          <fpage>589</fpage>
          <lpage>596</lpage>
          <pub-id pub-id-type="doi">10.1001/jamainternmed.2023.1838</pub-id>
        </element-citation>
      </ref>
      <ref id="B21">
        <label>21</label>
        <mixed-citation>21. Stanceski K, Zhong S, Zhang X, Khadra S, Tracy M, Koria L, et al. The Quality and Safety of Using Generative AI to Produce Patient-Centred Discharge Instructions. NPJ Digit Med. 2024;7(1):329. doi: 10.1038/s41746-024-01336-w.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Stanceski</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Zhong</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Khadra</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Tracy</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Koria</surname>
              <given-names>L</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>The Quality and Safety of Using Generative AI to Produce Patient-Centred Discharge Instructions</article-title>
          <source>NPJ Digit Med</source>
          <year>2024</year>
          <volume>7</volume>
          <issue>1</issue>
          <size units="pages">329</size>
          <pub-id pub-id-type="doi">10.1038/s41746-024-01336-w</pub-id>
        </element-citation>
      </ref>
      <ref id="B22">
        <label>22</label>
        <mixed-citation>22. Armoundas AA, Narayan SM, Arnett DK, Spector-Bagdady K, Bennett DA, Celi LA, et al. Use of Artificial Intelligence in Improving Outcomes in Heart Disease: A Scientific Statement from the American Heart Association. Circulation. 2024;149(14):e1028-50. doi: 10.1161/CIR.0000000000001201.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Armoundas</surname>
              <given-names>AA</given-names>
            </name>
            <name>
              <surname>Narayan</surname>
              <given-names>SM</given-names>
            </name>
            <name>
              <surname>Arnett</surname>
              <given-names>DK</given-names>
            </name>
            <name>
              <surname>Spector-Bagdady</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Bennett</surname>
              <given-names>DA</given-names>
            </name>
            <name>
              <surname>Celi</surname>
              <given-names>LA</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Use of Artificial Intelligence in Improving Outcomes in Heart Disease: A Scientific Statement from the American Heart Association</article-title>
          <source>Circulation</source>
          <year>2024</year>
          <volume>149</volume>
          <issue>14</issue>
          <fpage>e1028</fpage>
          <lpage>e1050</lpage>
          <pub-id pub-id-type="doi">10.1161/CIR.0000000000001201</pub-id>
        </element-citation>
      </ref>
      <ref id="B23">
        <label>23</label>
        <mixed-citation>23. Regmi K, Jones L. A Systematic Review of the Factors - Enablers and Barriers - Affecting E-Learning in Health Sciences Education. BMC Med Educ. 2020;20(1):91. doi: 10.1186/s12909-020-02007-6.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Regmi</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Jones</surname>
              <given-names>L</given-names>
            </name>
          </person-group>
          <article-title>A Systematic Review of the Factors - Enablers and Barriers - Affecting E-Learning in Health Sciences Education</article-title>
          <source>BMC Med Educ</source>
          <year>2020</year>
          <volume>20</volume>
          <issue>1</issue>
          <size units="pages">91</size>
          <pub-id pub-id-type="doi">10.1186/s12909-020-02007-6</pub-id>
        </element-citation>
      </ref>
      <ref id="B24">
        <label>24</label>
        <mixed-citation>24. Brown MT, Bussell JK. Medication Adherence: WHO Cares? Mayo Clin Proc. 2011;86(4):304-14. doi: 10.4065/mcp.2010.0575.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Brown</surname>
              <given-names>MT</given-names>
            </name>
            <name>
              <surname>Bussell</surname>
              <given-names>JK</given-names>
            </name>
          </person-group>
          <article-title>Medication Adherence: WHO Cares?</article-title>
          <source>Mayo Clin Proc</source>
          <year>2011</year>
          <volume>86</volume>
          <issue>4</issue>
          <fpage>304</fpage>
          <lpage>314</lpage>
          <pub-id pub-id-type="doi">10.4065/mcp.2010.0575</pub-id>
        </element-citation>
      </ref>
      <ref id="B25">
        <label>25</label>
        <mixed-citation>25. Tam TYC, Sivarajkumar S, Kapoor S, Stolyar AV, Polanska K, McCarthy KR, et al. A Framework for Human Evaluation of Large Language Models in Healthcare Derived from Literature Review. NPJ Digit Med. 2024;7(1):258. doi: 10.1038/s41746-024-01258-7.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Tam</surname>
              <given-names>TYC</given-names>
            </name>
            <name>
              <surname>Sivarajkumar</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Kapoor</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Stolyar</surname>
              <given-names>AV</given-names>
            </name>
            <name>
              <surname>Polanska</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>McCarthy</surname>
              <given-names>KR</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>A Framework for Human Evaluation of Large Language Models in Healthcare Derived from Literature Review</article-title>
          <source>NPJ Digit Med</source>
          <year>2024</year>
          <volume>7</volume>
          <issue>1</issue>
          <size units="pages">258</size>
          <pub-id pub-id-type="doi">10.1038/s41746-024-01258-7</pub-id>
        </element-citation>
      </ref>
      <ref id="B26">
        <label>26</label>
        <mixed-citation>26. Jasinska-Stroschein M, Dymek J, Drozd M, Sierpniowska O, Jedra A, Stankiewicz A, et al. An Evaluation of Written Materials for Supporting Hypertensive Patient Education and Counselling When Performing a New Medicine Service in Poland. BMC Med Educ. 2024;24(1):521. doi: 10.1186/s12909-024-05523-x.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Jasinska-Stroschein</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Dymek</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Drozd</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Sierpniowska</surname>
              <given-names>O</given-names>
            </name>
            <name>
              <surname>Jedra</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Stankiewicz</surname>
              <given-names>A</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>An Evaluation of Written Materials for Supporting Hypertensive Patient Education and Counselling When Performing a New Medicine Service in Poland</article-title>
          <source>BMC Med Educ</source>
          <year>2024</year>
          <volume>24</volume>
          <issue>1</issue>
          <size units="pages">521</size>
          <pub-id pub-id-type="doi">10.1186/s12909-024-05523-x</pub-id>
        </element-citation>
      </ref>
      <ref id="B27">
        <label>27</label>
        <mixed-citation>27. Akinrinmade AO, Adebile TM, Ezuma-Ebong C, Bolaji K, Ajufo A, Adigun AO, et al. Artificial Intelligence in Healthcare: Perception and Reality. Cureus. 2023;15(9):e45594. doi: 10.7759/cureus.45594.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Akinrinmade</surname>
              <given-names>AO</given-names>
            </name>
            <name>
              <surname>Adebile</surname>
              <given-names>TM</given-names>
            </name>
            <name>
              <surname>Ezuma-Ebong</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Bolaji</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Ajufo</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Adigun</surname>
              <given-names>AO</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Artificial Intelligence in Healthcare: Perception and Reality</article-title>
          <source>Cureus</source>
          <year>2023</year>
          <volume>15</volume>
          <issue>9</issue>
          <elocation-id>e45594</elocation-id>
          <pub-id pub-id-type="doi">10.7759/cureus.45594</pub-id>
        </element-citation>
      </ref>
      <ref id="B28">
        <label>28</label>
        <mixed-citation>28. Colasacco CJ, Born HL. A Case of Artificial Intelligence Chatbot Hallucination. JAMA Otolaryngol Head Neck Surg. 2024;150(6):457-8. doi: 10.1001/jamaoto.2024.0428.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Colasacco</surname>
              <given-names>CJ</given-names>
            </name>
            <name>
              <surname>Born</surname>
              <given-names>HL</given-names>
            </name>
          </person-group>
          <article-title>A Case of Artificial Intelligence Chatbot Hallucination</article-title>
          <source>JAMA Otolaryngol Head Neck Surg</source>
          <year>2024</year>
          <volume>150</volume>
          <issue>6</issue>
          <fpage>457</fpage>
          <lpage>458</lpage>
          <pub-id pub-id-type="doi">10.1001/jamaoto.2024.0428</pub-id>
        </element-citation>
      </ref>
      <ref id="B29">
        <label>29</label>
        <mixed-citation>29. Gondode P, Duggal S, Mahor V. Artificial Intelligence Hallucinations in Anaesthesia: Causes, Consequences and Countermeasures. Indian J Anaesth. 2024;68(7):658-61. doi: 10.4103/ija.ija_203_24.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Gondode</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Duggal</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Mahor</surname>
              <given-names>V</given-names>
            </name>
          </person-group>
          <article-title>Artificial Intelligence Hallucinations in Anaesthesia: Causes, Consequences and Countermeasures</article-title>
          <source>Indian J Anaesth</source>
          <year>2024</year>
          <volume>68</volume>
          <issue>7</issue>
          <fpage>658</fpage>
          <lpage>661</lpage>
          <pub-id pub-id-type="doi">10.4103/ija.ija_203_24</pub-id>
        </element-citation>
      </ref>
      <ref id="B30">
        <label>30</label>
        <mixed-citation>30. Hatem R, Simmons B, Thornton JE. A Call to Address AI "Hallucinations" and How Healthcare Professionals Can Mitigate Their Risks. Cureus. 2023;15(9):e44720. doi: 10.7759/cureus.44720.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Hatem</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Simmons</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Thornton</surname>
              <given-names>JE</given-names>
            </name>
          </person-group>
          <article-title>A Call to Address AI "Hallucinations" and How Healthcare Professionals Can Mitigate Their Risks</article-title>
          <source>Cureus</source>
          <year>2023</year>
          <volume>15</volume>
          <issue>9</issue>
          <elocation-id>e44720</elocation-id>
          <pub-id pub-id-type="doi">10.7759/cureus.44720</pub-id>
        </element-citation>
      </ref>
      <ref id="B31">
        <label>31</label>
        <mixed-citation>31. Sousa-Pinto B, Vieira RJ, Marques-Cruz M, Bognanni A, Gil-Mata S, Jankin S, et al. Artificial Intelligence-Supported Development of Health Guideline Questions. Ann Intern Med. 2024;177(11):1518-29. doi: 10.7326/ANNALS-24-00363.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Sousa-Pinto</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Vieira</surname>
              <given-names>RJ</given-names>
            </name>
            <name>
              <surname>Marques-Cruz</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Bognanni</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Gil-Mata</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Jankin</surname>
              <given-names>S</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Artificial Intelligence-Supported Development of Health Guideline Questions</article-title>
          <source>Ann Intern Med</source>
          <year>2024</year>
          <volume>177</volume>
          <issue>11</issue>
          <fpage>1518</fpage>
          <lpage>1529</lpage>
          <pub-id pub-id-type="doi">10.7326/ANNALS-24-00363</pub-id>
        </element-citation>
      </ref>
      <ref id="B32">
        <label>32</label>
        <mixed-citation>32. Sun G, Zhou YH. AI in Healthcare: Navigating Opportunities and Challenges in Digital Communication. Front Digit Health. 2023;5:1291132. doi: 10.3389/fdgth.2023.1291132.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Sun</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Zhou</surname>
              <given-names>YH</given-names>
            </name>
          </person-group>
          <article-title>AI in Healthcare: Navigating Opportunities and Challenges in Digital Communication</article-title>
          <source>Front Digit Health</source>
          <year>2023</year>
          <volume>5</volume>
          <size units="pages">1291132</size>
          <pub-id pub-id-type="doi">10.3389/fdgth.2023.1291132</pub-id>
        </element-citation>
      </ref>
      <ref id="B33">
        <label>33</label>
        <mixed-citation>33. Sauerbrei A, Kerasidou A, Lucivero F, Hallowell N. The Impact of Artificial Intelligence on the Person-Centred, Doctor-Patient Relationship: Some Problems and Solutions. BMC Med Inform Decis Mak. 2023;23(1):73. doi: 10.1186/s12911-023-02162-y.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Sauerbrei</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Kerasidou</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Lucivero</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Hallowell</surname>
              <given-names>N</given-names>
            </name>
          </person-group>
          <article-title>The Impact of Artificial Intelligence on the Person-Centred, Doctor-Patient Relationship: Some Problems and Solutions</article-title>
          <source>BMC Med Inform Decis Mak</source>
          <year>2023</year>
          <volume>23</volume>
          <issue>1</issue>
          <size units="pages">73</size>
          <pub-id pub-id-type="doi">10.1186/s12911-023-02162-y</pub-id>
        </element-citation>
      </ref>
      <ref id="B34">
        <label>34</label>
        <mixed-citation>34. Wang G, Gao K, Liu Q, Wu Y, Zhang K, Zhou W, et al. Potential and Limitations of ChatGPT 3.5 and 4.0 as a Source of COVID-19 Information: Comprehensive Comparative Analysis of Generative and Authoritative Information. J Med Internet Res. 2023;25:e49771. doi: 10.2196/49771.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Wang</surname>
              <given-names>G</given-names>
            </name>
            <name>
              <surname>Gao</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Liu</surname>
              <given-names>Q</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Zhou</surname>
              <given-names>W</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Potential and Limitations of ChatGPT 3.5 and 4.0 as a Source of COVID-19 Information: Comprehensive Comparative Analysis of Generative and Authoritative Information</article-title>
          <source>J Med Internet Res</source>
          <year>2023</year>
          <volume>25</volume>
          <elocation-id>e49771</elocation-id>
          <pub-id pub-id-type="doi">10.2196/49771</pub-id>
        </element-citation>
      </ref>
      <ref id="B35">
        <label>35</label>
        <mixed-citation>35. Tang Y, Xiao Z, Li X, Zhang Q, Chan EW, Wong IC, et al. Large Language Model in Medical Information Extraction from Titles and Abstracts with Prompt Engineering Strategies: A Comparative Study of GPT-3.5 and GPT-4 [Internet]. medRxiv; 2024 [cited 2025 Feb 28]. p. 2024.03.20.24304572. Available from: <ext-link ext-link-type="uri" xlink:href="https://www.medrxiv.org/content/10.1101/2024.03.20.24304572v1">https://www.medrxiv.org/content/10.1101/2024.03.20.24304572v1</ext-link> </mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Tang</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Xiao</surname>
              <given-names>Z</given-names>
            </name>
            <name>
              <surname>Li</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>Q</given-names>
            </name>
            <name>
              <surname>Chan</surname>
              <given-names>EW</given-names>
            </name>
            <name>
              <surname>Wong</surname>
              <given-names>IC</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>Large Language Model in Medical Information Extraction from Titles and Abstracts with Prompt Engineering Strategies: A Comparative Study of GPT-3.5 and GPT-4</article-title>
          <comment>Internet</comment>
          <source>medRxiv</source>
          <year>2024</year>
          <date-in-citation content-type="cited-date">cited 2025 Feb 28</date-in-citation>
          <size units="pages">2024.03.20.24304572</size>
          <ext-link ext-link-type="uri" xlink:href="https://www.medrxiv.org/content/10.1101/2024.03.20.24304572v1">https://www.medrxiv.org/content/10.1101/2024.03.20.24304572v1</ext-link>
        </element-citation>
      </ref>
      <ref id="B36">
        <label>36</label>
        <mixed-citation>36. Zhou J, Gandomi AH, Chen F, Holzinger A. Evaluating the Quality of Machine Learning Explanations: A Survey on Methods and Metrics. Electronics. 2021;10(5):593. doi: 10.3390/electronics10050593.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Zhou</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Gandomi</surname>
              <given-names>AH</given-names>
            </name>
            <name>
              <surname>Chen</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Holzinger</surname>
              <given-names>A</given-names>
            </name>
          </person-group>
          <article-title>Evaluating the Quality of Machine Learning Explanations: A Survey on Methods and Metrics</article-title>
          <source>Electronics</source>
          <year>2021</year>
          <volume>10</volume>
          <issue>5</issue>
          <size units="pages">593</size>
          <pub-id pub-id-type="doi">10.3390/electronics10050593</pub-id>
        </element-citation>
      </ref>
      <ref id="B37">
        <label>37</label>
        <mixed-citation>37. Johri S, Jeong J, Tran BA, Schlessinger DI, Wongvibulsin S, Barnes LA, et al. An Evaluation Framework for Clinical Use of Large Language Models in Patient Interaction Tasks. Nat Med. 2025;31(1):77-86. doi: 10.1038/s41591-024-03328-5.</mixed-citation>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Johri</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Jeong</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Tran</surname>
              <given-names>BA</given-names>
            </name>
            <name>
              <surname>Schlessinger</surname>
              <given-names>DI</given-names>
            </name>
            <name>
              <surname>Wongvibulsin</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Barnes</surname>
              <given-names>LA</given-names>
            </name>
            <etal>et al</etal>
          </person-group>
          <article-title>An Evaluation Framework for Clinical Use of Large Language Models in Patient Interaction Tasks</article-title>
          <source>Nat Med</source>
          <year>2025</year>
          <volume>31</volume>
          <issue>1</issue>
          <fpage>77</fpage>
          <lpage>86</lpage>
          <pub-id pub-id-type="doi">10.1038/s41591-024-03328-5</pub-id>
        </element-citation>
      </ref>
    </ref-list>
    <fn-group>
      <fn fn-type="other">
        <label>Vincula&#231;&#227;o acad&#234;mica:</label>
        <p> N&#227;o h&#225; vincula&#231;&#227;o deste estudo a programas de p&#243;s-gradua&#231;&#227;o.</p>
      </fn>
      <fn fn-type="other">
        <label>Aprova&#231;&#227;o &#233;tica e consentimento informado:</label>
        <p> Este artigo n&#227;o cont&#233;m estudos com humanos ou animais realizados por nenhum dos autores.</p>
      </fn>
      <fn fn-type="other">
        <label>Uso de Intelig&#234;ncia Artificial:</label>
        <p> Durante a prepara&#231;&#227;o deste trabalho, o(s) autor(es) usaram ChatGPT e Grammarly para aprimorar comandos e corrigir gram&#225;tica e pontua&#231;&#227;o. Ap&#243;s o uso desta ferramenta/servi&#231;o, o(s) autor(es) revisaram e editaram o conte&#250;do conforme necess&#225;rio e assumem total responsabilidade pelo conte&#250;do do artigo publicado.</p>
      </fn>
      <fn fn-type="data-availability" specific-use="data-in-article">
        <label>Disponibilidade de Dados:</label>
        <p> Os conte&#250;dos subjacentes ao texto da pesquisa est&#227;o contidos no manuscrito.</p>
      </fn>
      <fn fn-type="financial-disclosure">
        <label>Fontes de financiamento:</label>
        <p> O presente estudo n&#227;o teve fontes de financiamento externas.</p>
      </fn>
      <fn fn-type="supplementary-material" id="fn_ast1">
        <label>*Material suplementar</label>
        <p>Para informa&#231;&#227;o adicional do Material Suplementar 1, por favor, <ext-link ext-link-type="uri" xlink:href="https://abccardiol.org/supplementary-material/2026/12302/2025-0498_AO_Suplemental_1.pdf">clique aqui</ext-link>. </p>
        <p>Para informa&#231;&#227;o adicional do Material Suplementar 2, por favor, <ext-link ext-link-type="uri" xlink:href="https://abccardiol.org/supplementary-material/2026/12302/2025-0498_AO_Suplemental_2.pdf">clique aqui</ext-link>. </p>
      </fn>
    </fn-group>
  </back>
  <sub-article article-type="translation" id="TRen" xml:lang="en">
    <front-stub>
      <article-id pub-id-type="doi">10.36660/abc.20250498i</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Original Article</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Assessment of ChatGPT-4.0 versus ChatGPT-Mini in Generating Guideline-Based Hypertension Content</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-4570-869X</contrib-id>
          <name>
            <surname>Ata&#237;des</surname>
            <given-names>R&#244;mullo Jos&#233; Costa</given-names>
          </name>
          <role>Conception and design of the research</role>
          <role>Acquisition of data</role>
          <role>Analysis and interpretation of the data and Statistical analysis</role>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Campos</surname>
            <given-names>Marcos Adriano Garcia</given-names>
          </name>
          <role>Conception and design of the research</role>
          <role>Acquisition of data</role>
          <role>Analysis and interpretation of the data and Statistical analysis</role>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff2001">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Souza</surname>
            <given-names>Jo&#227;o V&#237;tor Perez de</given-names>
          </name>
          <role>Analysis and interpretation of the data and Statistical analysis</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff2001">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Rocha</surname>
            <given-names>Rafael Cardoso</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff3001">
            <sup>3</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0009-0008-7777-6727</contrib-id>
          <name>
            <surname>Lacalle</surname>
            <given-names>Almir Alamino</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff3001">
            <sup>3</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Vieira</surname>
            <given-names>Ciro Bezerra</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-6242-0885</contrib-id>
          <name>
            <surname>Artioli</surname>
            <given-names>Thiago</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff4001">
            <sup>4</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-5610-7921</contrib-id>
          <name>
            <surname>Medeiros</surname>
            <given-names>Tiago Cordeiro</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff5001">
            <sup>5</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0381-3344</contrib-id>
          <name>
            <surname>Souza</surname>
            <given-names>Erito Marques de</given-names>
            <suffix>Filho</suffix>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff6001">
            <sup>6</sup>
          </xref>
          <xref ref-type="aff" rid="aff7001">
            <sup>7</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-0884-4190</contrib-id>
          <name>
            <surname>Gismondi</surname>
            <given-names>Ronaldo</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff7001">
            <sup>7</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6352-3330</contrib-id>
          <name>
            <surname>Campana</surname>
            <given-names>&#201;rika Maria Gon&#231;alves</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff8001">
            <sup>8</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-1261-5785</contrib-id>
          <name>
            <surname>Romeo</surname>
            <given-names>Francisco Jos&#233;</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff9001">
            <sup>9</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Razuk</surname>
            <given-names>Victor</given-names>
          </name>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff9001">
            <sup>9</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Vissoci</surname>
            <given-names>Jo&#227;o Ricardo Nickenig</given-names>
          </name>
          <role>Conception and design of the research</role>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff2001">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Lopes</surname>
            <given-names>Renato Delascio</given-names>
          </name>
          <role>Conception and design of the research</role>
          <role>Writing of the manuscript</role>
          <role>Critical revision of the manuscript for content</role>
          <xref ref-type="aff" rid="aff10001">
            <sup>10</sup>
          </xref>
        </contrib>
      </contrib-group>
      <aff id="aff1001">
        <label>1</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Faculdade de Medicina da Universidade de S&#227;o Paulo, S&#227;o Paulo, SP &#8211; Brazil</institution>
      </aff>
      <aff id="aff2001">
        <label>2</label>
        <country country="US">USA</country>
        <institution content-type="original">Departamento de Medicina de Emerg&#234;ncia - Duke University, Durham &#8211; USA</institution>
      </aff>
      <aff id="aff3001">
        <label>3</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Divis&#227;o de Cardiologia do Hospital das Cl&#237;nicas de Ribeir&#227;o Preto da Universidade de S&#227;o Paulo, Ribeir&#227;o Preto, SP &#8211; Brazil</institution>
      </aff>
      <aff id="aff4001">
        <label>4</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Instituto Dante Pazzanese de Cardiologia, S&#227;o Paulo, SP &#8211; Brazil</institution>
      </aff>
      <aff id="aff5001">
        <label>5</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Hospital Universit&#225;rio - Universidade Federal do Maranh&#227;o, S&#227;o Lu&#237;s, MA &#8211; Brazil</institution>
      </aff>
      <aff id="aff6001">
        <label>6</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal Rural do Rio de Janeiro, Serop&#233;dica, RJ &#8211; Brazil</institution>
      </aff>
      <aff id="aff7001">
        <label>7</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Federal Fluminense, Niter&#243;i, RJ &#8211; Brazil</institution>
      </aff>
      <aff id="aff8001">
        <label>8</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade do Estado do Rio de Janeiro, Rio de Janeiro, RJ &#8211; Brazil</institution>
      </aff>
      <aff id="aff9001">
        <label>9</label>
        <country country="US">USA </country>
        <institution content-type="original">Departamento de Medicina Cardiovascular &#8211; University of Miami Miller School of Medicine/Jackson Memorial Hospital, Miami &#8211; USA </institution>
      </aff>
      <aff id="aff10001">
        <label>10</label>
        <country country="US">USA</country>
        <institution content-type="original">Duke Clinical Research Institute Cardiology, Durham &#8211; USA</institution>
      </aff>
      <author-notes>
        <corresp id="c01001"><label>Mailing Address:</label> R&#244;mullo Jos&#233; Costa Ata&#237;des Av. Dr. Arnaldo, 455. Postal Code 01246-903, Cerqueira C&#233;sar, Pacaembu, SP &#8211; Brazil E-mail: romulloataides2@hotmail.com </corresp>
        <fn fn-type="edited-by">
          <label>Editor responsible for the review:</label>
          <p>Gl&#225;ucia Maria Moraes de Oliveira</p>
        </fn>
        <fn fn-type="coi-statement">
          <label>Potential conflict of interest:</label>
          <p>No potential conflict of interest relevant to this article was reported.</p>
        </fn>
      </author-notes>
      <abstract>
        <title>Abstract</title>
        <sec>
          <title>Background</title>
          <p> Artificial intelligence (AI) language models are increasingly used to generate patient education materials. However, their accuracy, completeness, and adherence to clinical guidelines remain uncertain.</p>
        </sec>
        <sec>
          <title>Objectives</title>
          <p> To compare ChatGPT-Mini and ChatGPT-4.0 in the generation of hypertension education content with respect to accuracy, completeness, structural quality using the Ensuring Quality Information for Patients (EQIP), response consistency, and alignment with established guidelines.</p>
        </sec>
        <sec>
          <title>Methods</title>
          <p> A standardized set of 31 hypertension-related questions was submitted to both models. Outputs were independently evaluated by 10 blinded clinicians using a modified EQIP score, a 5-point accuracy scale, and a 3-point completeness scale. Response consistency was assessed using BERTScore. Between-model comparisons were performed using the two-sided Wilcoxon rank-sum test (p &lt; 0.05). Effect sizes were reported as Hodges&#8211;Lehmann (HL) median differences and Cliff&#8217;s delta (&#948;), both with 95% CIs. Inter-rater reliability was estimated using the intraclass correlation coefficient (ICC; two-way random effects model, absolute agreement).</p>
        </sec>
        <sec>
          <title>Results</title>
          <p> Central tendency measures favored ChatGPT-4.0, although differences were small. Median scores were as follows: accuracy, 4.10 (3.70-4.20) versus 3.73 (3.60-4.05); completeness, 1.26 (1.17-1.41) versus 1.10 (0.96-1.23); and total EQIP score, 19.5 (18.0-25.0) versus 18.5 (16.0-23.0) for ChatGPT-4.0 and ChatGPT-Mini, respectively. HL median differences were small, with 95% CIs crossing zero (accuracy: +0.37, &#8722;0.25 to +0.50; completeness: +0.16, &#8722;0.06 to +0.36; EQIP: +1.0, &#8722;1.0 to +6.0). Cliff&#8217;s &#948; values were consistently small and positive across primary outcomes, indicating only modest stochastic dominance of ChatGPT-4.0. Identification clarity tended to be higher with ChatGPT-4.0, whereas response consistency measured by BERTScore F1 was generally higher for ChatGPT-Mini (&gt; 0.92 versus 0.885-0.932). Inter-rater reliability was good to excellent across all measures (ICC &gt; 0.80).</p>
        </sec>
        <sec>
          <title>Conclusions</title>
          <p> ChatGPT-4.0 demonstrated small, non-significant improvements in accuracy, completeness, and structural quality compared with ChatGPT-Mini. Effect sizes were modest, and all 95% CIs included zero. ChatGPT-Mini produced more consistent responses. These findings underscore the importance of routinely reporting effect sizes with 95% CIs and support the use of standardized evaluation methods and real-time validation frameworks for AI-generated medical education content.</p>
        </sec>
      </abstract>
      <abstract abstract-type="graphical">
        <p>
          <fig id="f01001">
            <label>Central Illustration:</label>
            <caption>
              <title>Assessment of ChatGPT-4.0 versus ChatGPT-Mini in Generating Guideline-Based Hypertension Content</title>
            </caption>
            <attrib>CC: American College of Cardiology; AHA: American Heart Association; AI: artificial intelligence; EQIP: Ensuring Quality Information for Patients; ESC: European Society of Cardiology; ESH: European Society of Hypertension; ICC: intraclass correlation coefficient.</attrib>
            <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf01-en.tif"/></fig>
        </p>
      </abstract>
      <kwd-group xml:lang="en">
        <title>Keywords</title>
        <kwd>Artificial Intelligence</kwd>
        <kwd>Hypertension</kwd>
        <kwd>Patient Education as Topic</kwd>
      </kwd-group>
    </front-stub>
    <body>
      <sec sec-type="intro">
        <title>Introduction</title>
        <p>The use of artificial intelligence (AI) in hypertension management offers substantial potential.<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref></sup> However, current evidence indicates that its use remains insufficiently systematized, particularly regarding evaluation criteria and patient-centered approaches.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B4">4</xref></sup> For example, AI-driven tools such as chatbots have demonstrated utility in supporting hypertension self-management, yet existing studies consistently highlight the need for more robust evaluation frameworks and clearer, more consistent guidance for patients.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
        <p>Almagazzachi et al. investigated the use of generative AI for patient education in hypertension and reported that, despite providing meaningful support, these tools show inconsistencies in alignment with established clinical guidelines, including those of the European Society of Hypertension (ESH).<sup><xref ref-type="bibr" rid="B6">6</xref></sup> In addition, variability in chatbot performance across different languages has been documented, further emphasizing the necessity for standardized methodologies to ensure accuracy, reliability, and guideline adherence.<sup><xref ref-type="bibr" rid="B7">7</xref></sup> Adopting a user-centered approach in the development of AI solutions for hypertension management may enhance systematization and improve patient outcomes.<sup><xref ref-type="bibr" rid="B5">5</xref></sup> Therefore, although AI demonstrates considerable promise in this field, a more structured and patient-focused framework is required to fully realize its potential in hypertension care.<sup><xref ref-type="bibr" rid="B8">8</xref></sup></p>
        <p>Several studies have attempted to systematize the evaluation of AI-generated educational information for patients.<sup><xref ref-type="bibr" rid="B9">9</xref>,<xref ref-type="bibr" rid="B10">10</xref></sup> Nevertheless, these efforts have not incorporated a systematic process for developing evaluation questions based on expert consensus.<sup><xref ref-type="bibr" rid="B11">11</xref></sup></p>
        <p>The present study aims to assess the performance of two ChatGPT model versions, ChatGPT-Mini and ChatGPT-4.0, in generating hypertension-related medical content. Validated instruments were used to evaluate the quality of patient information and to examine the strengths and limitations of AI in medical education and patient care.<sup><xref ref-type="bibr" rid="B11">11</xref>,<xref ref-type="bibr" rid="B12">12</xref></sup> Furthermore, by applying a structured and standardized evaluation of AI-generated responses, this study seeks to identify areas in which AI-produced content may be deficient, thereby providing relevant insights to inform future AI development in health care.<sup><xref ref-type="bibr" rid="B13">13</xref></sup> This systematic approach enables an assessment of how effectively AI models deliver accurate, patient-centered educational materials and supports the refinement of AI-based strategies for medical information dissemination.</p>
      </sec>
      <sec sec-type="methods">
        <title>Methods</title>
        <sec>
          <title>Study design</title>
          <p>This study employed a survey-based design (<xref ref-type="fig" rid="f02001">Figure 1</xref>). Questions addressing the diagnosis, treatment, and prognosis of hypertension were submitted to ChatGPT-Mini and ChatGPT-4.0 (<xref ref-type="sec" rid="suppl01_en">Supplementary Table 1</xref>). The generated responses were compared with recommendations from the 2023 ESH guidelines and the 2017 American College of Cardiology/American Heart Association (ACC/AHA) hypertension guidelines. All responses were evaluated by independent cardiologists and cardiology fellows.</p>
          <p>
            <fig id="f02001">
              <label>Figure 1</label>
              <caption>
                <title>&#8211; Study flowchart for the evaluation of artificial intelligence-generated hypertension content. ACC: American College of Cardiology; AHA: American Heart Association; EQIP: Ensuring Quality Information for Patients; ESH: European Society of Hypertension.</title>
              </caption>
              <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf02-en.tif"/></fig>
          </p>
        </sec>
        <sec>
          <title>Data collection</title>
          <p>The research team developed 31 questions related to hypertension (Table 1; see <xref ref-type="sec" rid="suppl01_en">Supplementary Material 1</xref> for additional details). The content and structure of each question were guided by the Ensuring Quality Information for Patients (EQIP) score.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> The EQIP score is a validated tool designed to assess the quality of information provided to patients in health care settings, with a focus on accuracy, reliability, and usefulness in supporting informed health-related decision-making.</p>
          <p>Each question was submitted 10 times in English to both ChatGPT versions between July 25 and July 30, 2024. All responses were systematically recorded by initiating new conversations using unconditional prompts in temporary chat mode. In cases in which the response format did not adequately address the original question, the prompt &#8220;<italic>Be specific and incorporate medical knowledge about hypertension from the guideline of hypertension from AHA/ACC 2017 and ESH 2023</italic>&#8221; was appended to the conversation as needed.</p>
        </sec>
        <sec>
          <title>Expert rating of GPT responses</title>
          <p>A total of 10 physicians (five cardiologists and five cardiology fellows) from multiple institutions in Brazil and the United States independently evaluated the responses generated by ChatGPT (<xref ref-type="fig" rid="f03001">Figure 2</xref>). Evaluators received written instructions detailing how to interpret the responses, apply the evaluation criteria, and compare the content with the 2017 ACC/AHA and 2023 ESH hypertension guidelines. Although evaluators were permitted to consult the guidelines, they were encouraged to rely on their clinical expertise when assessing response quality.</p>
          <p>
            <fig id="f03001">
              <label>Figure 2</label>
              <caption>
                <title>&#8211; Example of the evaluation of artificial intelligence-generated hypertension responses. EQIP: Ensuring Quality Information for Patients.</title>
              </caption>
              <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf03-en.tif"/></fig>
          </p>
          <p>All evaluators were blinded to the source of the responses and were unaware of whether the content had been generated by ChatGPT-Mini or ChatGPT-4.0. Ratings were collected using surveys created in REDCap (<xref ref-type="sec" rid="suppl01_en">Supplementary Material 1</xref>).</p>
        </sec>
        <sec>
          <title>Measures</title>
          <p>Based on guideline-derived information and clinical knowledge, evaluators classified each response using three components: the EQIP score to assess overall information quality and two Likert-type scales to independently assess accuracy and completeness (<xref ref-type="sec" rid="suppl01_en">Supplementary Material 1</xref>).</p>
          <p>The EQIP score was adapted for the evaluation of responses generated by ChatGPT. It consists of criteria covering three domains: content (assessing accuracy, relevance, and comprehensiveness), identification (evaluating clarity and ease of identifying information), and data structure (examining organization, readability, and presentation of information). The modified EQIP instrument allows a maximum score of 31 points, distributed across the content (maximum 16 points), identification (maximum 5 points), and structural data (maximum 10 points) domains.</p>
          <p>Each EQIP criterion was scored using a three-point system: 1 (correct response), 0.5 (partially correct response), or 0 (incorrect response). The total EQIP score was calculated as the sum of points across all criteria, with higher scores indicating higher-quality information. For all responses, one point was automatically awarded for the item &#8220;<italic>Can you date the above-named information?</italic>&#8221;, as this element could potentially reveal the version of ChatGPT being evaluated.</p>
          <p>Accuracy was assessed using a five-point scale: completely incorrect (no correct information and contradictory to hypertension guidelines); more incorrect than correct (predominantly incorrect and largely inconsistent with guidelines); approximately correct and incorrect (a similar proportion of correct and incorrect information, partially aligned with guidelines); more correct than incorrect (predominantly correct and generally aligned with guidelines); and completely correct (fully accurate, comprehensive, and entirely consistent with hypertension guidelines).</p>
          <p>Completeness was evaluated using a three-point scale reflecting how thoroughly each response addressed the question: incomplete (key aspects missing or insufficiently addressed); adequate (all requested aspects addressed with the minimum information required); and comprehensive (all aspects addressed with additional relevant context beyond expectations).</p>
          <p>Response consistency across the 10 repetitions of each question was assessed using BERTScore, which quantifies semantic similarity based on precision, recall, and F1 scores. Results were presented graphically to compare content similarity across responses.<sup><xref ref-type="bibr" rid="B15">15</xref></sup></p>
        </sec>
        <sec>
          <title>ChatGPT sampling</title>
          <p>The minimum number of raters required to achieve an intraclass correlation coefficient (ICC) of 0.8, with a precision of &#177; 0.1 and a 95% CI, when evaluating the 31 items generated by each ChatGPT version, was estimated to be 10.</p>
        </sec>
        <sec>
          <title>Statistical analysis</title>
          <p>The distribution of continuous variables was assessed using the Shapiro&#8211;Wilk test. As all continuous variables exhibited non-normal distributions, data are presented as medians and interquartile ranges (IQRs). Comparisons between ChatGPT-Mini and ChatGPT-4.0 were performed using the Wilcoxon rank-sum test (two-sided, p &lt; 0.05).</p>
          <p>Effect sizes were reported with 95% CIs, including Hodges&#8211;Lehmann median differences to estimate location shifts and Cliff&#8217;s delta (&#948;) to assess stochastic dominance. 95% CIs were derived using Walsh averages or nonparametric bootstrap methods, as appropriate. Inter-rater agreement among evaluators was quantified using ICC (two-way random effects model, absolute agreement), with 95% CIs.</p>
          <p>Comparisons between ChatGPT-4.0 and ChatGPT-Mini across accuracy, completeness, and EQIP domains were conducted using Cliff&#8217;s delta, with &#948; values greater than 0 indicating superiority of ChatGPT-4.0. All estimates are reported with 95% CIs, and statistical significance was defined as p &lt; 0.05. Statistical analyses were performed using R (version 4.2.2). BERTScore analyses were conducted in Google Colab using the <italic>bert-score</italic> package.</p>
        </sec>
      </sec>
      <sec sec-type="results">
        <title>Results</title>
        <sec>
          <title>Study sample and data inclusion</title>
          <p>All responses generated by ChatGPT-Mini and ChatGPT-4.0 were included in the analysis, with no exclusions or missing data. The standardized set of 31 questions was applied uniformly to both models. Across the 10 repetitions per question, no instances of AI hallucination were identified. In addition, no follow-up prompts were required to clarify responses, as all outputs adequately addressed the original questions.</p>
        </sec>
        <sec>
          <title>Accuracy and completeness</title>
          <p>As shown in Table 2, both models achieved high scores for accuracy and completeness. Median values were numerically higher for ChatGPT-4.0; however, no statistically significant differences were observed between the two models for either outcome.</p>
        </sec>
        <sec>
          <title>Overall quality (EQIP scores)</title>
          <p>The EQIP score, which integrates content, identification, and structural domains, indicated broadly comparable overall quality between the two models. As presented in Table 2, ChatGPT-4.0 tended to achieve numerically higher total EQIP scores as well as higher content and identification subscores, compared with ChatGPT-Mini. Structural scores were very similar between models, resulting in comparable overall EQIP totals.</p>
          <p>Effect-size analyses using Cliff&#8217;s &#948; favored ChatGPT-4.0 across all EQIP domains; however, all corresponding 95% CIs included zero (Table 3), indicating that these differences were not statistically robust. <xref ref-type="fig" rid="f04001">Figure 3</xref> illustrates the distribution of EQIP scores across domains for each model, demonstrating generally favorable evaluations by raters for content, identification, and structure. In this figure, each horizontal line corresponds to an individual EQIP item: items 1-16 represent content, items 17-20 represent identification, and items 21-30 represent structure. Across all items, both models received consistently favorable and largely comparable ratings.</p>
          <p>
            <fig id="f04001">
              <label>Figure 3</label>
              <caption>
                <title>&#8211; Modified EQIP score&#8211;based comparison of ChatGPT responses across hypertension questions. Responses generated by ChatGPT-Mini and ChatGPT-4.0 across 30 hypertension-related questions were evaluated using a modified EQIP score. Responses were classified as 1 (correct), 0.5 (partially correct), or 0 (incorrect). The x-axis represents the proportion of responses in each category, and the y-axis lists the evaluated questions. EQIP: Ensuring Quality Information for Patients.</title>
              </caption>
              <graphic xlink:href="1678-4170-abc-123-02-e20250498-50498-gf04-en.tif"/></fig>
          </p>
        </sec>
        <sec>
          <title>BERTScore evaluations</title>
          <p>BERTScore analyses revealed differences between models in precision, recall, and response consistency. ChatGPT-Mini consistently demonstrated higher F1 scores, frequently exceeding 0.92, whereas ChatGPT-4.0 exhibited F1 scores ranging from 0.8851 to 0.9316 (see <xref ref-type="sec" rid="suppl02_en">Supplementary Material 2</xref>).</p>
          <p>For precision, ChatGPT-Mini achieved a score of 0.9356 compared with 0.8643 for ChatGPT-4.0 in response to a question regarding the definition of hypertension. Similarly, for recall, ChatGPT-Mini scored 0.9558, whereas ChatGPT-4.0 scored 0.9330 in a related follow-up question.</p>
          <p>Overall, ChatGPT-Mini showed greater consistency in response stability, with F1 scores rarely falling below 0.90. In contrast, ChatGPT-4.0 demonstrated greater variability, including an F1 score of 0.8757 in response to a question addressing non-pharmacological treatments for hypertension.</p>
        </sec>
        <sec>
          <title>Inter-rater reliability</title>
          <p>Inter-rater reliability was consistently high across all evaluated metrics, with ICCs exceeding 0.80 for both models.</p>
          <p>For EQIP scores, ICCs were 0.844 (95% CI, 0.677-0.952) for ChatGPT-Mini and 0.840 (95% CI, 0.665-0.951) for ChatGPT-4.0. For accuracy scores, ICCs were 0.742 (95% CI, 0.456-0.922) for ChatGPT-Mini and 0.688 (95% CI, 0.339-0.906) for ChatGPT-4.0. For completeness scores, ICCs were identical for both models at 0.824 (95% CI, 0.627-0.947 for ChatGPT-Mini and 0.628-0.947 for ChatGPT-4.0).</p>
          <p>These ICC values indicate moderate to strong inter-rater agreement, supporting the reliability of the evaluation methodology used in this study. Overall, the findings suggest superior performance of ChatGPT-4.0 across key quality metrics (Central Illustration).</p>
        </sec>
      </sec>
      <sec sec-type="discussion">
        <title>Discussion</title>
        <p>As AI becomes increasingly integrated into patient education, persistent concerns remain regarding the accuracy, completeness, and consistency of AI-generated medical information. Although AI holds substantial potential to improve health care accessibility, its capacity to deliver reliable, guideline-adherent content remains a critical challenge, particularly in the management of chronic conditions such as hypertension. In this study, we systematically evaluated the performance of ChatGPT-Mini and ChatGPT-4.0 in generating patient-centered hypertension education, with a focus on guideline alignment, informational quality, and response consistency.</p>
        <p>Our findings indicate that ChatGPT-4.0 demonstrated incremental improvements over ChatGPT-Mini in accuracy, completeness, and response structure. However, these differences were modest and did not reach statistical significance, likely reflecting small effect sizes. ChatGPT-4.0 achieved a slightly higher mean accuracy score than ChatGPT-Mini (4.02 &#177; 0.35 vs 3.86 &#177; 0.45), along with a higher median completeness score (1.26 vs 1.10). Despite these numerical gains, both models showed notable limitations in content completeness. These observations are consistent with prior studies reporting that newer AI models improve linguistic coherence and structural organization<sup><xref ref-type="bibr" rid="B11">11</xref></sup> but may still lack sufficient depth and specificity when addressing complex clinical recommendations.<sup><xref ref-type="bibr" rid="B16">16</xref>,<xref ref-type="bibr" rid="B17">17</xref></sup></p>
        <p>To evaluate the quality of AI-generated medical content systematically, we applied established assessment instruments, including the EQIP score and Likert-type scales. The EQIP score, originally developed to assess written patient education materials,<sup><xref ref-type="bibr" rid="B18">18</xref></sup> provided a structured framework for evaluating clarity, relevance, and reliability in AI-generated responses.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> Although EQIP has not been formally validated for AI-generated content, its use in this study highlights the growing need for standardized evaluation frameworks tailored to AI-driven health care communication. Likert scales further supported the assessment of perceived accuracy and completeness, capturing professional judgments regarding the trustworthiness of AI-generated information.<sup><xref ref-type="bibr" rid="B12">12</xref></sup></p>
        <p>ChatGPT-4.0 achieved a slightly higher overall EQIP score than ChatGPT-Mini (21.4 &#177; 3.7 vs 19.5 &#177; 4.0), suggesting modest improvements in the quality of hypertension-related educational content. The most pronounced differences were observed in the content domain (9.15 &#177; 2.32 vs 8.35 &#177; 2.32) and the identification domain (3.80 &#177; 1.14 vs 2.85 &#177; 1.08), reflecting improved comprehensiveness and transparency. Nevertheless, these differences were not statistically significant, indicating that enhancements in presentation and structure may not consistently translate into more reliable or guideline-adherent content.<sup><xref ref-type="bibr" rid="B19">19</xref></sup></p>
        <p>When examining the proportion of fully correct responses, ChatGPT-4.0 outperformed ChatGPT-Mini, producing correct answers (score = 1) in approximately 70%-80% of evaluated questions, compared with fewer than 50% for ChatGPT-Mini. Despite this improvement, adherence to established hypertension guidelines, including the 2017 ACC/AHA and 2023 ESH recommendations, remained inconsistent. Although both models generally aligned with core guideline principles, responses frequently lacked clinical specificity or deviated from evidence-based standards. This pattern mirrors findings from broader evaluations of AI-generated patient education, where reported accuracy rates range from 71.7% to 94.3%, yet deficits in depth and clinical precision remain common.<sup><xref ref-type="bibr" rid="B20">20</xref></sup> ChatGPT-Mini demonstrated a higher proportion of incorrect or partially correct responses, particularly for questions requiring detailed clinical explanations. ChatGPT-4.0 reduced the frequency of incomplete responses to below 20% for most variables; however, 10%-30% of responses still fell into lower-quality categories, underscoring persistent informational gaps.</p>
        <p>These findings are consistent with a previous study evaluating ChatGPT-4.0 across five hepato-pancreatico-biliary conditions, which reported a median EQIP score of 16 (IQR, 14.5-18), indicating substantial deficiencies in citation accuracy and structural coherence.<sup><xref ref-type="bibr" rid="B10">10</xref></sup> While our study observed higher EQIP scores for hypertension-related content, both investigations highlight the continued limitations of AI models in fully aligning with evidence-based medical recommendations. Importantly, although newer models may improve readability and organization, their factual accuracy does not appear to have advanced proportionally.<sup><xref ref-type="bibr" rid="B21">21</xref></sup> This reinforces the need for rigorous validation processes and ongoing human oversight to ensure that improvements in clarity do not compromise adherence to clinical guidelines.<sup><xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B22">22</xref></sup></p>
        <p>Despite these limitations, ChatGPT-4.0 demonstrates potential as a supplementary tool for patient education in hypertension management. Its capacity to rapidly generate guideline-informed explanations may reduce the informational burden on health professionals, particularly for initial discussions of treatment strategies and lifestyle interventions.<sup><xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref></sup> Prior research suggests that better-informed patients are more likely to adhere to prescribed therapies, potentially improving outcomes in chronic conditions such as hypertension.<sup><xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref></sup> Nevertheless, AI-generated content alone may not yet provide sufficiently comprehensive or actionable guidance, especially given ongoing public skepticism toward AI applications in health care.<sup><xref ref-type="bibr" rid="B25">25</xref>-<xref ref-type="bibr" rid="B27">27</xref></sup></p>
        <p>A critical concern in AI-generated medical content is the phenomenon of &#8220;hallucinations,&#8221; in which models produce plausible but incorrect information. Notably, no hallucinations were identified in either model across the 10 repeated interactions in this study, which suggests reliable performance within this specific clinical domain. However, previous studies indicate that hallucinations remain a broader risk in health care AI, particularly for complex or less well-defined topics.<sup><xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref></sup> This risk, combined with known sensitivity to prompt design, underscores the continued importance of clinician oversight in clinical and educational settings.<sup><xref ref-type="bibr" rid="B30">30</xref></sup></p>
        <p>AI is also increasingly influencing the development of clinical guidelines and patient communication by facilitating the generation of clinical questions and supporting real-time interactions. AI-driven systems have shown promise in prioritizing clinical recommendations and streamlining information delivery.<sup><xref ref-type="bibr" rid="B31">31</xref>,<xref ref-type="bibr" rid="B32">32</xref></sup> Nonetheless, inconsistent adherence to clinical standards remains a major limitation, necessitating ongoing refinement of these technologies.<sup><xref ref-type="bibr" rid="B33">33</xref></sup></p>
        <p>Response consistency was further evaluated using BERTScore metrics, which quantify semantic similarity across repeated responses. ChatGPT-Mini demonstrated greater response stability, achieving higher precision (0.9356 vs 0.8643) and recall (0.9558 vs 0.9330) than ChatGPT-4.0. In contrast, ChatGPT-4.0 exhibited greater variability, with F1 scores ranging from 0.8851 to 0.9316. These findings suggest that ChatGPT-Mini may be better suited for structured patient education materials requiring high consistency, whereas ChatGPT-4.0&#8217;s variability may enhance engagement in more dynamic, conversational contexts. Similar patterns have been reported in prior studies, where newer models prioritize flexibility at the expense of strict consistency.<sup><xref ref-type="bibr" rid="B34">34</xref>,<xref ref-type="bibr" rid="B35">35</xref></sup></p>
        <p>Finally, inter-rater reliability was high across all evaluated domains, with ICC values exceeding 0.80 for EQIP, accuracy, and completeness scores. This strong agreement among evaluators supports the robustness of the assessment methodology and indicates that observed differences reflect genuine variations in model performance rather than rater subjectivity.</p>
        <p>In summary, this study contributes to the growing literature on the role of AI in health care communication and patient education. While advances in AI models have improved clarity and structural organization, ensuring accuracy, consistency, and guideline adherence remains essential for their safe integration into clinical practice.<sup><xref ref-type="bibr" rid="B26">26</xref>,<xref ref-type="bibr" rid="B36">36</xref></sup> Our findings highlight the urgent need for standardized, rigorous evaluation frameworks for AI-generated medical content. Emerging approaches, such as structured assessment models including the Conversational Reasoning Assessment Framework for Testing in Medicine, may offer valuable tools for systematically evaluating AI performance in health care contexts.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> Future research should focus on real-time validation mechanisms that dynamically link AI-generated responses to up-to-date clinical guidelines and promote interdisciplinary collaboration among clinicians, AI developers, and regulatory authorities to establish robust ethical, safety, and quality standards for AI use in health care.<sup><xref ref-type="bibr" rid="B25">25</xref></sup></p>
        <sec>
          <title>Study limitations</title>
          <p>Although this study provides relevant insights into the performance of ChatGPT-Mini and ChatGPT-4.0 in patient education, several limitations should be acknowledged. First, the analysis was restricted to a single clinical domain, hypertension, which limits the generalizability of the findings to other medical conditions. Because AI performance may vary across specialties, future investigations should evaluate AI-generated content across a broader range of clinical topics.</p>
          <p>Second, despite the use of structured evaluation tools such as EQIP and BERTScore, the assessment relied on expert human judgment, which is inherently subjective, even with standardized rating criteria. Variability in expert interpretation may have influenced scoring, underscoring the need for complementary automated and consensus-based validation approaches in future research.</p>
          <p>Third, the evaluation was limited to English-language outputs, thereby restricting applicability to non&#8211;English-speaking populations. As prior studies have demonstrated variability in AI performance across languages, further research is warranted to assess linguistic differences in AI-generated medical content and to ensure equitable access to accurate patient education globally.</p>
          <p>In addition, the AI models were assessed in a controlled, static testing environment and were not exposed to real-world patient interactions or iterative feedback. AI-generated medical content is inherently dynamic, influenced by model updates and evolving clinical guidelines. Future studies should incorporate real-time validation strategies that continuously benchmark AI outputs against current clinical standards to maintain accuracy and relevance.</p>
          <p>Finally, this study focused exclusively on ChatGPT-Mini and ChatGPT-4.0. Other large language models with distinct architectures and training paradigms (e.g., Claude, Gemini, and Med-PaLM) may demonstrate different performance characteristics. Comparative evaluations involving multiple AI systems are therefore necessary to more comprehensively define the role of AI in patient education.</p>
        </sec>
      </sec>
      <sec sec-type="conclusions">
        <title>Conclusion</title>
        <p>This study demonstrates that although ChatGPT-4.0 offers improved structural coherence compared with ChatGPT-Mini, it does not produce statistically significant gains in factual completeness or response consistency. These findings highlight the importance of implementing standardized evaluation frameworks, real-time validation mechanisms, and continued human oversight to support the safe and effective use of AI in patient education.</p>
        <p>Future research should emphasize dynamic validation strategies and the development of universally accepted evaluation standards to ensure that medical AI systems remain clinically reliable and ethically responsible. While AI technologies have the potential to enhance health care accessibility, human supervision remains essential to prevent misinformation and preserve patient trust. Ultimately, the responsible integration of AI into medical practice will depend on sustained interdisciplinary collaboration to uphold rigorous standards of accuracy, safety, and ethical integrity.</p>
      </sec>
      <sec id="suppl-en" sec-type="supplementary-material">
        <title>Supplemental Materials</title>
        <supplementary-material id="suppl01_en" mime-subtype="pdf" mimetype="application">
          <label>Supplemental Content 1</label>
          <media mime-subtype="pdf" mimetype="application" xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/cf0dad04c1dcc423600ea13700e5be97ebec7c4f.pdf"/>
        </supplementary-material>
        <supplementary-material id="suppl02_en" mime-subtype="pdf" mimetype="application">
          <label>Supplemental Content 2</label>
          <media mime-subtype="pdf" mimetype="application" xlink:href="https://minio.scielo.br/documentstore/1678-4170/6Jyy5jHtZVmY6Sm6DXVrdNw/294918e6c92161ac3dbbf06bd155d73431bfe8a1.pdf"/>
        </supplementary-material>
      </sec>
    </body>
    <back>
      <fn-group>
        <fn fn-type="other">
          <label>Study association:</label>
          <p> This study is not associated with any thesis or dissertation work.</p>
        </fn>
        <fn fn-type="other">
          <label>Ethics approval and consent to participate:</label>
          <p> This article does not contain any studies with human participants or animals performed by any of the authors.</p>
        </fn>
        <fn fn-type="other">
          <label>Use of Artificial Intelligence:</label>
          <p> During the preparation of this work, the author(s) used ChatGPT and Grammaly for refining commands and correcting grammar, punctuation. After using this tool/service, the author(s) reviewed and edited the content as needed and take full responsibility for the content of the published article.</p>
        </fn>
        <fn fn-type="data-availability" specific-use="data-in-article">
          <label>Data Availability Statement:</label>
          <p> The underlying content of the research text is contained within the manuscript.</p>
        </fn>
        <fn fn-type="financial-disclosure">
          <label>Sources of funding:</label>
          <p> There were no external funding sources for this study.</p>
        </fn>
        <fn fn-type="supplementary-material" id="fn_ast1001">
          <label>*Supplemental Materials</label>
          <p>For additional information Supplemental Material 1, please <ext-link ext-link-type="uri" xlink:href="https://abccardiol.org/supplementary-material/2026/12302/2025-0498_AO_Suplemental_1.pdf">click here</ext-link>. </p>
          <p>For additional information Supplemental Material 2, please <ext-link ext-link-type="uri" xlink:href="https://abccardiol.org/supplementary-material/2026/12302/2025-0498_AO_Suplemental_2.pdf">click here</ext-link>. </p>
        </fn>
      </fn-group>
    </back>
  </sub-article>
</article>
