To determine the clinical characteristics of children with acute heart failure syndrome in the emergency room and identify factors associated with poor outcomes.

Rivers State University Teaching Hospital (formerly known as Braithwaite Memorial Specialist Hospital) is a tertiary hospital located in Port Harcourt, Rivers State, Nigeria. The pediatric department contains 80 beds, including 12 in the emergency ward. The pediatric emergency department treats an average of 1724 children annually. Patients admitted through the emergency department are first evaluated by the resident physician, whose activity is overseen by a supervising consultant. Admitted patients are given urgent treatment, followed by transfer to the main pediatric ward within 48hrs of admission if they are determined to be clinically stable. Patients who respond to treatment can also be discharged from the emergency room within 24 hours.

This is a retrospective review of prospectively collected data from acute heart failure records of the pediatric emergency ward between March 2014 and July 2015. Children who presented in the emergency ward with the cardinal signs of heart failure were identified by the attending physician. The criteria for diagnosing heart failure were the presence of hepatomegaly with any two of the following:⁸ 1) significant tachycardia, which is resting heart rate greater than the following cut-offs for different ages: 160bpm in infants, 150bpm in children >1-2 years of age, 140bpm in children >2-4 years, 120bpm in children >4-6 years and 100bpm in children >6 years (febrile children were allowed an additional 10bpm for every degree rise in temperature above normal); 2) tachypnea, which is resting respiratory rate above the following cut-offs; 60 bpm in infants <2 months old, 50 bpm in infants 2-12 months old, 40 bpm in children >1 year old; 3) cardiomegaly, which is a displaced apex beat in the presence of a central trachea or a cardiothoracic ratio of >60% in children ≤5 years of age and >50% in older children. Children diagnosed with heart failure whose parents gave consent were included in the study. Apart from excluding children whose parents who did not give consent, we also excluded neonates (children aged ≤ 28 days old), since they are usually admitted into the special-care baby unit, which is separate from the pediatric emergency ward.

After initial resuscitative treatment had been carried out to stabilize the patients, a member of the research team approached the parents and patient to inform them about the study and to receive formal consent for inclusion. Once recruited, the patients were followed up throughout their stay in the hospital, from the emergency room to the main hospital ward stay until discharge. An interviewer administered a questionnaire to the parents, and a study number was allocated to the patient. Data was collected using a structured proforma and was updated daily. The following information was collected: age, gender, referral pattern, social class, weight, temperature, respiratory rate, heart rate, presence of hepatomegaly, previous treatment, modified Ross score, packed cell volume, white blood cell count, primary diagnosis, treatments given, diagnosis, length of hospital stay, and outcome of admission.

Social class was categorized according to the parent’s level of education and occupation, from class 1 (highest) to class 5 (lowest).¹³ A hemoglobin level of less than 8g/dl was described as severe anemia.¹⁴ The modified Ross criteria were used to classify the severity of heart failure.

Nutritional status was classified using weight-for-age Z-scores^15-17

The data were entered into an Excel spreadsheet and analyzed using IBM SPSS statistics version 23. Univariate descriptive statistics were used to explain the patient’s characteristics with measures of central tendency and dispersion. Kaplan-Meier survival curves were generated for resolution of heart failure by 1 week (168 hrs) after admission, and a log rank test was performed to compare the time to resolution of heart failure in different sub groups. P-values ≤0.05 were considered significant. Post hoc identification of risk factors for poor outcomes was performed using univariate logistic regression for all variables. Variables with a p-value <0.1 or with a clinical application relevant to children with a non-cardiac cause of heart failure were put into a multivariate logistic regression model. Backward stepwise elimination was used to determine variables included in the final model. The final model was internally validated using the Hosmer and Lemeshow goodness-of-fit test, with an acceptable fitness set at a p >0.05, and a ROC curve of the new model was constructed to assess its discriminatory accuracy.

The primary outcome measures were the persistence of heart failure >4 days after admission or death. The results are reported as odds ratio with 95% confidence interval.

Informed consent was required from all participating patients, although refusal did not affect the patient’s hospital management. Ethical clearance was obtained from the ethics committee of the Rivers State Hospital Management Board, Port Harcourt, Nigeria.

The included children ranged from one month to 16 years of age, with a median age of 11 months (IQR: 5- 38 months). There were 45(51%) males, with a male-to-female ratio of 1.1:1. Most patients 72(81%) lived in urban areas, belonged to the middle class 52(58%), and lived with both of their parents 80(90%). Most presented directly to the hospital 53(60%) without being referred, although 67(75%) had taken medication before presenting to the hospital. Of those who had taken medication prior to admission, 28(42%) took prescription medication while 10(15%) took herbal medications (Table 1).

Table 1

Variable		All patients (n=89) (%)
Age months (Median, 25th,75th percentile)		11 ( 5, 38)
1month to ≥ 5 yrs	72 (81)
>5 to 10 yrs	11(12)
>10 yrs	6 (7)
Gender
Female	44 (49)
Male	45 (51)
Residence
Urban	72 (81)
Rural	17 (19)
Who child resides with
Both parents	80(90)
Single parent	5(6)
other	4(4)
Social class of parents
Upper class (1 to 1.7 )	9(10)
Middle class (1.8 to 3.3)	52(58)
Lower class (3.4 to 5)	28(32)
Prior treatment
Yes to medications	67(75)
No to medications	22(25)
Prior Medications (N= 67)
Prescribed	28(42)
Over-the-counter	29(43)
Herbal	10((15)

Data are expressed in (%) unless stated otherwise.

Forty-six (52%) of the children were malnourished, of whom 32(36%) were severely malnourished. The median hemoglobin level was 8.6(IQR 6.3, 11) g/dl, and 41(46%) patients were severely anemic (with a hemoglobin level of less than 8g/dl), while 10(11%) had dysmorphic features suggestive of Down syndrome. Sixty-seven (75%) had moderate heart failure according to the modified Ross classification. Non-cardiac disorders contributed to heart failure in 74(83%) cases. The most common causes of heart failure were bronchopneumonia 32(36%), sepsis 21(24%), and congenital heart disease 15(17%). A blood transfusion was given to 41(46%) patients, while 64(72%) received furosemide (Table 2).

Table 2

Variable	All patients (n=89) (%)
Nutritional status Z-score (median, IQR)	-2.1 (-3.7, -0.5)
>2SD (Overweight)	4(4)
2 to -2SD (Normal)	39(44)
<-2 to -3SD ( Moderate Malnutrition)	14(16)
<-3SD (Severe Malnutrition)	32(36)
Hemoglobin in g/dl (median, IQR)	8.6 (6.3, 11)
Hb <8g/dl	41(46)
Hb >8g/dl	48(54)
Dysmorphic features
Yes	10(11)
No	79(89)
Modified Ross score
3 to <7 (Mild HF)	20(23)
7 to <11 (Mod HF)	67(75)
11 to 12 (severe HF)	2(2)
Cause of Heart Failure
Bronchopneumonia	32(36)
Sepsis	21(24)
Congenital heart defect	15(17)
Severe Malaria	10(11)
Sickle cell	8(9)
Tuberculosis	3(3)
Treatment received
Blood transfusion	41(46)
Furosemide	64 (72)
Captopril	15(17)
Spironolactone	15(17)
Digoxin	15(17)

Seventy-four (83%) of the patients were discharged home, and the median time for resolving heart failure was 60 hours. Seventy-five percent of the discharged patients were discharged by 108 hrs. Of the 10(11%) who died, the median time of death was 180 hrs (range 2 to 24 days), while the median time for those who left against medical advice was 156 hrs. One patient was referred to another hospital within 2 h of admission due to a health worker strike that interrupted services (Table 3). Of the 10 children who died, 3 had bronchopneumonia and 5 had sepsis; only one had a diagnosed congenital heart defect and one had severe malaria. The median time to resolution of heart failure with a modified Ross score <2 was significantly longer for malnourished children than those with a normal nutritional status (72 Vs 48hrs, log-rank : 0.001; see Figure 2). Although a modified Ross score >7 was associated with a relatively longer recovery time (72 vs, 48 hrs.), the difference was not statistically significant (log-rank: 0.2; see Figure 3). Children with a modified Ross score >7 and those with sepsis were more likely to die, with odds ratios of 8.8 (CI 1.2 to 72.5, p 0.02) and 3.9 (CI 1.01 to 15.2, p 0.04), respectively.

Table 3

Outcome	N (%)	Median in hours (1QR)
Heart failure resolved	74(83)	60 (60; 108)
Death	10(11)	180 (60; 348)
Discharged against medical advice	4(5)	156 (48;216)
Referred	1(1)	2 (2)

Figura 2

Figura 3

Children less than 2 years of age were more likely to have poor outcomes (OR 3.1, CI 1.2 to 8.5, p = 0.02), which were defined as persistence of heart failure 4 days after admission or death. Residential location and family size did not affect the odds of a poor outcome. Dysmorphic features significantly increased the odds of a poor outcome (OR 7.2 95% CI 1.5 to 36, P = 0.009), while gender, nutritional status, modified Ross score at admission, prior treatment and anemia did not affect the odds of a poor outcome. Heart failure due to congenital heart disease was associated with a higher likelihood of a poor outcome OR 3.6, 95% CI 1.1 to 12, P = 0.02. Each unit increase in Z-score decreased the odds of a poor outcome (OR 0.77, CI 0.63 to 0.95, p = 0.016; see Table 4).

Table 4

Variable	Poor Outcome (n= 36)(%)	Odds ratio	95% confidence interval		P
			Lower	Upper
Residence
Urban	32(44)	2.6	0.7	8.8	0.1
Rural	4(24)
Who child lives with
Both parents	35(45)	0.1	0.01	1.3	0.06
Others	1(11)
Age
<2yrs	29(49)	3.1	1.2	8.5	0.02
>2yrs	7(23)
Gender
Female	20(45)	0.6	0.3	1.6	0.34
Male	16(35)
Referred
Yes	18(50)	1.9	0.8	4.6	0.13
No	18(34)
Social class
Upper	8(38)	1.1	0.4	3.2	0.8
Lower	28(41)
Family size
>4 children	7(35)	1.3	0.5	3.7	0.6
<4 children	29(42)
Nutritional status
Severe Malnutrition	16(50)	1.9	0.8	4.5	0.17
No severe Malnutrition	20(35)
Z-score Weight/age	-	0.77	0.63	0.95	0.016
Dysmorphia
Yes	8(80)	7.2	1.5	36	0.009
No	28(35)
Modified Ross >7
Yes	22(61)	1.5	0.6	3.5	0.3
No	141(38)
Prior treatment
Yes	7(32)	1.6	0.5	4.5	0.3
No	29(43)
Herbs
Yes	5(50)	1.5	0.4	5.6	0.3
No	31(39)
PCV <24
Yes	14(34)	0.6	0.3	1.4	0.2
No	22(46)
Sepsis
Yes	9(42)	1.1	0.4	3.0	0.06
No	27(40)
Bronchopneumonia
Yes	10(31)	0.5	0.2	1.3	0.1
No	26(46)
Coronary heart disease
Yes	10(67)	3.6	1.1	12	0.02
No	26(35)
Malaria
Yes	2(20)	0.3	0.6	1.6	0.1
No	34(43)
Blood transfusion
Yes	15(36.5)	0.7	0.3	1.7	0.5
No	21(43)

Using the clinical relevance of variables for non-cardiac causes of heart failure and a p-value cut-off of 0.1 to select the model of best fit, the following variables were added into multivariable logistic regression model (Table 5): age less than 2 years, weight-for-age score, and modified Ross score >7. This model explained 13.9% (R²) of the variance in outcome of heart failure and correctly classified 66.3% of the cases. The area under curve for the predictability of poor outcome was 64% (95% CI; 52 to 76%) Figure 4.

Table 5

		B	SE	Wald	df	Sigma	Exp(B)	95% CI for EXP(B)
								Lower	Upper
Step 1a	Z-scoreWA	-.202	.112	3.261	1	.071	.817	.656	1.017
	<2yr(1)	.837	.535	2.449	1	.118	2.310	.810	6.594
	Rossgt7(1)	.310	.462	.450	1	.502	1.364	.551	3.375
	Constant	-1.585	.525	9.108	1	.003	.205

Hosmer and Lemeshow Test: Sig: 0.5, Nagelkerke R Square: .139 B: / Exp(B): / Wald: / SE: standard error / df: degrees of freedom / <2yr(1): age less than 2 years / Z-scoreWA: weight-for-age score / Rossgt7(1): modified Ross score >7.

Figura 4

This study has some limitations. Since the data was collected for registry purposes, it involved the limitations inherent to a retrospective study, including possible confounders that were not measured and problems with missing data.