<!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">JCZCM336rBq3Ym3XyqLWV4D</article-id>
      <article-id specific-use="scielo-v2" pub-id-type="publisher-id">S0066-782X2022001200006</article-id>
      <article-id pub-id-type="publisher-id">abc.20220355</article-id>
      <article-id pub-id-type="doi">10.36660/abc.20220355</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 Psicom&#233;trica da Prova de T&#237;tulo de Especialista em Cardiologia da Sociedade Brasileira de Cardiologia</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-6466-1287</contrib-id>
          <name>
            <surname>Marinho</surname>
            <given-names>Gustavo Eug&#234;nio Martins</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
          <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</role>
          <role>An&#225;lise estat&#237;stica</role>
          <role>Obten&#231;&#227;o de financiamento e Reda&#231;&#227;o do manuscrito</role>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-2684-0142</contrib-id>
          <name>
            <surname>Peixoto</surname>
            <given-names>Jos&#233; Maria</given-names>
          </name>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
          <xref ref-type="aff" rid="aff3">
            <sup>3</sup>
          </xref>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6537-7568</contrib-id>
          <name>
            <surname>Knopfholz</surname>
            <given-names>Jos&#233;</given-names>
          </name>
          <xref ref-type="aff" rid="aff4">
            <sup>4</sup>
          </xref>
          <xref ref-type="aff" rid="aff5">
            <sup>5</sup>
          </xref>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6599-5702</contrib-id>
          <name>
            <surname>Andrade</surname>
            <given-names>Marcus Vinicius Santos</given-names>
          </name>
          <xref ref-type="aff" rid="aff6">
            <sup>6</sup>
          </xref>
          <xref ref-type="aff" rid="aff7">
            <sup>7</sup>
          </xref>
          <xref ref-type="aff" rid="aff8">
            <sup>8</sup>
          </xref>
          <role>Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante</role>
        </contrib>
      </contrib-group>
      <aff id="aff1">
        <label> 1 </label>
        <institution content-type="orgdiv1">Faculdade de Medicina</institution>
        <institution content-type="orgname">Universidade Jos&#233; do Ros&#225;rio Vellano</institution>
        <addr-line>
          <named-content content-type="city">Alfenas</named-content>
          <named-content content-type="state">MG</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Faculdade de Medicina na Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS), Alfenas , MG &#8211; Brasil </institution>
      </aff>
      <aff id="aff2">
        <label> 2 </label>
        <institution content-type="orgname">Universidade Jos&#233; do Ros&#225;rio Vellano</institution>
        <addr-line>
          <named-content content-type="city">Belo Horizonte</named-content>
          <named-content content-type="state">MG</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS), Belo Horizonte , MG &#8211; Brasil </institution>
      </aff>
      <aff id="aff3">
        <label> 3 </label>
        <institution content-type="orgdiv1">Programa de Mestrado Profissional em Ensino em Sa&#250;de</institution>
        <institution content-type="orgname">Universidade Jos&#233; do Ros&#225;rio Vellano</institution>
        <addr-line>
          <named-content content-type="city">Belo Horizonte</named-content>
          <named-content content-type="state">MG</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Programa de Mestrado Profissional em Ensino em Sa&#250;de da Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS-BH), Belo Horizonte , MG &#8211; Brasil </institution>
      </aff>
      <aff id="aff4">
        <label> 4 </label>
        <institution content-type="orgname">Pontif&#237;cia Universidade Cat&#243;lica do Paran&#225;</institution>
        <addr-line>
          <named-content content-type="city">Curitiba</named-content>
          <named-content content-type="state">PR</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Pontif&#237;cia Universidade Cat&#243;lica do Paran&#225; , Curitiba , PR &#8211; Brasil </institution>
      </aff>
      <aff id="aff5">
        <label> 5 </label>
        <institution content-type="orgdiv1">Programa de Especializa&#231;&#227;o em Educa&#231;&#227;o em Sa&#250;de</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"> Programa de Especializa&#231;&#227;o em Educa&#231;&#227;o em Sa&#250;de da Universidade de S&#227;o Paulo , S&#227;o Paulo , SP &#8211; Brasil </institution>
      </aff>
      <aff id="aff6">
        <label> 6 </label>
        <institution content-type="orgname">Escola Bahiana de Medicina</institution>
        <addr-line>
          <named-content content-type="city">Salvador</named-content>
          <named-content content-type="state">BA</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Escola Bahiana de Medicina , Salvador , BA &#8211; Brasil </institution>
      </aff>
      <aff id="aff7">
        <label> 7 </label>
        <institution content-type="orgdiv1">Hospital Santa Izabel</institution>
        <institution content-type="orgname">Santa Casa da Bahia</institution>
        <addr-line>
          <named-content content-type="city">Salvador</named-content>
          <named-content content-type="state">BA</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Hospital Santa Izabel - Santa Casa da Bahia , Salvador , BA &#8211; Brasil </institution>
      </aff>
      <aff id="aff8">
        <label> 8 </label>
        <institution content-type="orgname">Hospital Alian&#231;a</institution>
        <addr-line>
          <named-content content-type="city">Salvador</named-content>
          <named-content content-type="state">BA</named-content>
        </addr-line>
        <country country="BR">Brasil</country>
        <institution content-type="original"> Hospital Alian&#231;a , Salvador , BA &#8211; Brasil </institution>
      </aff>
      <author-notes>
        <corresp id="c01">Correspond&#234;ncia: Gustavo Eug&#234;nio Martins Marinho &#8226; Rodovia MG-179 Km 0 s/n. CEP 37132-440, Bairro Trevo Alfenas, Alfenas, MG &#8211; Brasil E-mail: <email>gustavoeugenio@cardiol.br</email> </corresp>
        <fn fn-type="conflict">
          <p>Potencial conflito de interesse</p>
          <p>N&#227;o h&#225; conflito com o presente artigo</p>
        </fn>
      </author-notes>
      <pub-date date-type="pub" publication-format="electronic">
        <day>25</day>
        <month>11</month>
        <year>2022</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>10</month>
        <year>2022</year>
      </pub-date>
      <volume>119</volume>
      <issue>5 suppl 1</issue>
      <fpage>6</fpage>
      <lpage>13</lpage>
      <history>
        <date date-type="received">
          <day>17</day>
          <month>05</month>
          <year>2022</year>
        </date>
        <date date-type="rev-recd">
          <day>13</day>
          <month>07</month>
          <year>2022</year>
        </date>
        <date date-type="accepted">
          <day>20</day>
          <month>07</month>
          <year>2022</year>
        </date>
      </history>
      <permissions>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/" xml:lang="en">
          <license-p>This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.</license-p>
        </license>
      </permissions>
      <abstract>
        <title>Resumo</title>
        <sec>
          <title>Fundamento</title>
          <p>A Sociedade Brasileira de Cardiologia promove anualmente a prova para obten&#231;&#227;o do t&#237;tulo de especialista em Cardiologia, sendo a Comiss&#227;o Julgadora do T&#237;tulo de Especialista em Cardiologia respons&#225;vel pela sua organiza&#231;&#227;o e aplica&#231;&#227;o. A an&#225;lise psicom&#233;trica dos itens de uma prova, por meio da Teoria de Resposta ao Item (TRI) pode fornecer informa&#231;&#245;es robustas e contribuir para o aprimoramento cont&#237;nuo dessa avalia&#231;&#227;o.</p>
        </sec>
        <sec>
          <title>Objetivos</title>
          <p>Avaliar as propriedades psicom&#233;tricas da prova do T&#237;tulo de Especialista em Cardiologia no ano de 2019, em rela&#231;&#227;o aos par&#226;metros da TRI.</p>
        </sec>
        <sec>
          <title>M&#233;todos</title>
          <p>Estudo observacional, com a an&#225;lise psicom&#233;trica das 120 quest&#245;es da prova realizada por 1120 (mil cento e vinte) candidatos para a obten&#231;&#227;o do t&#237;tulo de especialista em Cardiologia, no ano de 2019.</p>
        </sec>
        <sec>
          <title>Resultados</title>
          <p>A an&#225;lise da prova pela TRI mostrou 32,2% dos itens com poder de discrimina&#231;&#227;o &#8220;alto&#8221; ou &#8220;muito alto&#8221;, 49,2% dos itens categorizados como &#8220;f&#225;cil&#8221; ou &#8220;muito f&#225;cil&#8221; e 41,5% apresentavam alta probabilidade de <italic>acerto ao acaso</italic> . Foram identificados 69 itens com problemas em rela&#231;&#227;o aos par&#226;metros da TRI e, portanto, com baixo poder de avaliar a profici&#234;ncia do candidato.</p>
        </sec>
        <sec>
          <title>Conclus&#245;es</title>
          <p>A an&#225;lise psicom&#233;trica da prova de t&#237;tulo de Especialista em Cardiologia pela TRI demonstrou um alto percentual de quest&#245;es f&#225;ceis, com cerca de dois ter&#231;os dos itens com alta probabilidade de <italic>acerto ao acaso.</italic> Esses dados poder&#227;o desencadear uma s&#233;rie de discuss&#245;es e propostas para a constru&#231;&#227;o das futuras provas em cardiologia.</p>
        </sec>
      </abstract>
      <kwd-group xml:lang="pt">
        <kwd>Especializa&#231;&#227;o</kwd>
        <kwd>Cardiologia</kwd>
        <kwd>Psicometria</kwd>
      </kwd-group>
      <counts>
        <fig-count count="10"/>
        <table-count count="8"/>
        <equation-count count="4"/>
        <ref-count count="4"/>
        <page-count count="8"/>
      </counts>
    </article-meta>
  </front>
  <body>
    <sec sec-type="intro">
      <title>Introdu&#231;&#227;o</title>
      <p>A busca pelo t&#237;tulo de especialista tem se tornado uma constante entre os m&#233;dicos brasileiros. As raz&#245;es envolvem desde o ganho de conhecimento, permiss&#227;o para participar de concursos, at&#233; o ingresso em cooperativas m&#233;dicas no mercado de trabalho, tornando-se evidente que a titula&#231;&#227;o aumenta o prest&#237;gio do profissional e de sua especialidade.</p>
      <p>O T&#237;tulo de Especialista em Cardiologia (TEC) existe na Sociedade Brasileira de Cardiologia (SBC) desde 1968, por&#233;m somente foi regulamentado pela Associa&#231;&#227;o M&#233;dica Brasileira (AMB) e pelo Conselho Federal de Medicina (CFM) por meio da Resolu&#231;&#227;o n. 1286/89. Nesse contexto, surge, em 1992, a Comiss&#227;o Julgadora do T&#237;tulo de Especialista em Cardiologia (CJTEC). <sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
      <p>A prova do TEC consta de 120 quest&#245;es de m&#250;ltipla escolha, com cinco alternativas cada, sendo apenas uma correta. H&#225; uma preocupa&#231;&#227;o em rela&#231;&#227;o ao grau de dificuldade das quest&#245;es selecionadas para a prova e, dessa forma, a comiss&#227;o classifica caracteriza-as como dif&#237;ceis, m&#233;dias e f&#225;ceis. No entanto, at&#233; o momento, essa classifica&#231;&#227;o &#233; feita de modo subjetivo, ou seja, de acordo com a opini&#227;o dos membros da CJTEC e n&#227;o atrav&#233;s de uma metodologia psicom&#233;trica, que avalia o grau de dificuldade para quem faz a prova. <sup><xref ref-type="bibr" rid="B2">2</xref></sup></p>
      <p>A Teoria de Resposta ao Item (TRI), tem sido utilizada recentemente como avalia&#231;&#227;o psicom&#233;trica na an&#225;lise e na interpreta&#231;&#227;o dos resultados nos diversos cen&#225;rios de provas e concursos. <sup><xref ref-type="bibr" rid="B2">2</xref></sup></p>
      <p>At&#233; o momento, n&#227;o foi realizada nenhuma avalia&#231;&#227;o psicom&#233;trica do exame do TEC e, em virtude da import&#226;ncia da prova, &#233; imprescind&#237;vel conhecer se esse modelo de avalia&#231;&#227;o fornece realmente uma medida coerente e confi&#225;vel do ponto de vista t&#233;cnico. Diante do exposto, esse estudo teve como objetivo avaliar as propriedades psicom&#233;tricas da prova do TEC no ano de 2019, em rela&#231;&#227;o &#224; TRI.</p>
    </sec>
    <sec sec-type="methods">
      <title>M&#233;todos</title>
      <sec>
        <title>Desenho do estudo</title>
        <p>Foi um estudo observacional, com a an&#225;lise psicom&#233;trica das 120 quest&#245;es da prova realizada por 1120 candidatos, para a obten&#231;&#227;o do TEC, no dia 27 de outubro de 2019, das 13:00 &#224;s 18:00 horas, em uma Universidade Privada de S&#227;o Paulo.</p>
      </sec>
      <sec>
        <title>Crit&#233;rios de inclus&#227;o e exclus&#227;o</title>
        <p>Foram inclu&#237;dos todos os gabaritos entregues pelos candidatos que realizaram a prova para obten&#231;&#227;o do TEC em 2019. Foram exclu&#237;das duas quest&#245;es ap&#243;s a fase recursal, al&#233;m da prova de um candidato que respondeu apenas duas quest&#245;es.</p>
      </sec>
      <sec>
        <title>Amostra</title>
        <p>A Amostra, que teve duas quest&#245;es exclu&#237;das na fase recursal, foi constitu&#237;da por gabaritos de 118 quest&#245;es, que foram respondidas e entregues pelos m&#233;dicos que realizaram a prova para a obten&#231;&#227;o do TEC no ano de 2019.</p>
      </sec>
      <sec>
        <title>Coleta de Dados</title>
        <p>Os dados dos candidatos foram extra&#237;dos diretamente do banco de dados da empresa respons&#225;vel pela elabora&#231;&#227;o da prova (Segmento Farma Editores Ltda., em parceria com a Simples Detalhe Assessoria, Planejamento e organiza&#231;&#227;o de Eventos Ltda. e a Picsis inform&#225;tica ind&#250;stria e com&#233;rcio Ltda.), e organizados em planilhas Microsoft Excel <sup>&#174;</sup> .</p>
        <p>A partir desses dados, foram geradas planilhas separadas para os dados de identifica&#231;&#227;o e para os dados referentes &#224;s notas de cada prova. Os nomes dos candidatos foram exclu&#237;dos das planilhas, com o objetivo de manter o sigilo da pesquisa, sendo a identifica&#231;&#227;o de cada candidato feita por um n&#250;mero.</p>
      </sec>
      <sec>
        <title>Aspectos &#233;ticos</title>
        <p>Foram utilizadas bases de dados secund&#225;rias, sem identifica&#231;&#227;o dos participantes, de modo que n&#227;o houve necessidade de utiliza&#231;&#227;o de Termo de Consentimento Livre e Esclarecido (TCLE). No entanto, para a obten&#231;&#227;o do banco de dados, foi firmado um Termo de Consentimento de Utiliza&#231;&#227;o do Banco de Dados (TCUD), o qual foi encaminhado, inicialmente, para a SBC e, posteriormente, ao Comit&#234; de &#201;tica em Pesquisa (CEP), com n&#250;mero do parecer: 4.030.702.</p>
      </sec>
      <sec>
        <title>An&#225;lise estat&#237;stica</title>
        <p>Realizamos uma avalia&#231;&#227;o psicom&#233;trica da prova aplicada aos candidatos ao TEC em 2019 pela SBC, utilizando a TRI. A TRI busca determinar o n&#237;vel de aptid&#227;o do candidato (tra&#231;o latente, par&#226;metro <italic>teta</italic> (&#952;)] e a probabilidade de um indiv&#237;duo com certo n&#237;vel de aptid&#227;o em responder acertadamente os itens conforme o grau de dificuldade.</p>
        <p>Para an&#225;lise do tra&#231;o latente, a TRI avalia os seguintes par&#226;metros:</p>
        <list list-type="alpha-lower">
          <list-item>
            <p>Item Discrimina&#231;&#227;o ( <italic>a</italic> ): consiste na avalia&#231;&#227;o da aptid&#227;o do item em distinguir indiv&#237;duos com habilidades diferentes;</p>
          </list-item>
          <list-item>
            <p>Item Dificuldade ( <italic>b</italic> ): trata da habilidade m&#237;nima que um respondente precisa para ter uma grande probabilidade de dar a resposta correta;</p>
          </list-item>
          <list-item>
            <p>Acerto ao Acaso ( <italic>c</italic> ): a probabilidade de um respondente com baixa profici&#234;ncia responder corretamente um item.</p>
          </list-item>
        </list>
        <p>Desta forma, a TRI procura medir vari&#225;veis n&#227;o observ&#225;veis (tra&#231;o latente) que possam influenciar as respostas dadas aos itens, utilizando a aferi&#231;&#227;o das vari&#225;veis observ&#225;veis (respostas aos itens), e estabelecendo uma rela&#231;&#227;o entre a habilidade do respondente e os par&#226;metros do item com a probabilidade de acerto ao item. Assim, quanto maior a aptid&#227;o de um candidato, maior ser&#225; sua probabilidade de responder corretamente o item no instrumento de avalia&#231;&#227;o.</p>
        <p>Duas importantes suposi&#231;&#245;es da TRI s&#227;o a Unidimensionalidade que postula que h&#225; apenas uma aptid&#227;o (&#952;) respons&#225;vel pela realiza&#231;&#227;o de um conjunto de itens de um teste, e a Independ&#234;ncia local, que implica que o desempenho do sujeito em um item n&#227;o afeta o desempenho em outro, pois cada item &#233; respondido em fun&#231;&#227;o da aptid&#227;o dominante (&#952;) do candidato para aquele item.</p>
        <p>O modelo estat&#237;stico da TRI predominantemente utilizado no Brasil &#233; o modelo Log&#237;stico unidimensional de tr&#234;s par&#226;metros. Os modelos Log&#237;sticos unidimensional com um e dois par&#226;metros n&#227;o s&#227;o convenientes na an&#225;lise do presente estudo uma vez que, pelos resultados obtidos no modelo de tr&#234;s par&#226;metros, o <italic>acerto ao acaso</italic> variou muito entre os 120 itens da prova aplicada em 2019.</p>
      </sec>
      <sec>
        <title>Metodologia de c&#225;lculos da TRI:</title>
        <p>Modelo log&#237;stico Unidimensional da TRI com 3 par&#226;metros</p>
        <disp-formula id="e1">
          <mml:math>
            <mml:mi>P</mml:mi>
            <mml:mfenced close=")" open="(" separators="|">
              <mml:mrow>
                <mml:msub>
                  <mml:mi>U</mml:mi>
                  <mml:mrow>
                    <mml:mi>i</mml:mi>
                    <mml:mi>j</mml:mi>
                  </mml:mrow>
                </mml:msub>
                <mml:mo>=</mml:mo>
                <mml:mn>1</mml:mn>
                <mml:mo>&#8739;</mml:mo>
                <mml:msub>
                  <mml:mi>&#952;</mml:mi>
                  <mml:mi>j</mml:mi>
                </mml:msub>
              </mml:mrow>
            </mml:mfenced>
            <mml:mo>=</mml:mo>
            <mml:msub>
              <mml:mi>c</mml:mi>
              <mml:mi>i</mml:mi>
            </mml:msub>
            <mml:mo>+</mml:mo>
            <mml:mfenced close=")" open="(" separators="|">
              <mml:mrow>
                <mml:mn>1</mml:mn>
                <mml:mo>&#8722;</mml:mo>
                <mml:msub>
                  <mml:mi>c</mml:mi>
                  <mml:mi>i</mml:mi>
                </mml:msub>
              </mml:mrow>
            </mml:mfenced>
            <mml:mfrac>
              <mml:mn>1</mml:mn>
              <mml:mrow>
                <mml:mn>1</mml:mn>
                <mml:mo>+</mml:mo>
                <mml:msup>
                  <mml:mi>e</mml:mi>
                  <mml:mrow>
                    <mml:mo>&#8722;</mml:mo>
                    <mml:mi>D</mml:mi>
                    <mml:msub>
                      <mml:mi>a</mml:mi>
                      <mml:mi>i</mml:mi>
                    </mml:msub>
                    <mml:mfenced close=")" open="(" separators="|">
                      <mml:mrow>
                        <mml:msub>
                          <mml:mi>&#952;</mml:mi>
                          <mml:mi>j</mml:mi>
                        </mml:msub>
                        <mml:mo>&#8722;</mml:mo>
                        <mml:msub>
                          <mml:mi>b</mml:mi>
                          <mml:mi>i</mml:mi>
                        </mml:msub>
                      </mml:mrow>
                    </mml:mfenced>
                  </mml:mrow>
                </mml:msup>
              </mml:mrow>
            </mml:mfrac>
          </mml:math>
        </disp-formula>
        <p>com i = 1, 2, ..., I e j = 1, 2, ..., n, onde:</p>
        <list list-type="simple">
          <list-item>
            <p>- <bold>U</bold> <sub><bold>ij</bold></sub> &#233; uma vari&#225;vel dicot&#244;mica que assume os valores 1, quando o indiv&#237;duo j responde corretamente o item i, ou 0 quando o indiv&#237;duo j n&#227;o responde corretamente ao item i.</p>
          </list-item>
          <list-item>
            <p>- <bold>&#952;</bold> <sub><bold>j</bold></sub> representa a habilidade (tra&#231;o latente) do j-&#233;simo indiv&#237;duo.</p>
          </list-item>
          <list-item>
            <p>- <bold>P(U</bold> <sub><bold>ij</bold></sub> <bold>=1</bold> |&#952; <sub><bold>j</bold></sub> <bold>)</bold> &#233; a probabilidade de um indiv&#237;duo j com habilidade &#952;j responder corretamente o item i e &#233; chamada de Fun&#231;&#227;o de Resposta do Item &#8211; FRI.</p>
          </list-item>
          <list-item>
            <p>- <bold>b</bold> <sub><bold>i</bold></sub> &#233; o par&#226;metro de dificuldade (ou de posi&#231;&#227;o) do item i, medido na mesma escala da habilidade.</p>
          </list-item>
          <list-item>
            <p>- <bold>a</bold> <sub><bold>i</bold></sub> &#233; o par&#226;metro de discrimina&#231;&#227;o (ou de inclina&#231;&#227;o) do item i, com valor proporcional &#224; inclina&#231;&#227;o da Curva Caracter&#237;stica do Item &#8212; CCI no ponto bi.</p>
          </list-item>
          <list-item>
            <p>- <bold>c</bold> <sub><bold>i</bold></sub> &#233; o par&#226;metro do item que representa a probabilidade de indiv&#237;duos com baixa habilidade responderem corretamente o item i (muitas vezes referido como a probabilidade de acerto casual).</p>
          </list-item>
          <list-item>
            <p>- <bold>D</bold> &#233; um fator de escala, constante e igual a 1.</p>
          </list-item>
        </list>
        <p>Os valores dos par&#226;metros <italic>a, b e c</italic> , s&#227;o calculados por meio de pr&#233;-testagens (calibragem) dos itens, utilizando o m&#233;todo da m&#225;xima verossimilhan&#231;a ( <bold>L</bold> , de <italic>likelihood</italic> ), que trabalha com as derivadas. Sua f&#243;rmula &#233; a seguinte:</p>
        <disp-formula id="e2">
          <mml:math>
            <mml:mrow>
              <mml:mi mathvariant="normal">L</mml:mi>
            </mml:mrow>
            <mml:mfenced close=")" open="(" separators="|">
              <mml:mrow>
                <mml:msub>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">u</mml:mi>
                  </mml:mrow>
                  <mml:mrow>
                    <mml:mn>1</mml:mn>
                    <mml:msup>
                      <mml:mi>s</mml:mi>
                      <mml:mrow>
                        <mml:mi>&#8242;</mml:mi>
                      </mml:mrow>
                    </mml:msup>
                  </mml:mrow>
                </mml:msub>
                <mml:msub>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">u</mml:mi>
                  </mml:mrow>
                  <mml:mrow>
                    <mml:mn>2</mml:mn>
                    <mml:msup>
                      <mml:mi>s</mml:mi>
                      <mml:mrow>
                        <mml:msup>
                          <mml:mi>s</mml:mi>
                          <mml:mrow>
                            <mml:mi>&#8242;</mml:mi>
                          </mml:mrow>
                        </mml:msup>
                      </mml:mrow>
                    </mml:msup>
                  </mml:mrow>
                </mml:msub>
                <mml:mo>&#8230;</mml:mo>
                <mml:mo>,</mml:mo>
                <mml:msub>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">u</mml:mi>
                  </mml:mrow>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">n</mml:mi>
                      <mml:mi mathvariant="normal">s</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                </mml:msub>
                <mml:mo>&#8739;</mml:mo>
                <mml:mi>&#952;</mml:mi>
              </mml:mrow>
            </mml:mfenced>
            <mml:mo>=</mml:mo>
            <mml:munderover>
              <mml:mrow>
                <mml:mo>&#8719;</mml:mo>
              </mml:mrow>
              <mml:mrow>
                <mml:mi>i</mml:mi>
                <mml:mo>=</mml:mo>
                <mml:mn>1</mml:mn>
              </mml:mrow>
              <mml:mi>n</mml:mi>
            </mml:munderover>
            <mml:mrow>
              <mml:mo> </mml:mo>
            </mml:mrow>
            <mml:mrow>
              <mml:mi mathvariant="normal">P</mml:mi>
              <mml:mi mathvariant="normal">i</mml:mi>
            </mml:mrow>
            <mml:mo stretchy="false">(</mml:mo>
            <mml:mi>&#952;</mml:mi>
            <mml:mrow>
              <mml:mi mathvariant="normal">s</mml:mi>
            </mml:mrow>
            <mml:msup>
              <mml:mo stretchy="false">)</mml:mo>
              <mml:mrow>
                <mml:msub>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">u</mml:mi>
                  </mml:mrow>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">s</mml:mi>
                      <mml:mi mathvariant="normal">i</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                </mml:msub>
              </mml:mrow>
            </mml:msup>
            <mml:mrow>
              <mml:mi mathvariant="normal">Q</mml:mi>
              <mml:mi mathvariant="normal">i</mml:mi>
            </mml:mrow>
            <mml:mo stretchy="false">(</mml:mo>
            <mml:mi>&#952;</mml:mi>
            <mml:mrow>
              <mml:mi mathvariant="normal">s</mml:mi>
            </mml:mrow>
            <mml:msup>
              <mml:mo stretchy="false">)</mml:mo>
              <mml:mrow>
                <mml:mn>1</mml:mn>
                <mml:mo>&#8722;</mml:mo>
                <mml:msub>
                  <mml:mi>u</mml:mi>
                  <mml:mrow>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">s</mml:mi>
                      <mml:mi mathvariant="normal">i</mml:mi>
                    </mml:mrow>
                  </mml:mrow>
                </mml:msub>
              </mml:mrow>
            </mml:msup>
          </mml:math>
        </disp-formula>
        <p>Em que:</p>
        <list list-type="simple">
          <list-item>
            <p>- <bold>i</bold> = 1, 2, ..., n itens</p>
          </list-item>
          <list-item>
            <p>- <bold>u</bold> <sub><bold>is</bold></sub> = resposta do sujeito a cada item (1 = acertou, 0 = errou)</p>
          </list-item>
        </list>
        <p>Para se calcular a aptid&#227;o/profici&#234;ncia de um candidato, temos que achar o m&#225;ximo da fun&#231;&#227;o acima. Primeiramente, identifica-se a probabilidade de acerto [(Pi(&#952;)] de cada um dos itens do teste utilizando um dos tr&#234;s modelos da TRI (1PL, 2PL, 3PL &#8211; na presente pesquisa foi utilizado o modelo de tr&#234;s par&#226;metros &#8211; 3PL). Posteriormente, de forma emp&#237;rica, substituem-se os valores de &#952; numa faixa de - 5 a + 5 (-5,00 &#8804; &#952; &#8804; +5,00, normalmente, utiliza-se -3,00 &#8804; &#952; &#8804; +3,00) ou emprega-se o algoritmo de itera&#231;&#227;o de Newton-Raphson para se calcular o m&#225;ximo da fun&#231;&#227;o L. Esse m&#225;ximo, com base nos valores de &#952;, &#233; a aptid&#227;o/profici&#234;ncia do candidato obtido no teste aplicado.</p>
      </sec>
      <sec>
        <title>Curva Caracter&#237;stica do Item (CCI)</title>
        <p>O modelo matem&#225;tico que define a TRI &#233; uma fun&#231;&#227;o de probabilidade. Portanto, sua imagem estar&#225; sempre no intervalo [0,1]. O n&#250;mero U <sub>ij</sub> =1|&#952; <sub>j</sub> ) pode ser identificado pela propor&#231;&#227;o de respostas corretas ao item i no grupo de indiv&#237;duos com habilidade &#952;j. Essa rela&#231;&#227;o &#233; descrita por uma curva sigmoide, em que o eixo horizontal representa a escala de aptid&#227;o e o eixo vertical a probabilidade do indiv&#237;duo com uma habilidade &#952;j dar a resposta correta ao item i. Podemos destacar duas ass&#237;ntotas horizontais e notam-se, com uma certa precis&#227;o, os tr&#234;s par&#226;metros do item.</p>
      </sec>
      <sec>
        <title>Curva de Informa&#231;&#227;o &#8211; I(&#952;)</title>
        <p>Precis&#227;o da informa&#231;&#227;o significa a exatid&#227;o que um item representa aquilo que ele pretende medir. Neste contexto, precis&#227;o significa o qu&#227;o bem o item prediz o crit&#233;rio ou representa o tra&#231;o latente (&#952;). Assim, a fun&#231;&#227;o de informa&#231;&#227;o da TRI segue o c&#225;lculo do erro de estima&#231;&#227;o, isto &#233;, o quanto o escore obtido pelo sujeito num teste se afasta do seu escore verdadeiro. O pr&#243;prio conceito de fun&#231;&#227;o de informa&#231;&#227;o &#233; o rec&#237;proco da vari&#226;ncia, ou seja: I = 1 / S <sup>2</sup> . A fun&#231;&#227;o de informa&#231;&#227;o corresponde ao conceito de carga fatorial do item da an&#225;lise fatorial, na vis&#227;o do modelo do tra&#231;o latente, pois a carga fatorial representa a covari&#226;ncia entre o item (representa&#231;&#227;o comportamental) e o tra&#231;o latente (teta). A curva de informa&#231;&#227;o do teste mostra a quantidade de informa&#231;&#227;o fornecida pelo teste a um certo n&#237;vel teta; apresenta a amplitude do teta para a qual o teste fornece informa&#231;&#227;o confi&#225;vel, dizendo que fora dessa amplitude o teste produz mais informa&#231;&#227;o err&#244;nea (erro) sobre o teta que informa&#231;&#227;o correta. Assim, a curva de informa&#231;&#227;o tem interface com ambos os par&#226;metros dos testes, isto &#233;, validade e precis&#227;o, mas n&#227;o se confunde com nenhum dos dois. A representa&#231;&#227;o dessa informa&#231;&#227;o do item assemelha-se a uma curva de tipo normal (em forma de &#8220;sino&#8221;).</p>
        <p>Na presente an&#225;lise, foi adotado o crit&#233;rio de, pelo menos, 25% de &#8220;acerto ao acaso&#8221; como insatisfat&#243;rio para um determinado item da prova. Assim, como foram aplicadas 1120 provas, 5% a mais de &#8220;acerto ao acaso&#8221; al&#233;m do esperado (20%) &#233; considerado muito alto e, portanto, o item avaliado apresenta algum problema em sua formula&#231;&#227;o ou nas op&#231;&#245;es de respostas. O &#8220;acerto ao acaso&#8221; &#233; demonstr&#225;vel pela falta de coer&#234;ncia do candidato, em errar itens f&#225;ceis e, de modo contradit&#243;rio, acertar itens dif&#237;ceis, teoricamente sem profici&#234;ncia para tal.</p>
      </sec>
    </sec>
    <sec sec-type="results">
      <title>Resultados</title>
      <p>Os resultados apresentados referem-se &#224; an&#225;lise psicom&#233;trica dos 118 itens da prova aplicada aos candidatos ao TEC em 2019, utilizando um modelo log&#237;stico unidimensional de tr&#234;s par&#226;metros da TRI: discrimina&#231;&#227;o (a), dificuldade (b) e acerto ao acaso (c).</p>
      <p>Durante a an&#225;lise, verificou-se que um item (Quest&#227;o n&#186; 110) apresentou n&#237;vel negativo para o par&#226;metro <italic>discrimina&#231;&#227;o</italic> ( <italic>a</italic> = - 0,174), sugerindo que quanto maior o n&#237;vel de conhecimento do candidato menor a chance de acertar o item, um resultado incoerente com o objetivo do par&#226;metro, por essa raz&#227;o, esse item n&#227;o foi inclu&#237;do no modelo da an&#225;lise final.</p>
      <p>A <xref ref-type="table" rid="t1">Tabela 1</xref> apresenta a distribui&#231;&#227;o dos 118 itens da prova em rela&#231;&#227;o ao par&#226;metro <italic>discrimina&#231;&#227;o</italic> . Observa-se que 18,7% desses apresentavam &#8220;muito baixo&#8221; ou &#8220;baixo&#8221; poder de <italic>discrimina&#231;&#227;o</italic> (a &#8804; 0,65); 49,1% &#8220;moderada&#8221; <italic>discrimina&#231;&#227;o</italic> (0,651 &lt; a &#8804; 1,350) e 32,2% dos itens apresentavam &#8220;alto&#8221; ou &#8220;muito alto&#8221; poder de <italic>discrimina&#231;&#227;o</italic> (a &#8805; 1,351).</p>
      <p>
        <table-wrap id="t1">
          <label>Tabela 1</label>
          <caption>
            <title>Distribui&#231;&#227;o dos itens da prova em rela&#231;&#227;o ao par&#226;metro <italic>discrimina&#231;&#227;o</italic> da Teoria de Resposta ao Item (TRI)</title>
          </caption>
          <table frame="hsides" rules="groups">
            <colgroup>
              <col/>
              <col/>
              <col/>
            </colgroup>
            <thead>
              <tr>
                <th align="left">Classifica&#231;&#227;o do poder de discrimina&#231;&#227;o (a)</th>
                <th>Frequ&#234;ncia (n)</th>
                <th>%</th>
              </tr>
            </thead>
            <tbody>
              <tr>
                <td>&#8804; 0,35 (muito baixa)</td>
                <td align="center">12</td>
                <td align="center">10,2</td>
              </tr>
              <tr>
                <td>De 0,351 a 0,650 (baixa)</td>
                <td align="center">10</td>
                <td align="center">8,5</td>
              </tr>
              <tr>
                <td>De 0,651 a 1,350 (moderada)</td>
                <td align="center">58</td>
                <td align="center">49,1</td>
              </tr>
              <tr>
                <td>De 1,351 a 1,700 (alta)</td>
                <td align="center">25</td>
                <td align="center">21,2</td>
              </tr>
              <tr>
                <td>&gt; 1,700 (muito alta)</td>
                <td align="center">13</td>
                <td align="center">11,0</td>
              </tr>
              <tr>
                <td>Total</td>
                <td align="center">118</td>
                <td align="center">100,0</td>
              </tr>
            </tbody>
          </table>
          <table-wrap-foot>
            <fn id="TFN1">
              <p>Base de Dados: 1.120 candidatos. Nota: 2 itens da prova anulados (itens 23 e 46).</p>
            </fn>
          </table-wrap-foot>
        </table-wrap>
      </p>
      <p>A <xref ref-type="table" rid="t3">Tabela 2</xref> apresenta a distribui&#231;&#227;o dos 118 itens da prova em rela&#231;&#227;o ao par&#226;metro <italic>dificuldade</italic> . Observa-se que 49,2% desses foram classificados como &#8220;f&#225;cil&#8221; ou &#8220;muito f&#225;cil&#8221; ( <italic>b</italic> &lt; -0,52); 22,0% classificados com <italic>dificuldade</italic> moderada (-0,51 &#8804; <italic>b</italic> &#8804; 0,51); e 28,8% dos itens foram classificados como dif&#237;cil ou muito dif&#237;cil ( <italic>b</italic> &#8805; 0,52).</p>
      <p>
        <table-wrap id="t3">
          <label>Tabela 2</label>
          <caption>
            <title>Distribui&#231;&#227;o dos itens da prova em rela&#231;&#227;o ao par&#226;metro <italic>dificuldade</italic> pela TRI</title>
          </caption>
          <table frame="hsides" rules="groups">
            <colgroup>
              <col/>
              <col/>
              <col/>
            </colgroup>
            <thead>
              <tr>
                <th align="left">Classifica&#231;&#227;o do par&#226;metro dificuldade (b)</th>
                <th>Frequ&#234;ncia (n)</th>
                <th>%</th>
              </tr>
            </thead>
            <tbody>
              <tr>
                <td>&#8804; -1,28 (Muito f&#225;cil)</td>
                <td align="center">31</td>
                <td align="center">26,3</td>
              </tr>
              <tr>
                <td>De -1,27 a - 0,52 (F&#225;cil)</td>
                <td align="center">27</td>
                <td align="center">22,9</td>
              </tr>
              <tr>
                <td>De -0,51 a 0,51 (Moderada)</td>
                <td align="center">26</td>
                <td align="center">22,0</td>
              </tr>
              <tr>
                <td>De 0,52 a 1,27 (Dif&#237;cil)</td>
                <td align="center">19</td>
                <td align="center">16,1</td>
              </tr>
              <tr>
                <td>&#8805; 1,28 (Muito dif&#237;cil)</td>
                <td align="center">15</td>
                <td align="center">12,7</td>
              </tr>
              <tr>
                <td>Total</td>
                <td align="center">118</td>
                <td align="center">100,0</td>
              </tr>
            </tbody>
          </table>
          <table-wrap-foot>
            <attrib>Fonte: Elaborada pelos autores; Base de Dados: 1.120 candidatos. Nota: 2 itens da prova anulados (itens 23 e 46).</attrib>
          </table-wrap-foot>
        </table-wrap>
      </p>
      <p>A <xref ref-type="table" rid="t2">Tabela 3</xref> apresenta a distribui&#231;&#227;o dos 118 itens da prova em rela&#231;&#227;o ao par&#226;metro <italic>acerto ao acaso</italic> , onde observa-se que 41,5% dos itens apresentaram alta probabilidade de <italic>acerto ao acaso</italic> , conforme a metodologia da TRI.</p>
      <p>
        <table-wrap id="t2">
          <label>Tabela 3</label>
          <caption>
            <title>Distribui&#231;&#227;o dos itens da prova em rela&#231;&#227;o aos percentuais de <italic>acertos ao acaso</italic> pela teoria de resposta ao item</title>
          </caption>
          <table frame="hsides" rules="groups">
            <colgroup>
              <col/>
              <col/>
              <col/>
            </colgroup>
            <thead>
              <tr>
                <th align="left">Percentual de acertos ao acaso (c)</th>
                <th>Frequ&#234;ncia (n)</th>
                <th>%</th>
              </tr>
            </thead>
            <tbody>
              <tr>
                <td>&#8804; 10,0%</td>
                <td align="center">48</td>
                <td align="center">40,7</td>
              </tr>
              <tr>
                <td>De 10,1 a 25,0%</td>
                <td align="center">21</td>
                <td align="center">17,8</td>
              </tr>
              <tr>
                <td>De 25,1 a 40,0%</td>
                <td align="center">20</td>
                <td align="center">16,9</td>
              </tr>
              <tr>
                <td>De 40,1 a 60,0%</td>
                <td align="center">19</td>
                <td align="center">16,1</td>
              </tr>
              <tr>
                <td>&gt; 60,0%</td>
                <td align="center">10</td>
                <td align="center">8,5</td>
              </tr>
              <tr>
                <td>Total</td>
                <td align="center">118</td>
                <td align="center">100,0</td>
              </tr>
            </tbody>
          </table>
          <table-wrap-foot>
            <attrib>Fonte: Elaborada pelos autores; Base de Dados: 1.120 candidatos. Nota: 2 itens da prova anulados (nos 23 e 46).</attrib>
          </table-wrap-foot>
        </table-wrap>
      </p>
      <p>Sobre a CCI, 58,5% dos itens foram considerados insatisfat&#243;rios. J&#225; em rela&#231;&#227;o a curva de informa&#231;&#227;o, 78.8% dos itens foram satisfat&#243;rios ( <xref ref-type="table" rid="t4">Tabela 4</xref> ).</p>
      <p>
        <table-wrap id="t4">
          <label>Tabela 4</label>
          <caption>
            <title>Distribui&#231;&#227;o dos itens da prova segundo classifica&#231;&#227;o (satisfat&#243;ria ou insatisfat&#243;ria) da Curva Caracter&#237;stica do Item e da Curva de Informa&#231;&#227;o da teoria de resposta ao item</title>
          </caption>
          <table frame="hsides" rules="groups">
            <colgroup>
              <col/>
              <col/>
              <col/>
            </colgroup>
            <thead>
              <tr>
                <th align="left">Curva Caracter&#237;stica do Item</th>
                <th>Frequ&#234;ncia (n)</th>
                <th>%</th>
              </tr>
            </thead>
            <tbody>
              <tr>
                <td>Satisfat&#243;ria</td>
                <td align="center">49</td>
                <td align="center">41,5</td>
              </tr>
              <tr>
                <td>Insatisfat&#243;ria</td>
                <td align="center">69</td>
                <td align="center">58,5</td>
              </tr>
              <tr>
                <td>
                  <bold>Curva de Informa&#231;&#227;o</bold>
                </td>
                <td align="center">
                  <bold>Frequ&#234;ncia (n)</bold>
                </td>
                <td align="center">
                  <bold>%</bold>
                </td>
              </tr>
              <tr>
                <td>Satisfat&#243;ria</td>
                <td align="center">93</td>
                <td align="center">78,8</td>
              </tr>
              <tr>
                <td>Insatisfat&#243;ria</td>
                <td align="center">25</td>
                <td align="center">21,2</td>
              </tr>
            </tbody>
          </table>
          <table-wrap-foot>
            <attrib>Fonte: Elaborada pelos autores; Base de Dados: 1.120 candidatos. Nota: 2 itens da prova anulados (nos 23 e 46).</attrib>
          </table-wrap-foot>
        </table-wrap>
      </p>
      <p>A an&#225;lise individual dos itens da prova pela TRI identificou 69 itens que apresentavam algum tipo de problema em rela&#231;&#227;o aos tr&#234;s par&#226;metros e, portanto, considerados com baixo poder de gerar informa&#231;&#227;o em rela&#231;&#227;o &#224; identifica&#231;&#227;o do tra&#231;o latente (&#952;), que avalia a profici&#234;ncia do candidato. Assim, os 49 itens restantes da prova foram analisados pela TRI e comparados ao modelo inicial da prova com 118 itens.</p>
      <p>A <xref ref-type="fig" rid="f01">Figura 1</xref> apresenta a CCI considerando os 118 itens da prova pelo modelo da TRI. O resultado mostra que quanto maior a aptid&#227;o (&#952;) do candidato, maior ser&#225; o n&#250;mero de itens com resposta correta. Um candidato com aptid&#227;o igual a 0 (&#952; = 0 &#8211; aptid&#227;o mediana, &#952; entre -1 e +1) &#233; esperado que acerte, aproximadamente, 80 dos 118 itens da prova (67,8%). Al&#233;m disso, um candidato com muito baixo n&#237;vel de aptid&#227;o (&#952; &lt; - 4,0) &#233; esperado que acerte pelo menos 36 dos 118 itens da prova (30,5%).</p>
      <p>
        <fig id="f01">
          <label>Figura 1</label>
          <caption>
            <title>Nota: T(&#952;) &#8211; estimada pela teoria de resposta ao item considerando-se os 118 itens da prova, para cada candidato de acordo com sua aptid&#227;o (&#952;).</title>
          </caption>
          <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf01.tif"/></fig>
      </p>
      <p>A <italic>Curva de Informa&#231;&#227;o</italic> ( <xref ref-type="fig" rid="f02">Figura 2</xref> ) para o conjunto dos 118 itens da prova mostra que a quantidade m&#225;xima de informa&#231;&#227;o recuperada sobre o racioc&#237;nio anal&#243;gico do candidato encontra-se em torno da mediana da aptid&#227;o, ou seja, valor de &#952; pr&#243;ximo a 0. Al&#233;m disso, para os valores extremos de &#952;, a prova produz mais erro de informa&#231;&#227;o do que informa&#231;&#227;o leg&#237;tima, e nos valores de &#952; entre -3,2 a +3,1 encontra-se o m&#225;ximo de informa&#231;&#227;o gerado pela avalia&#231;&#227;o.</p>
      <p>
        <fig id="f02">
          <label>Figura 2</label>
          <caption>
            <title>Curva de Informa&#231;&#227;o: I(&#952;) &#8211; e erro padr&#227;o gerados pela teoria de resposta ao item considerando os 118 itens da prova.</title>
          </caption>
          <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf02.tif"/></fig>
      </p>
      <p>A <xref ref-type="fig" rid="f03">Figura 3</xref> mostra a CCI para o conjunto dos 49 itens da prova que restaram ap&#243;s a extra&#231;&#227;o dos itens com problemas nos par&#226;metros da TRI. O resultado mostra que quanto maior a aptid&#227;o (&#952;) do candidato, maior ser&#225; o n&#250;mero de itens com resposta correta. Assim, &#233; esperado que um candidato com aptid&#227;o igual a 0 (&#952; = 0 &#8211; aptid&#227;o mediana, &#952; entre -1 e +1) acerte, aproximadamente, 32 dos 49 itens da prova (65,3%), e um candidato com n&#237;vel de aptid&#227;o muito baixo (&#952; &lt; -4,0) acerte pelo menos quatro dos 49 itens da prova (8,2%). Portanto, considerando os dados da TRI para os 49 itens, os candidatos precisar&#227;o de um n&#237;vel aptid&#227;o (&#952;) maior em rela&#231;&#227;o &#224; exigida para os 118 itens da prova.</p>
      <p>
        <fig id="f03">
          <label>Figura 3</label>
          <caption>
            <title>Nota: T(&#952;) &#8211; estimada pela teoria de resposta o item considerando os 49 itens da prova, para cada candidato de acordo com sua aptid&#227;o (&#952;).</title>
          </caption>
          <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf03.tif"/></fig>
      </p>
      <p>A <italic>Curva de Informa&#231;&#227;o</italic> ( <xref ref-type="fig" rid="f04">Figura 4</xref> ) para os 49 itens da prova mostra que a quantidade de informa&#231;&#227;o m&#225;xima recuperada sobre o racioc&#237;nio anal&#243;gico do candidato encontra-se, tamb&#233;m, em torno da mediana da aptid&#227;o, ou seja, valor de &#952; pr&#243;ximo de 0 (zero). Al&#233;m disso, para os valores extremos dos n&#237;veis de &#952;, a prova produz mais erro de informa&#231;&#227;o do que informa&#231;&#227;o leg&#237;tima, sendo que para valores de &#952; variando entre -4,0 a +3,2 encontra-se o m&#225;ximo de informa&#231;&#227;o gerado pela avalia&#231;&#227;o.</p>
      <p>
        <fig id="f04">
          <label>Figura 4</label>
          <caption>
            <title>Curva de Informa&#231;&#227;o: I(&#952;) &#8211; e erro padr&#227;o gerados pela teoria de resposta ao item considerando os 49 itens da prova.</title>
          </caption>
          <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf04.tif"/></fig>
      </p>
      <p>A <xref ref-type="fig" rid="f05">Figura 5</xref> mostra resultado da profici&#234;ncia gerado pela TRI, considerando-se os 49 itens exclu&#237;dos da prova inicialmente aplicada. Nota-se uma curva t&#237;pica de Gauss, que mostra o n&#237;vel m&#233;dio de profici&#234;ncia dos candidatos com um padr&#227;o normal de distribui&#231;&#227;o.</p>
      <p>
        <fig id="f05">
          <label>Figura 5</label>
          <caption>
            <title>Resultado da profici&#234;ncia gerados pela TRI.</title>
          </caption>
          <attrib>Fonte: Elaborado pelos autores.</attrib>
          <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf05.tif"/></fig>
      </p>
    </sec>
    <sec sec-type="discussion">
      <title>Discuss&#227;o</title>
      <p>O objetivo do presente estudo foi avaliar os itens da prova para obten&#231;&#227;o do TEC em 2019, em rela&#231;&#227;o aos seus par&#226;metros psicom&#233;tricos pela TRI. At&#233; ent&#227;o, o &#250;nico par&#226;metro conhecido pela CJTEC era o grau de dificuldade dos itens, julgados como f&#225;ceis, m&#233;dios ou dif&#237;ceis, com base no conhecimento e na experi&#234;ncia dos integrantes da comiss&#227;o. No entanto esta forma de avalia&#231;&#227;o &#233; subjetiva e carece de validade.</p>
      <p>Em rela&#231;&#227;o ao par&#226;metro <italic>discrimina&#231;&#227;o</italic> , observou-se que apenas 32,2% dos itens apresentavam poder de <italic>discrimina&#231;&#227;o</italic> &#8220;alto&#8221; ou &#8220;muito alto&#8221;. Informa&#231;&#227;o relevante, uma vez que a <italic>discrimina&#231;&#227;o</italic> de um item relaciona-se &#224; sua capacidade de identificar candidatos com profici&#234;ncias distintas, por avaliar a probabilidade de indiv&#237;duos com diferentes profici&#234;ncias responderem corretamente o item. Fatos semelhantes foram observados nas provas do Exame Nacional de Desempenho dos Estudantes (ENADE), nos anos de 2010, 2011 e 2012, que ao serem submetidas &#224; an&#225;lise psicom&#233;trica p&#244;de-se identificar quest&#245;es com baixo poder de discrimina&#231;&#227;o, contribuindo tecnicamente para melhorias na elabora&#231;&#227;o futura de novos itens para as provas do ENADE. <sup><xref ref-type="bibr" rid="B3">3</xref> , <xref ref-type="bibr" rid="B4">4</xref></sup></p>
      <p>A respeito do par&#226;metro <italic>dificuldade</italic> , observou-se que 49,2% dos itens da prova, eram categorizados pela TRI como &#8220;f&#225;cil&#8221; ou &#8220;muito f&#225;cil&#8221; e apenas 22% como de &#8220;moderada&#8221; dificuldade. Isso denota um desbalanceamento da prova em rela&#231;&#227;o &#224; psicometria, que recomenda a seguinte propor&#231;&#227;o de facilidade dos itens: muito f&#225;ceis (10%), f&#225;ceis (20%), moderados (40%), dif&#237;ceis (20%) e muito dif&#237;ceis (10%). <sup><xref ref-type="bibr" rid="B4">4</xref></sup> A propor&#231;&#227;o de itens &#8220;dif&#237;ceis e &#8220;muito dif&#237;ceis&#8221; se encontrava adequada. Chama a aten&#231;&#227;o que a prova do TEC em 2019 apresentava predominantemente itens considerados &#8220;f&#225;ceis&#8221;.</p>
      <p>Sobre o par&#226;metro <italic>acerto ao acaso</italic> , constatou-se que 41,5% dos itens da prova do TEC apresentavam grande possibilidade de <italic>acerto ao acaso</italic> , um percentual elevado para uma avalia&#231;&#227;o certificativa do porte da prova do TEC. A CCI foi insatisfat&#243;ria para 58,5% dos itens e a <italic>curva de informa&#231;&#227;o</italic> foi satisfat&#243;ria em 78,8% dos itens, o que demonstra que o acerto aos itens n&#227;o apresentava boa correla&#231;&#227;o &#224; profici&#234;ncia do candidato, apesar de ser capaz de medir o tra&#231;o latente.</p>
      <p>Ao proceder a an&#225;lise individual dos itens da prova, constatou-se que 69 itens apresentavam problemas em rela&#231;&#227;o aos par&#226;metros da TRI avaliados e, portanto, com baixo poder de informa&#231;&#227;o em rela&#231;&#227;o &#224; identifica&#231;&#227;o do tra&#231;o latente dos candidatos. Apesar disso, a CCI era coerente em rela&#231;&#227;o &#224; aptid&#227;o do candidato e seu n&#250;mero de acertos aos itens, ou seja, quanto maior a aptid&#227;o do candidato, maior o n&#250;mero de itens corretos. Por&#233;m, a CCI demonstrou que candidatos com n&#237;vel de aptid&#227;o baixa eram capazes de acertar at&#233; 30,5% dos itens da prova. Em rela&#231;&#227;o &#224; identifica&#231;&#227;o de itens deficientes, o resultado foi semelhante ao observado na prova da Olimp&#237;ada Brasileira de Matem&#225;tica das Escolas P&#250;blicas em 2016, na qual 11 de suas 20 quest&#245;es eram deficientes em rela&#231;&#227;o aos par&#226;metros da Teoria Cl&#225;ssica de Teste. <sup><xref ref-type="bibr" rid="B3">3</xref></sup></p>
      <p>Ao retirar os itens com problemas nos par&#226;metros da TRI da prova original, restaram 49 itens, que ao serem avaliados em conjunto como um &#8220;modelo alternativo&#8221; de prova, mantiveram as mesmas caracter&#237;sticas psicom&#233;tricas da CCI da prova original e uma distribui&#231;&#227;o normal da m&#233;dia de profici&#234;ncia dos candidatos. No entanto, este modelo reduziu de 30,5% para 8,2% o percentual de candidatos que mesmo com baixa profici&#234;ncia acertariam os itens da prova. Essa redu&#231;&#227;o significativa deve-se &#224; diminui&#231;&#227;o do <italic>acerto ao acaso</italic> dos itens, resultado relevante observado nesse &#8220;modelo alternativo&#8221; de prova orientado pela TRI.</p>
      <p>Nesse sentido, observa-se a import&#226;ncia da an&#225;lise dos par&#226;metros psicom&#233;tricos em uma prova, os quais apresentam medidas matem&#225;ticas, contribuindo para a constru&#231;&#227;o de um exame composto por itens &#8220;calibrados&#8221;, e o aprimoramento do instrumento de avalia&#231;&#227;o.</p>
      <p>At&#233; onde se sabe, este &#233; o primeiro estudo a avaliar as caracter&#237;sticas psicom&#233;tricas de uma prova de t&#237;tulo de especialista da AMB e seus resultados contribuir&#227;o para reflex&#245;es e aprimoramentos desses instrumentos certificadores. Por esse motivo, n&#227;o foram encontradas outras refer&#234;ncias bibliogr&#225;ficas que permitissem comparar os resultados encontrados com os de outras sociedades de especialidades, mas h&#225; publica&#231;&#245;es dispon&#237;veis em outros cen&#225;rios.</p>
      <p>O presente estudo oportuniza a discuss&#227;o sobre o modelo atual de confec&#231;&#227;o da prova do TEC. Nele, os itens s&#227;o elaborados por um conjunto heterog&#234;neo de pessoas, que n&#227;o discutem a prova como um instrumento &#250;nico, e as provas n&#227;o t&#234;m as mesmas caracter&#237;sticas psicom&#233;tricas a cada ano, impossibilitando a comparabilidade no tempo.</p>
      <p>Esses dados contribuem para que a CJTEC possa, inclusive, avaliar o n&#250;mero de quest&#245;es que s&#227;o necess&#225;rias na prova do TEC, uma vez que, pela TRI, um modelo ajustado com 49 itens apresentou os mesmos resultados certificadores. A possibilidade da redu&#231;&#227;o do n&#250;mero de quest&#245;es de uma prova, orientada por m&#233;todos psicom&#233;tricos, pode produzir um instrumento de avalia&#231;&#227;o capaz de discriminar com maior precis&#227;o os candidatos merecedores do TEC, e de lhes oferecer um exame menos cansativo, favorecendo inclusive, um melhor desempenho dos candidatos. Assim, a probabilidade de concess&#227;o de t&#237;tulos favorecida pelos <italic>acertos ao acaso</italic> seria menor, aprimorando-se a identifica&#231;&#227;o dos participantes proficientes, com um padr&#227;o de respostas coerentes em rela&#231;&#227;o aos par&#226;metros estudados.</p>
      <p>Com base nos achados deste estudo e seguindo as tend&#234;ncias de outras institui&#231;&#245;es que j&#225; utilizam a TRI para a sele&#231;&#227;o dos itens de suas avalia&#231;&#245;es, <sup><xref ref-type="bibr" rid="B4">4</xref></sup> essa metodologia pode incrementar, de forma impactante, a qualidade das provas de t&#237;tulos das diversas especialidades da AMB, contribuindo para a identifica&#231;&#227;o de candidatos com as compet&#234;ncias esperadas para o exerc&#237;cio de sua especialidade no Brasil.</p>
      <p>Ao apoiar este estudo, a SBC demonstra o compromisso no aprimoramento do seu instrumento de certifica&#231;&#227;o profissional, a prova do TEC. Os resultados deste estudo in&#233;dito s&#227;o relevantes para o aprimoramento t&#233;cnico da elabora&#231;&#227;o dos itens para a prova de t&#237;tulo da SBC, e servir&#225; de referencial para outras Sociedades de Especialidades da AMB.</p>
      <sec>
        <title>Limita&#231;&#245;es e perspectivas</title>
        <p>O presente estudo apresenta algumas limita&#231;&#245;es. Para a obten&#231;&#227;o de melhores resultados pela TRI, &#233; desej&#225;vel a constru&#231;&#227;o de um banco de dados com itens previamente utilizados e calibrados. Isso n&#227;o foi poss&#237;vel, uma vez que este &#233; o primeiro estudo realizado com uma prova do TEC e, provavelmente, tamb&#233;m de um exame para concess&#227;o do t&#237;tulo de especialista da AMB. Outra limita&#231;&#227;o diz respeito ao banco de dados avaliado, pois embora tenhamos considerado a prova realizada no ano de 2019, todas as edi&#231;&#245;es da prova do TEC foram independentes, apesar de seguirem a mesma metodologia de elabora&#231;&#227;o. Por isso, n&#227;o &#233; poss&#237;vel afirmar que os resultados aqui apresentados possam ser extrapolados para os concursos anteriormente realizados. Apesar disso, o estudo apresenta importantes contribui&#231;&#245;es para que a SBC e AMB possam aprimorar os instrumentos certificativos para concess&#227;o de t&#237;tulos de especialistas.</p>
      </sec>
    </sec>
    <sec sec-type="conclusions">
      <title>Conclus&#227;o</title>
      <p>O presente estudo permitiu verificar as propriedades psicom&#233;tricas da prova do TEC de 2019 usando a TRI. O exame apresentou um maior percentual de quest&#245;es f&#225;ceis, com cerca de um ter&#231;o de itens com alto poder de discrimina&#231;&#227;o, e os demais necessitando melhorias na elabora&#231;&#227;o, uma vez que apresentaram elevada probabilidade de acerto ao acaso. O estudo sugere que uma avalia&#231;&#227;o com menor n&#250;mero de quest&#245;es seria capaz de apresentar as mesmas caracter&#237;sticas psicom&#233;tricas da prova analisada, mas com um potencial de reduzir o acerto ao acaso dos itens. Os resultados deste trabalho contribuem para o aprimoramento da prova do TEC, importante instrumento que certifica o especialista em cardiologia no Brasil.</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <title>Refer&#234;ncias</title>
      <ref id="B1">
        <label>1</label>
        <mixed-citation>1. Sociedade Brasileira de Cardiologia. Regimento da Comiss&#227;o de Julgamento do T&#237;tulo de Especialista em Cardiologia da Sociedade Brasileira de Cardiologia CJTEC. Rio de Janeiro: SBC; 2018.</mixed-citation>
        <element-citation publication-type="report">
          <person-group person-group-type="author">
            <collab>Sociedade Brasileira de Cardiologia</collab>
          </person-group>
          <source>Regimento da Comiss&#227;o de Julgamento do T&#237;tulo de Especialista em Cardiologia da Sociedade Brasileira de Cardiologia CJTEC</source>
          <publisher-loc>Rio de Janeiro</publisher-loc>
          <publisher-name>SBC</publisher-name>
          <year>2018</year>
        </element-citation>
      </ref>
      <ref id="B2">
        <label>2</label>
        <mixed-citation>2. Vilarinho APL. Uma Proposta de An&#225;lise de Desempenho dos Estudantes e de Valoriza&#231;&#227;o da Primeira Fase da OBMEP [dissertation]. Bras&#237;lia: Universidade de Bras&#237;lia; 2015.</mixed-citation>
        <element-citation publication-type="thesis">
          <person-group person-group-type="author">
            <name>
              <surname>Vilarinho</surname>
              <given-names>APL</given-names>
            </name>
          </person-group>
          <source>Uma Proposta de An&#225;lise de Desempenho dos Estudantes e de Valoriza&#231;&#227;o da Primeira Fase da OBMEP</source>
          <comment>dissertation</comment>
          <publisher-loc>Bras&#237;lia</publisher-loc>
          <publisher-name>Universidade de Bras&#237;lia</publisher-name>
          <year>2015</year>
        </element-citation>
      </ref>
      <ref id="B3">
        <label>3</label>
        <mixed-citation>3. Kn&#252;pfer REN, Amaral A, Henning E. An&#225;lise Cl&#225;ssica de Testes: Uma Proposta de An&#225;lise de Desempenho dos Estudantes na Primeira Fase da OBMEP. Joinville: Universidade Federal de Santa Catarina; 2016.</mixed-citation>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Kn&#252;pfer</surname>
              <given-names>REN</given-names>
            </name>
            <name>
              <surname>Amaral</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Henning</surname>
              <given-names>E</given-names>
            </name>
          </person-group>
          <source>An&#225;lise Cl&#225;ssica de Testes: Uma Proposta de An&#225;lise de Desempenho dos Estudantes na Primeira Fase da OBMEP</source>
          <publisher-loc>Joinville</publisher-loc>
          <publisher-name>Universidade Federal de Santa Catarina</publisher-name>
          <year>2016</year>
        </element-citation>
      </ref>
      <ref id="B4">
        <label>4</label>
        <mixed-citation>4. Oliveira ALS. Avalia&#231;&#227;o psicom&#233;trica da medida do componente de forma&#231;&#227;o geral da prova do exame nacional de desempenho de estudantes (ENADE) de 2010, 2011 e 2012 [dissertation]. Florian&#243;polis: Universidade Federal de Santa Catarina; 2017.</mixed-citation>
        <element-citation publication-type="thesis">
          <person-group person-group-type="author">
            <name>
              <surname>Oliveira</surname>
              <given-names>ALS</given-names>
            </name>
          </person-group>
          <source>Avalia&#231;&#227;o psicom&#233;trica da medida do componente de forma&#231;&#227;o geral da prova do exame nacional de desempenho de estudantes (ENADE) de 2010, 2011 e 2012</source>
          <comment>dissertation</comment>
          <publisher-loc>Florian&#243;polis</publisher-loc>
          <publisher-name>Universidade Federal de Santa Catarina</publisher-name>
          <year>2017</year>
        </element-citation>
      </ref>
    </ref-list>
    <fn-group>
      <fn fn-type="other">
        <p>Vincula&#231;&#227;o acad&#234;mica</p>
        <p>Este artigo &#233; parte de disserta&#231;&#227;o de mestrado de Gustavo Eug&#234;nio Martins Marinho pela Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS).</p>
      </fn>
      <fn fn-type="other">
        <p>Aprova&#231;&#227;o &#233;tica e consentimento informado</p>
        <p>Este artigo n&#227;o cont&#233;m estudos com humanos ou animais realizados por nenhum dos autores.</p>
      </fn>
    </fn-group>
    <fn-group>
      <fn fn-type="supported-by">
        <p>Fontes de financiamento: O presente estudo n&#227;o teve fontes de financiamento externas.</p>
      </fn>
    </fn-group>
  </back>
  <sub-article article-type="translation" id="TRen" xml:lang="en">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Original Article</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Psychometric Evaluation of the Cardiology Certification Exam of the Brazilian Society of Cardiology</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0001-6466-1287</contrib-id>
          <name>
            <surname>Marinho</surname>
            <given-names>Gustavo Eug&#234;nio Martins</given-names>
          </name>
          <xref ref-type="aff" rid="aff1001">
            <sup>1</sup>
          </xref>
          <role>Conception and design of the research</role>
          <role>Acquisition of data</role>
          <role>Analysis and interpretation of the data, Statistical analysis</role>
          <role>Obtaining financing and Writing of the manuscript</role>
          <role>Critical revision of the manuscript for important intellectual content</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-2684-0142</contrib-id>
          <name>
            <surname>Peixoto</surname>
            <given-names>Jos&#233; Maria</given-names>
          </name>
          <xref ref-type="aff" rid="aff2001">
            <sup>2</sup>
          </xref>
          <xref ref-type="aff" rid="aff3001">
            <sup>3</sup>
          </xref>
          <role>Critical revision of the manuscript for important intellectual content</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6537-7568</contrib-id>
          <name>
            <surname>Knopfholz</surname>
            <given-names>Jos&#233;</given-names>
          </name>
          <xref ref-type="aff" rid="aff4001">
            <sup>4</sup>
          </xref>
          <xref ref-type="aff" rid="aff5001">
            <sup>5</sup>
          </xref>
          <role>Critical revision of the manuscript for important intellectual content</role>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6599-5702</contrib-id>
          <name>
            <surname>Andrade</surname>
            <given-names>Marcus Vinicius Santos</given-names>
          </name>
          <xref ref-type="aff" rid="aff6001">
            <sup>6</sup>
          </xref>
          <xref ref-type="aff" rid="aff7001">
            <sup>7</sup>
          </xref>
          <xref ref-type="aff" rid="aff8001">
            <sup>8</sup>
          </xref>
          <role>Critical revision of the manuscript for important intellectual content</role>
        </contrib>
      </contrib-group>
      <aff id="aff1001">
        <label>1</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Faculdade de Medicina na Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS), Alfenas, MG &#8211; Brazil</institution>
      </aff>
      <aff id="aff2001">
        <label>2</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS), Belo Horizonte, MG &#8211; Brazil</institution>
      </aff>
      <aff id="aff3001">
        <label>3</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Programa de Mestrado Profissional em Ensino em Sa&#250;de da Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS-BH), Belo Horizonte, MG &#8211; Brazil</institution>
      </aff>
      <aff id="aff4001">
        <label>4</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Pontif&#237;cia Universidade Cat&#243;lica do Paran&#225;, Curitiba, PR &#8211; Brazil</institution>
      </aff>
      <aff id="aff5001">
        <label>5</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Programa de Especializa&#231;&#227;o em Educa&#231;&#227;o em Sa&#250;de da Universidade de S&#227;o Paulo, S&#227;o Paulo, SP &#8211; Brazil</institution>
      </aff>
      <aff id="aff6001">
        <label>6</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Escola Bahiana de Medicina, Salvador, BA &#8211; Brazil</institution>
      </aff>
      <aff id="aff7001">
        <label>7</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Hospital Santa Izabel - Santa Casa da Bahia, Salvador, BA &#8211; Brazil</institution>
      </aff>
      <aff id="aff8001">
        <label>8</label>
        <country country="BR">Brazil</country>
        <institution content-type="original">Hospital Alian&#231;a, Salvador, BA &#8211; Brazil</institution>
      </aff>
      <author-notes>
        <corresp id="c01001">Mailing Address: Gustavo Eug&#234;nio Martins Marinho &#8226; Rodovia MG-179 Km 0 s/n. Postal code 37132-440, Bairro Trevo Alfenas, Alfenas, MG &#8211; Brazil E-mail: gustavoeugenio@cardiol.br</corresp>
        <fn fn-type="conflict">
          <p>Potential Conflict of Interest</p>
          <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>The Cardiology Certification Exam is issued annually by the Brazilian Cardiology Society and set and applied by the Judging Committee for the Cardiologist Title (CJTEC). The psychometric analysis of the exam items using the Item Response Theory (IRT) may provide robust data that can help in the continuous improvement of this instrument.</p>
        </sec>
        <sec>
          <title>Objectives</title>
          <p>To evaluate the psychometric properties of the 2019 Cardiology Certification Exam in relation to the IR parameters.</p>
        </sec>
        <sec>
          <title>Methods</title>
          <p>This was an observational study, with psychometric analysis of the 120 questions of the exam taken by 1,120 candidates for the title of Cardiologist in 2019.</p>
        </sec>
        <sec>
          <title>Results</title>
          <p>The IRT analysis revealed that 32.2% of the items had a &#8220;high&#8221; or &#8220;very high&#8221; discriminating power, 49.2% were categorized as &#8220;easy&#8221; or &#8220;very easy&#8221;, and 41.5% showed a high probability of a correct guessing. Sixty-nine deficient items in terms of the IRT parameters were identified, which were then considered poorly effective in evaluating the candidate&#8217;s ability.</p>
        </sec>
        <sec>
          <title>Conclusions</title>
          <p>The psychometric analysis of the 2019 Cardiology Certification Exam by the IRT revealed a high percentage of easy questions, with nearly two thirds of the items with a high probability of correct guessing. These data may serve as a basis for a series of discussions and proposals for the elaboration of future certificate exams in Cardiology.</p>
        </sec>
      </abstract>
      <kwd-group xml:lang="en">
        <kwd>Specialization</kwd>
        <kwd>Cardiology</kwd>
        <kwd>Psychometrics</kwd>
      </kwd-group>
    </front-stub>
    <body>
      <sec sec-type="intro">
        <title>Introduction</title>
        <p>The title of specialist has become a constant goal among Brazilian physicians. The reasons range from knowledge gain, prerequisite to participate in public calls, to becoming a member of medical cooperatives in the labor market, evidencing that medical titles enhance both professional status and the prestige of the specialty.</p>
        <p>The Cardiology Certification Exam (CCE) has been issued by the Brazilian Cardiology Society (SBC) since 1968, but was legalized only in 1989 by the Brazilian Medical Association (AMB) and the Federal Council of Medicine (CFM) by the 1286/89 resolution. In this context, in 1992, the Judging Committee for the Cardiologist title (CJTEC) was created.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
        <p>The CCE consists of 120 multiple-choice questions with five choices with one correct answer each. There is a concern regarding the difficulty level of the questions, and in this respect, the CJTEC classify them as highly, moderately or little difficult. However, this classification has been done subjectively, <italic>i.e</italic> ., according to the opinion of the CJTEC members, without the use of a psychometric methodology that evaluates the degree of difficulty faced by the applicants.<sup><xref ref-type="bibr" rid="B2">2</xref></sup></p>
        <p>The item response theory (IRT) has been recently used as a psychometric method for the analysis and interpretation of the results in different scenarios of exams and public calls.<sup><xref ref-type="bibr" rid="B2">2</xref></sup></p>
        <p>So far, the CCE has not undergone a psychometric test, and considering the importance of this exam, it is essential to know whether this method of evaluation provides a reliable and coherent measure from the technical point of view. Based on this, this study aimed to assess the psychometric properties of the 2019 CCE in relation to the IRT.</p>
      </sec>
      <sec sec-type="methods">
        <title>Methods</title>
        <sec>
          <title>Study design</title>
          <p>This was an observational study, with psychometric analysis of 120 questions of the CCE taken by 1,120 applicants to obtain the title of cardiologist. The CCE was administered on October 27, 2019, from 13h to 18h at the Universidade Privada de S&#227;o Paulo.</p>
        </sec>
        <sec>
          <title>Inclusion and exclusion criteria</title>
          <p>All the answer keys delivered by the candidates who applied for the CCE in 2019 were included. After the appealing phase, two questions and one exam from an applicant who answered only two questions of the test were excluded.</p>
        </sec>
        <sec>
          <title>Sample</title>
          <p>After the exclusion of two questions in the appealing phase, the sample consisted of answer keys of 118 questions, answered by physicians who applied for the CCE in 2019.</p>
        </sec>
        <sec>
          <title>Data Collection</title>
          <p>Data were collected from the database of the agency responsible for elaborating the exam (Segmento Farma Editores Ltda., with the help of Simples Detalhe Assessoria, Planejamento e organiza&#231;&#227;o de Eventos Ltda. and Picsis inform&#225;tica ind&#250;stria e com&#233;rcio Ltda.) and plotted in Excel spreadsheets.</p>
          <p>Separate spreadsheets were then generated, with identification data and exam scores. The names of the candidates were deleted from the spreadsheets for the sake of confidentiality, and the applicants were identified by numbers.</p>
        </sec>
        <sec>
          <title>Ethical aspects</title>
          <p>Informed consent was waived since secondary databases were used, <italic>i.e.</italic> without participants&#8217; identification. However, to construct the database, a consent form for the use of the data was signed, which was first sent to the SBC and then to the ethics committee (approval number 4.030.702).</p>
        </sec>
        <sec>
          <title>Statistical analysis</title>
          <p>We performed a psychometric assessment of the 2019 CCE, offered by the SBC, using the IRT. The IRT aims to determine the applicant&#8217;s ability level (latent trait, theta [&#952;]), and the probability that a person with a given ability level will answer correctly a set of items according to their difficulty level.</p>
          <p>For analysis of the latent trait, the IRT assesses the following parameters:</p>
          <list list-type="alpha-lower">
            <list-item>
              <p>Item Discrimination (a): performance of the item in differentiating between individuals possessing different levels of ability;</p>
            </list-item>
            <list-item>
              <p>Item Difficulty (b): minimum ability that a respondent must possess to be very likely to answer correctly;</p>
            </list-item>
            <list-item>
              <p>Guessing (ci): probability of a low-proficient respondent answering correctly an item.</p>
            </list-item>
          </list>
          <p>Therefore, the IRT attempts to measure unobservable variables (latent trait) that may influence the answers given to the items, by measuring observed variables (responses). Thus, IRT establishes a relationship or the respondent&#8217;s ability and the item parameters with the probability of endorsing the correct answer for an item. The higher the person&#8217;s ability, the higher the respondent&#8217;s probability of answering correctly the instrument&#8217;s items.</p>
          <p>Two important assumptions of the IRT are Unidimensionality, that assumes that there is only one latent trait (&#952;) affecting the responses observed for the items in the measure, and Local Independence, that assumes that the individual&#8217;s performance in separate items is mutually independent, since each answer is given according to the dominant ability (&#952;) to that item.</p>
          <p>In Brazil, the most widely used IRT model is the unidimensional three-parameter logistic model. The unidimensional models with one or two parameters are not suitable for the analysis in the present study, since the results obtained from the three-parameter model revealed a great variation in the guessing item between the 120 questions of the exam applied in 2019.</p>
          <p>IRT calculation methods:</p>
          <p>Unidimensional three-parameter logistic model</p>
          <disp-formula id="e1001">
            <mml:math>
              <mml:mi>P</mml:mi>
              <mml:mfenced close=")" open="(" separators="|">
                <mml:mrow>
                  <mml:msub>
                    <mml:mi>U</mml:mi>
                    <mml:mrow>
                      <mml:mi>i</mml:mi>
                      <mml:mi>j</mml:mi>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>=</mml:mo>
                  <mml:mn>1</mml:mn>
                  <mml:mo>&#8739;</mml:mo>
                  <mml:msub>
                    <mml:mi>&#952;</mml:mi>
                    <mml:mi>j</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:mfenced>
              <mml:mo>=</mml:mo>
              <mml:msub>
                <mml:mi>c</mml:mi>
                <mml:mi>i</mml:mi>
              </mml:msub>
              <mml:mo>+</mml:mo>
              <mml:mfenced close=")" open="(" separators="|">
                <mml:mrow>
                  <mml:mn>1</mml:mn>
                  <mml:mo>&#8722;</mml:mo>
                  <mml:msub>
                    <mml:mi>c</mml:mi>
                    <mml:mi>i</mml:mi>
                  </mml:msub>
                </mml:mrow>
              </mml:mfenced>
              <mml:mfrac>
                <mml:mn>1</mml:mn>
                <mml:mrow>
                  <mml:mn>1</mml:mn>
                  <mml:mo>+</mml:mo>
                  <mml:msup>
                    <mml:mi>e</mml:mi>
                    <mml:mrow>
                      <mml:mo>&#8722;</mml:mo>
                      <mml:mi>D</mml:mi>
                      <mml:msub>
                        <mml:mi>a</mml:mi>
                        <mml:mi>i</mml:mi>
                      </mml:msub>
                      <mml:mfenced close=")" open="(" separators="|">
                        <mml:mrow>
                          <mml:msub>
                            <mml:mi>&#952;</mml:mi>
                            <mml:mi>j</mml:mi>
                          </mml:msub>
                          <mml:mo>&#8722;</mml:mo>
                          <mml:msub>
                            <mml:mi>b</mml:mi>
                            <mml:mi>i</mml:mi>
                          </mml:msub>
                        </mml:mrow>
                      </mml:mfenced>
                    </mml:mrow>
                  </mml:msup>
                </mml:mrow>
              </mml:mfrac>
            </mml:math>
          </disp-formula>
          <p>with i = 1, 2, ..., I and j = 1, 2, ..., n, where:</p>
          <list list-type="simple">
            <list-item>
              <p>- <bold>U</bold> <sub><bold>ij</bold></sub> is a dichotomous variable that corresponds to 1, when the respondent j answers correctly the item i, or 0 when the respondent does not answer the item i correctly.</p>
            </list-item>
            <list-item>
              <p>- <bold>&#952;</bold> <sub><bold>j</bold></sub> represents the ability (latent trait) of the respondent number j.</p>
            </list-item>
            <list-item>
              <p>- <bold>P(U</bold> <sub><bold>ij</bold></sub> <bold>=1</bold> | &#952;<sub><bold>j</bold></sub> <bold>)</bold> is the probability of the individual j with a &#952;j ability to answer correctly the item i, and is called Item Response Function (IRF).</p>
            </list-item>
            <list-item>
              <p>- <bold>b</bold> <sub><bold>i</bold></sub> is the difficulty (or position) parameter, measured on the same scale as ability.</p>
            </list-item>
            <list-item>
              <p>- <bold>a</bold> <sub><bold>i</bold></sub> is the discrimination (or inclination) parameter of the item i, which is proportional to the inclination of the item characteristic curve (ICC) in the point bi</p>
            </list-item>
            <list-item>
              <p>- <bold>c</bold> <sub><bold>i</bold></sub> is the parameter that represents the probability of low-ability individuals answering correctly the item i by chance (often referred as the correct guessing probability)</p>
            </list-item>
            <list-item>
              <p>- <bold>D</bold> is a scale factor, constant (=1).</p>
            </list-item>
          </list>
          <p>Values of the <italic>a, b</italic> and <italic>c</italic> parameters are calculated by pre-testing (calibration) using the maximum likelihood ( <bold>L</bold> ) method, which works with derivatives and is defined as:</p>
          <disp-formula id="e2001">
            <mml:math>
              <mml:mrow>
                <mml:mi mathvariant="normal">L</mml:mi>
              </mml:mrow>
              <mml:mfenced close=")" open="(" separators="|">
                <mml:mrow>
                  <mml:msub>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">u</mml:mi>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:mn>1</mml:mn>
                      <mml:msup>
                        <mml:mrow>
                          <mml:mtext/>
                          <mml:mi mathvariant="normal">s</mml:mi>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:mi>&#8242;</mml:mi>
                        </mml:mrow>
                      </mml:msup>
                    </mml:mrow>
                  </mml:msub>
                  <mml:msub>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">u</mml:mi>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:mn>2</mml:mn>
                      <mml:msup>
                        <mml:mrow>
                          <mml:mtext/>
                          <mml:mi mathvariant="normal">s</mml:mi>
                        </mml:mrow>
                        <mml:mrow>
                          <mml:mi>&#8242;</mml:mi>
                        </mml:mrow>
                      </mml:msup>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>&#8230;</mml:mo>
                  <mml:mo>,</mml:mo>
                  <mml:msub>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">u</mml:mi>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mi mathvariant="normal">n</mml:mi>
                        <mml:mi mathvariant="normal">s</mml:mi>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msub>
                  <mml:mo>&#8739;</mml:mo>
                  <mml:mi>&#952;</mml:mi>
                </mml:mrow>
              </mml:mfenced>
              <mml:mo>=</mml:mo>
              <mml:munderover>
                <mml:mrow>
                  <mml:mo>&#8719;</mml:mo>
                </mml:mrow>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">i</mml:mi>
                  </mml:mrow>
                  <mml:mo>=</mml:mo>
                  <mml:mn>1</mml:mn>
                </mml:mrow>
                <mml:mrow>
                  <mml:mrow>
                    <mml:mi mathvariant="normal">n</mml:mi>
                  </mml:mrow>
                </mml:mrow>
              </mml:munderover>
              <mml:mrow>
                <mml:mo> </mml:mo>
              </mml:mrow>
              <mml:mi>Pi</mml:mi>
              <mml:mo>&#8289;</mml:mo>
              <mml:mo stretchy="false">(</mml:mo>
              <mml:mi>&#952;</mml:mi>
              <mml:mrow>
                <mml:mi mathvariant="normal">s</mml:mi>
              </mml:mrow>
              <mml:msup>
                <mml:mo stretchy="false">)</mml:mo>
                <mml:mrow>
                  <mml:msub>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">u</mml:mi>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mi mathvariant="normal">s</mml:mi>
                        <mml:mi mathvariant="normal">i</mml:mi>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msub>
                </mml:mrow>
              </mml:msup>
              <mml:mrow>
                <mml:mi mathvariant="normal">Q</mml:mi>
                <mml:mi mathvariant="normal">i</mml:mi>
              </mml:mrow>
              <mml:mo stretchy="false">(</mml:mo>
              <mml:mi>&#952;</mml:mi>
              <mml:mrow>
                <mml:mi mathvariant="normal">s</mml:mi>
              </mml:mrow>
              <mml:msup>
                <mml:mo stretchy="false">)</mml:mo>
                <mml:mrow>
                  <mml:mn>1</mml:mn>
                  <mml:mo>&#8722;</mml:mo>
                  <mml:msub>
                    <mml:mrow>
                      <mml:mi mathvariant="normal">u</mml:mi>
                    </mml:mrow>
                    <mml:mrow>
                      <mml:mrow>
                        <mml:mi mathvariant="normal">s</mml:mi>
                        <mml:mi mathvariant="normal">i</mml:mi>
                      </mml:mrow>
                    </mml:mrow>
                  </mml:msub>
                </mml:mrow>
              </mml:msup>
            </mml:math>
          </disp-formula>
          <p>The maximum likelihood ( <bold>L</bold> ) works with derivatives.</p>
          <p>Where:</p>
          <list list-type="simple">
            <list-item>
              <p>- <bold>i</bold> = 1, 2, ..., n items</p>
            </list-item>
            <list-item>
              <p>- <bold>u</bold> <sub><bold>is</bold></sub> = response of the individual to each item (1 = correct, 0 = wrong)</p>
            </list-item>
          </list>
          <p>To calculate the ability/ proficiency of the applicant, we have first to determine the maximum value of the function above. First, the probability of correct responses [(Pi(&#952;)] of each item is determined using one of the three IRT models &#8211; 1PL, 2PL or 3PL. In the present study, the three-parameter model (3PL) was used. Then, &#952; is empirically substituted with values ranging from -5 to +5 (-5,00 &#8804; &#952; &#8804; +5,00, usually -3,00 &#8804; &#952; &#8804; +3,00), or the Newton-Raphson iteration algorithm is used to calculate the maximum of the L function. Based on the &#952;, this maximum represents the applicant&#8217;s ability/proficiency.</p>
        </sec>
        <sec>
          <title>Item characteristic curve (ICC)</title>
          <p>The mathematical model that defines IRT is a probability function. Therefore, it will always be visualized within the interval [0,1]. The number U<sub>ij</sub>=1|&#952;<sub>j</sub>) can be identified by the proportion of correct answers to the item I in the group of individuals with ability &#952;j. This ability is described as a sigmoid curve, where the horizontal axis represents the ability level and the vertical axis the probability of the individual with ability &#952;j to give a correct response to the item i. Two horizontal asymptotes can be highlighted, and three parameters can be seen with some accuracy.</p>
        </sec>
        <sec>
          <title>Item information curve &#8211; I(&#952;)</title>
          <p>Informatics accuracy is the degree of accuracy in which the item represents what it intends to measure. In this context, accuracy means how well the item predicts the criterion or represents the latent trait (&#952;). Thus, the IRT information function follows the calculation of the estimation error, that is, how much the score obtained by an individual in a test differs from the real score. The concept of information function itself is the reciprocal of variance, <italic>i.e</italic> ., I = 1/S<sup>2</sup>. The information function corresponds to the concept of factorial load of the item of the factorial analysis, from the latent trait model perspective, since the factorial load represents the covariance between the item (behavioral representation) and the latent trait (theta). The test information curve depicts the amount of information yielded by the test at any ability level; it presents the amplitude of theta to which the test provides reliable information, and out of which the test provides more erroneous than correct information about theta. Thus, the information curve has an interface to both test parameters, i.e., validity and accuracy, but is not cofounded by any of them. Representation of the information item resemble a normal-type (bell-shape) curve.</p>
          <p>In the present analysis, a rate of correct guessing &#8805; 25% in an item of the exam was considered unsatisfactory. Then, of the 1,120 exams, 5% of correct guessing higher than the expected rate (20%) is considered very high, and thus the item evaluated has some problem in its formulation or in the answer choices. The correct guessing can be seen by the lack of coherence of the candidate in answering incorrectly easy questions or, in contrast, answering correctly difficult questions, with no ability for it.</p>
        </sec>
      </sec>
      <sec sec-type="results">
        <title>Results</title>
        <p>We present the results obtained from the psychometric analysis of 118 items of the exam the candidates applying for the CCE in 2019, using a three-parameter unidimensional logistic model of IRT: discrimination (a), difficulty (b) and guessing (c).</p>
        <p>In the analysis, one item (question number 110) revealed a negative level for the <italic>discrimination</italic> parameter ( <italic>a</italic> = - 0.174), suggesting that the higher the respondent&#8217;s knowledge level, the lower the probability of correct answer, which is inconsistent with the objective of the parameter. For this reason, this item was not included in the final analysis.</p>
        <p><xref ref-type="table" rid="t1001">Table 1</xref> presents the distribution of the 118 items of the exam by their discriminating power. Of these items, 18.7% showed a very low or low discriminating power (a &#8804; 0.65); 49.1% showed moderate discriminating power (0.651 &lt; a &#8804; 1.350) and 32.2% showed high or very high discriminating power (a &#8805; 1.351).</p>
        <p>
          <table-wrap id="t1001">
            <label>Table 1</label>
            <caption>
              <title>Distribution of the exam items by the item response theory (IRT) <italic>discrimination</italic> parameter</title>
            </caption>
            <table frame="hsides" rules="groups">
              <colgroup>
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead>
                <tr>
                  <th align="left">Classification of the discriminating power (a)</th>
                  <th>Frequency (n)</th>
                  <th>%</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td>&#8804; 0,35 (very low)</td>
                  <td align="center">12</td>
                  <td align="center">10.2</td>
                </tr>
                <tr>
                  <td>0.351 - 0.650 (low)</td>
                  <td align="center">10</td>
                  <td align="center">8.5</td>
                </tr>
                <tr>
                  <td>0.651 - 1.350 (moderate)</td>
                  <td align="center">58</td>
                  <td align="center">49.1</td>
                </tr>
                <tr>
                  <td>1.351 - 1.700 (high)</td>
                  <td align="center">25</td>
                  <td align="center">21.2</td>
                </tr>
                <tr>
                  <td>&gt; 1.700 (very high)</td>
                  <td align="center">13</td>
                  <td align="center">11.0</td>
                </tr>
                <tr>
                  <td>Total</td>
                  <td align="center">118</td>
                  <td align="center">100.0</td>
                </tr>
              </tbody>
            </table>
            <table-wrap-foot>
              <fn id="TFN1001">
                <p>Database: 1,120 candidates. Note: Two items cancelled (items 23 and 46)</p>
              </fn>
            </table-wrap-foot>
          </table-wrap>
        </p>
        <p><xref ref-type="table" rid="t3001">Table 2</xref> presents the distribution of the 118 items of the exam according to the <italic>difficulty</italic> parameter. Of these items, 49.2% were classified as easy or very easy (b &lt; -0,52); 22.0% were moderately difficult (-0.51 &#8804; b &#8804; 0.51); and 28.8% were classified as difficult or very difficult (b &#8805; 0.52).</p>
        <p>
          <table-wrap id="t3001">
            <label>Table 2</label>
            <caption>
              <title>Distribution of the exam items by the item response theory (IRT) <italic>difficulty</italic> parameter</title>
            </caption>
            <table frame="hsides" rules="groups">
              <colgroup>
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead>
                <tr>
                  <th align="left">Classification of the difficulty parameter (b)</th>
                  <th>Frequency (n)</th>
                  <th>%</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td>&#8804; -1.28 (very easy)</td>
                  <td align="center">31</td>
                  <td align="center">26.3</td>
                </tr>
                <tr>
                  <td>-1.27 &#8211; -0.52 (easy)</td>
                  <td align="center">27</td>
                  <td align="center">22.9</td>
                </tr>
                <tr>
                  <td>-0.51 - 0.51 (moderate)</td>
                  <td align="center">26</td>
                  <td align="center">22.0</td>
                </tr>
                <tr>
                  <td>0.52 &#8211; 1.27 (difficult)</td>
                  <td align="center">19</td>
                  <td align="center">16.1</td>
                </tr>
                <tr>
                  <td>&#8805; 1.28 (very difficult)</td>
                  <td align="center">15</td>
                  <td align="center">12.7</td>
                </tr>
                <tr>
                  <td>Total</td>
                  <td align="center">118</td>
                  <td align="center">100.0</td>
                </tr>
              </tbody>
            </table>
            <table-wrap-foot>
              <attrib>Source: The authors; database: 1,120 candidates. Note: Two items cancelled (items 23 and 46).</attrib>
            </table-wrap-foot>
          </table-wrap>
        </p>
        <p><xref ref-type="table" rid="t2001">Table 3</xref> presents the distribution of the 118 items of the exam according to the <italic>guessing</italic> parameter. Of these, 41.5% of the items showed a high probability of guessing correctly according to the IRT methodology.</p>
        <p>
          <table-wrap id="t2001">
            <label>Table 3</label>
            <caption>
              <title>Distribution of the exam items by the percentage of <italic>correct guessing</italic> according to the item response theory (IRT)</title>
            </caption>
            <table frame="hsides" rules="groups">
              <colgroup>
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead>
                <tr>
                  <th align="left">Percentage of correct guessing (c)</th>
                  <th>Frequency (n)</th>
                  <th>%</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td>&#8804; 10.0%</td>
                  <td align="center">48</td>
                  <td align="center">40.7</td>
                </tr>
                <tr>
                  <td>10.1 - 25.0%</td>
                  <td align="center">21</td>
                  <td align="center">17.8</td>
                </tr>
                <tr>
                  <td>25.1 - 40.0%</td>
                  <td align="center">20</td>
                  <td align="center">16.9</td>
                </tr>
                <tr>
                  <td>40.1 - 60.0%</td>
                  <td align="center">19</td>
                  <td align="center">16.1</td>
                </tr>
                <tr>
                  <td>&gt; 60.0%</td>
                  <td align="center">10</td>
                  <td align="center">8.5</td>
                </tr>
                <tr>
                  <td>Total</td>
                  <td align="center">118</td>
                  <td align="center">100.0</td>
                </tr>
              </tbody>
            </table>
            <table-wrap-foot>
              <attrib>Source: The authors; database: 1,120 candidates. Note: Two items cancelled (items 23 and 46).</attrib>
            </table-wrap-foot>
          </table-wrap>
        </p>
        <p>According to the ICC and the information curve, 58.5% and 78.8% of the items, respectively, were considered unsatisfactory ( <xref ref-type="table" rid="t4001">Table 4</xref> ).</p>
        <p>
          <table-wrap id="t4001">
            <label>Table 4</label>
            <caption>
              <title>Distribution of the exam items according to the item characteristic curve and the information curve of the item response theory</title>
            </caption>
            <table frame="hsides" rules="groups">
              <colgroup>
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead>
                <tr>
                  <th align="left">Item characteristic curve</th>
                  <th>Frequency (n)</th>
                  <th>%</th>
                </tr>
              </thead>
              <tbody>
                <tr>
                  <td>Satisfactory</td>
                  <td align="center">49</td>
                  <td align="center">41.5</td>
                </tr>
                <tr>
                  <td>Unsatisfactory</td>
                  <td align="center">69</td>
                  <td align="center">58.5</td>
                </tr>
                <tr>
                  <td>
                    <bold>Information curve</bold>
                  </td>
                  <td align="center">
                    <bold>Frequency (n)</bold>
                  </td>
                  <td align="center">
                    <bold>%</bold>
                  </td>
                </tr>
                <tr>
                  <td>Satisfactory</td>
                  <td align="center">93</td>
                  <td align="center">78.8</td>
                </tr>
                <tr>
                  <td>Unsatisfactory</td>
                  <td align="center">25</td>
                  <td align="center">21.2</td>
                </tr>
              </tbody>
            </table>
            <table-wrap-foot>
              <attrib>Source: The authors; database: 1,120 candidates. Note: Two items cancelled (items 23 and 46).</attrib>
            </table-wrap-foot>
          </table-wrap>
        </p>
        <p>Individual analysis of the exam items by the IRT identified 69 deficient items in relation to the three parameters, that were then considered to have a low probability of providing information about the latent trait (&#952;), which evaluates the ability of the candidate. Thus, the other 49 items were analyzed by the IRT and compared with the initial model composed of 118 items.</p>
        <p><xref ref-type="fig" rid="f01001">Figure 1</xref> shows the ICC considering the 118 items by the IRT method. The results showed that the higher the applicant&#8217;s ability (&#952;), the higher the number of correct answers. It is expected that a medium-ability respondent answers approximately 80 (out of 118, 67.8%) items correctly. In addition, a very low-ability candidate (&#952; &lt; -4.0) is expected to answer at least 36 (out of 118, 30.5%) items correctly.</p>
        <p>
          <fig id="f01001">
            <label>Figure 1</label>
            <caption>
              <title>Score: T(&#952;) &#8211; of each respondent, estimated by the item response theory (IRT) considering a total of 118 exam items, according to the candidate&#8217;s ability (&#952;).</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf01-en.tif"/></fig>
        </p>
        <p>The information curve ( <xref ref-type="fig" rid="f02001">Figure 2</xref> ) for the 118 items showed that the maximum amount of information about the logical reasoning of the candidate was near the median ability, <italic>i.e</italic> ., &#952; near zero. Besides, for the extreme values of &#952;, the exam produces more information error than legitimate information, and the maximum information generated by the exam is within &#952; values between -3.2 and +3.1.</p>
        <p>
          <fig id="f02001">
            <label>Figure 2</label>
            <caption>
              <title>Information curve: I(&#952;) &#8211; and standard error of each candidate, generated by the item response theory, according to the respondent&#8217;s ability (&#952;).</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf02-en.tif"/></fig>
        </p>
        <p><xref ref-type="fig" rid="f03001">Figure 3</xref> shows the ICC for the 49 items remaining after the items with problems related to the IRT were excluded. The result shows that the higher the ability (&#952;) the higher the number of correct responses. Thus, it is expected that a 0-ability candidate (&#952; = 0 &#8211; median ability, -1 &lt; &#952; &lt; +1) answers approximately 32 questions (out of 49, 65.3%) correctly, and a very low-ability candidate (&#952; &lt; -4.0) answers at least four (out of 49, 8.2%) correctly. Therefore, considering the IRT data for the 49 items, the candidates will require a higher ability level (&#952;) than that required for the 118 exam items.</p>
        <p>
          <fig id="f03001">
            <label>Figure 3</label>
            <caption>
              <title>Score: T(&#952;) &#8211; of each respondent, estimated by the item response theory (IRT) considering a total of 118 exam items, according to the candidate&#8217;s ability (&#952;).</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf03-en.tif"/></fig>
        </p>
        <p>The information curve ( <xref ref-type="fig" rid="f04001">Figure 4</xref> ) for the 49 items showed that the maximum amount of information about the logical reasoning of the candidate was also near the median ability, i.e., &#952; near zero. Besides, for the extreme values of &#952;, the exam produces more information error than legitimate information, and the maximum information generated by the exam is within &#952; values between -4.0 and +3.2.</p>
        <p>
          <fig id="f04001">
            <label>Figure 4</label>
            <caption>
              <title>Information curve: I(&#952;) &#8211; and standard error generated by the item response theory, considering the 49 exam items.</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf04-en.tif"/></fig>
        </p>
        <p><xref ref-type="fig" rid="f05001">Figure 5</xref> depicts the results of ability generated by the IRT, considering the 49 items excluded from the exam initially applied. As can be seen, the mean ability level of the candidates shows a normal distribution, illustrated by a Gaussian pattern of data distribution.</p>
        <p>
          <fig id="f05001">
            <label>Figure 5</label>
            <caption>
              <title>Results of ability generated by the item response theory.</title>
            </caption>
            <attrib>Source: The authors.</attrib>
            <graphic xlink:href="1678-4170-abc-119-5suppl1-0006-20355-gf05-en.tif"/></fig>
        </p>
      </sec>
      <sec sec-type="discussion">
        <title>Discussion</title>
        <p>The aim of the present study was to analyze the items of the 2019 CCE regarding the psychometric parameters using the IRT. So far, the only known parameter was the degree of difficulty of the questions, categorized as easy, moderately difficult, or difficult, based on the knowledge and experience of the CJTEC members, who participated in the test formulation. However, this method of evaluation is subjective and lacks validity.</p>
        <p>Regarding the <italic>discrimination</italic> parameter, only 32.2% of the items showed a &#8220;high&#8221; or &#8220;very high&#8221; discriminating power. This is a relevant information, since the discrimination of an item is related to its capacity to identify candidates with different ability levels, as the parameter measures the probability of individuals with different ability levels to answer an item correctly. Similar data were observed in the Brazilian National Exam for the Assessment of Student Performance (ENADE, <italic>Exame Nacional de Desempenho dos Estudantes</italic> ) applied in 2010, 2011 and 2012. Psychometric analysis of these exams identified several questions with low discriminating power, providing technical contributions for the formulation of new items for the following exams.<sup><xref ref-type="bibr" rid="B3">3</xref> , <xref ref-type="bibr" rid="B4">4</xref></sup></p>
        <p>With respect to the <italic>difficulty</italic> parameter, 49.2% of the items were categorized by the IRT as &#8220;easy&#8221; or &#8220;very easy&#8221;, and only 22% as &#8220;moderately difficult&#8221;. This indicates that the CCE was unbalanced in terms of psychometry, which recommends the following proportion of the items by difficulty level &#8211; very easy (10%), easy (20%), moderately difficulty (40%), difficult (20%) and very difficult (10%).<sup><xref ref-type="bibr" rid="B4">4</xref></sup> The proportion of &#8220;difficult&#8221; and &#8220;very difficult&#8221; items was adequate. It is of note that the 2019 CCE was predominantly composed of &#8220;easy&#8221; items.</p>
        <p>As for the <italic>guessing</italic> parameter, 41.5% of the CCE items had high probability of correct guessing. This is a high percentage considering the importance of the CCE. The ICC was unsatisfactory for 58.5% of the items and the information curve was satisfactory for 78.8% of the items, which indicates that answering correctly the items did not have a good correlation with the respondents&#8217; ability, although it was able to measure the latent trait.</p>
        <p>Individual analysis of the exam items identified 69 items with problems related to the IRT parameters and that were then considered to have a low probability of providing information about the candidates&#8217; latent trait. Despite that, ICC was consistent regarding the candidate&#8217;s ability and the number of correct answers, <italic>i.e</italic> ., the higher the candidate&#8217;s ability, the higher the number of correct answers. Nevertheless, the ICC also revealed that low-ability respondents were able to answer up to 30.5% of the questions correctly. Similar result had been found in the 2016 Brazilian Mathematical Olympiad of Public Schools, in which 11 out of its 20 questions were deficient considering the classical test theory criteria.<sup><xref ref-type="bibr" rid="B3">3</xref></sup></p>
        <p>When the deficient items were removed from the original exam, the remaining 49 items were assessed as an &#8220;alternative model&#8221; of exam and maintained the same psychometric characteristics of the ICC of the original test and a normal distribution with the mean ability level of the candidates. However, with this model, the percentage of low-ability candidates who would answer the items correctly reduced from 30.5% to 8.2%. This significant reduction is attributed to a decrease in the percentage of correct guessing, which is a relevant result of the &#8220;alternative model&#8221; of exam, obtained by the IRT.</p>
        <p>Therefore, psychometric parameters have mathematical measures, and their analysis in certification exams allows the improvement and construction of more &#8220;calibrated&#8221; instruments.</p>
        <p>To the best of our knowledge, this is the first study to evaluate the psychometric characteristics of a specialist certification exam of the AMB, and the results will contribute to ideas and enhancement of this instrument. For this reason, we did not identify references of other medical societies or specialties to compare our results, although there are publications in other scenarios.</p>
        <p>The present study opens the discussion about the current model of elaboration of the CCE. In this model, the items are constructed by a heterogeneous group of people, who do not discuss the exam as a unique instrument. Also, the annual exams do not have similar psychometric characteristics, which precludes their comparability over time.</p>
        <p>In addition, our data contribute for the CJTEC to analyze the adequate number of questions of the CCE, since the IRT showed that an adjusted model of 49 items yielded the same certifying results. The possible reduction of the number of questions, when guided by psychometric methods, can produce an instrument able to discriminate, with greater accuracy, the candidates who are qualified for the title of cardiologist. Also, the exam would be less exhaustive, favoring a better performance of the candidates. Thus, the likelihood of passing the CCE due to a high percentage of correct answers by chance would be reduced, optimizing the identification of proficient professionals, able to give coherent answers in terms of the parameters evaluated.</p>
        <p>Based on our findings and on the trends observed in other institutions where the IRT has been used for the selection of their exams&#8217; items,<sup><xref ref-type="bibr" rid="B4">4</xref></sup> this method can strongly impact the quality of the AMB specialty certification exams, contributing to the identification of candidates with the competencies expected for their practice.</p>
        <p>The SBC supported this study, demonstrating its commitment in improving its professional certifying instrument, the CCE. The results of this unprecedented study are important for the technical improvement of the CCE items and will serve as a reference to other AMB specialty societies.</p>
        <sec>
          <title>Limitations and perspectives</title>
          <p>The present study has some limitations. First, better results of the IRT can be obtained if a database with previously calibrated items is used. However, this was not possible in our study, since this is the first one to evaluate the CCE, and probably the first to evaluate an AMB medical specialty certificate examination. Another limitation is related to the database used in the study. Although we have analyzed the CCE applied in 2019, all previous editions were independent despite having been elaborated using the same method. Thus, we cannot affirm that the results obtained from the present study can be extrapolated to previous years&#8217; editions. However, we do believe that the study provides important contributions for the SBC and the AMB to make improvements in their exams.</p>
        </sec>
      </sec>
      <sec sec-type="conclusions">
        <title>Conclusion</title>
        <p>This study allowed to determine the psychometric characteristics of the 2019 CCE by the IRT. The exam showed a high percentage of easy questions, with nearly one third of the questions with a high discriminating power and two thirds requiring improvements, as they had a high probability of correct guessing. The study suggests that an exam with a lower number of questions would show the same psychometric characteristics of the initial instrument, but with the potential to reduce the probability of guessing the answers correctly. These results contribute to the improvement of the CCE, an important certificate examination for the title of cardiologist in Brazil.</p>
      </sec>
    </body>
    <back>
      <fn-group>
        <fn fn-type="other">
          <p>Study Association</p>
          <p>This article is part of the thesis of master submitted by Gustavo Eug&#234;nio Martins Marinho, from Universidade Jos&#233; do Ros&#225;rio Vellano (UNIFENAS).</p>
        </fn>
        <fn fn-type="other">
          <p>thics approval and consent to participate</p>
          <p>This article does not contain any studies with human participants or animals performed by any of the authors.</p>
        </fn>
      </fn-group>
      <fn-group>
        <fn fn-type="supported-by">
          <p>Sources of Funding: There were no external funding sources for this study.</p>
        </fn>
      </fn-group>
    </back>
  </sub-article>
</article>
