Epidemiology and Outcomes of Acute Decompensated Heart Failure in Children

The PC⁴ is a quality improvement collaborative that collects data on all patients with primary cardiac disease admitted to the CICU service of participating hospitals.¹⁷ PC⁴ maintains a clinical registry to support research and quality improvement initiatives. At the time of this analysis, 23 centers were submitting cases to the PC⁴ Registry.

Each participating center has a trained data manager who has completed a certification exam. The data managers collect and enter data in accordance with the standardized PC⁴ Data Definitions Manual. The PC⁴ registry shares common terminology and definitions with applicable data points from the International Pediatric and Congenital Cardiac Code, Society of Thoracic Surgeons Congenital Heart Surgery Database, and American College of Cardiology Improving Pediatric and Adult Congenital Treatment Registry, as previously described.¹⁷ Participating centers are audited on a regular schedule and audit results suggest complete, accurate, and timely submission of data across centers, with the most recent published results demonstrating a major discrepancy rate of 0.6% across 29 476 fields.¹⁸ Because of the sensitive nature of the data collected for this study, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols may be sent to the PC4 Data Coordinating Center at [email protected].

The University of Michigan Institutional Review Board provides oversight for the PC⁴ Data Coordinating Center; this study was reviewed and approved with waiver of informed consent.

Patients ≤18 years of age admitted to CICUs at participating PC⁴ centers between August 2014 and April 2017 were included in this analysis if ADHF was identified at the time of admission to the CICU, either as the primary diagnosis or a concurrent comorbidity. The PC⁴ definition of ADHF, established by a consensus panel of physicians and nurses within the collaborative, was utilized for this analysis: the presence of documented systolic and/or diastolic ventricular dysfunction and receipt of at least one of the following critical care therapies: (1) continuous infusion of a vasoactive and/or diuretic agent, (2) respiratory support including mechanical ventilation as well as noninvasive high flow nasal cannula, continuous positive airway pressure, or bilevel positive airway pressure, or (3) any form of mechanical circulatory support during the CICU admission. Participating PC⁴ institution data collection teams were made up of clinical data abstractors as well as clinical champions (CICU physicians) who collaborated to ensure accurate interpretation of electronic medical records, including noninvasive imaging reports. Additional PC⁴ definitions of clinical conditions and complications utilized in this analysis can be found in Table I in the Data Supplement.

Cardiac diagnoses and operative procedures are defined according to the International Pediatric and Congenital Cardiac Code.¹⁹ The ADHF population was stratified by presence of underlying CHD versus lack of CHD (ie, structurally normal heart) and/or those having undergone prior heart transplantation, regardless of history of CHD before transplantation. Patients with ADHF with structurally normal hearts were grouped together with those having undergone prior heart transplantation. Patients undergoing any cardiothoracic surgical procedure during the ADHF CICU encounter, including extracorporeal membrane oxygenation (ECMO), ventricular assist device (VAD) placement, and/or heart transplantation, were also included. Therefore, patients with CHD were included if they met criteria for ADHF at the time of admission, whether or not they subsequently had cardiothoracic surgery in the same admission. CICU admissions were excluded from analysis if the patient presented with hospice care/do-not-resuscitate orders.

Twenty-three centers participating in the PC⁴ registry at the time of analysis had submitted qualifying cases. Both hospital-level and CICU admission-level descriptive analyses were performed with all inferential analyses utilizing the CICU admission as the episode of analysis.

Patient demographics (age, sex), presence of extracardiac abnormalities or chromosomal abnormalities/syndromes, weight at CICU admission, CICU resource utilization (mechanical ventilation, pharmacotherapies, arterial/venous access), CICU complications including cardiac arrest, and use of ECMO rescue during active CPR (E-CPR) were chosen a priori as potential factors associated with mortality in the ADHF population. Mortality in the CICU environment is described for the cohort as well as among the CHD and non-CHD groups of patients with ADHF.

Data are presented as frequency (percentage) for categorical variables and median with interquartile range for continuous variables. To identify patient and clinical characteristics associated with CICU mortality, univariate comparisons were performed using the Wilcoxon rank-sum test for continuous variables, and χ² or Fisher exact test as appropriate for categorical data.

Depictions of variability in mortality across PC⁴ centers are not annualized as centers contributed data to the registry at different time points during the study era. Raw (observed) mortality rates are presented as percentages. Several variables identified a priori as possible predictors of mortality were subsequently excluded from the final model.

To better understand predictors of mortality, we performed a stratified univariate analysis dividing the cohort into patients who underwent a cardiothoracic surgical procedure during their ADHF admission (N=310) versus those who did not (N=1184). Only nonsurgical ADHF admissions were subsequently included in our multivariate analysis of mortality for 2 reasons: (1) key differences in variables collected between surgical and nonsurgical patients made combining them in one model potentially misleading and (2) the small sample size of the surgical cohort limited additional multivariable analysis in this subgroup. Factors associated with CICU mortality among nonsurgical patients in unadjusted analysis (P<0.1) were included in a multivariable logistic regression model to determine independent associations (P<0.05) with mortality. We also tested models including laboratory values such as B-type natriuretic protein and creatinine, but there was a high percentage of missing values as these studies were not obtained on all patients at the time of admission. We evaluated missingness and found no difference in the distribution of missing data between patients who died and those survived. Addition of laboratory values (actual and predicted) did not improve model performance and resulted in larger standard errors for the coefficients, so we excluded laboratory variables from our final multivariable analysis. These results are not presented further.

Adjusted odds ratios and 95% CI for each predictor are reported. We ran the model accounting for clustering at the hospital-level using a hospital random-effect term. We also evaluated the model adding a random-effect term at the patient-level to account for multiple CICU admissions by the same patient.

We could not perform stratified multivariable analyses in the 2 disease subgroups (CHD and non-CHD + status posttransplant) due to sample size constraints. Therefore, to explore the impact of disease heterogeneity on predictors of mortality, we performed a second multivariable analysis and tested interactions with CHD. Key-independent variables associated with mortality in the initial model (age and exposure to mechanical ventilation, vasoactive infusions, and ECMO) were included with diagnosis as interaction terms to test whether they differentially impacted the effect of diagnosis on mortality. All analyses were performed using SAS Version 9.4 (SAS Institute, Cary, NC) or STATA Version 14 (StataCorp LLC, TX), with statistical significance at a P value <0.05.

Between August 1, 2014 and April 4, 2017, a total of 26 294 CICU admissions were submitted to the PC⁴ collaborative by 23 participant centers. Of these CICU admissions, 1494 unique ADHF admissions (1371 hospitalizations) were identified (Figure 1). Table 1 displays demographic, clinical, and laboratory characteristics at the time of CICU admission for the entire cohort as well as CHD and non-CHD subgroups. Patients with underlying CHD comprised 57% (N=852) of admissions. The median age at presentation for ADHF was 0.93 years (interquartile range [IQR], 0.1–9.3 years), and patients with CHD were significantly younger than patients without CHD. Patients with ADHF and CHD had significantly lower weight at CICU admission and were more likely to have extracardiac congenital anomalies, chromosomal abnormalities, and/or known syndromes when compared with patients without CHD. Only 4% of patients (58/1494) had undergone a previous cardiothoracic surgical procedure during the same hospitalization and among patients with prior heart transplantation presenting in ADHF, 80% (32/40) were identified as having rejection at the time of admission.

Therapies, procedures, and complications in the CICU are displayed in Table 2. Nearly 60% of ADHF admissions included invasive mechanical ventilation, for a median duration of 5.2 days (IQR, 2–13 days). Patients with CHD were more likely to receive mechanical ventilation and had higher rates of reintubation within 48 hours of planned extubation when compared with their non-CHD counterparts. Table II in the Data Supplement displays the spectrum of vasoactive/inotropic medications used in the care of patients with ADHF. Milrinone was the most commonly administered vasoactive agent (80% of admissions). Patients without CHD were treated with milrinone more frequently than their CHD counterparts while patients with CHD were exposed to more calcium and vasopressin infusions than patients without CHD. The frequency of vasoactive medication prescription varied considerably between individual centers. Milrinone had the highest rate of utilization during ADHF admissions among sites, ranging from 60% to 100% of ADHF admissions across the consortium (Table III in the Data Supplement).

The median length of CICU stay for all ADHF admissions was 7.9 days (IQR, 3.0–18.1 days, Table 3). Length of CICU stay was significantly longer for patients with CHD than patients without CHD (8.4 days [IQR, 3.2–18.9] versus 6.8 days [IQR, 3.0–17.5]; P=0.027). No difference in ICU readmission rates was appreciated between the 2 groups (179/852 [23%] CHD versus 122/642 [21%] non-CHD; P=0.43), although 22% of ADHF admissions were classified as an unplanned readmission to the CICU within 48 hours of previous discharge (N=301). Readmission to the hospital within 30 days (all-cause hospital readmission) was noted in 11% (N=147) of patients with ADHF with a higher readmission rate seen in the CHD cohort.

Complication rates are also displayed in Table 2. The most common complication was arrhythmia treated with pharmacological therapy and/or cardioversion/defibrillation. Of all complications, only central line-associated blood stream infection was significantly greater among patients with CHD in comparison to the non-CHD cohort. Stroke was observed in 69 (5%) ADHF CICU encounters. Of these encounters, 35 (50%) were in patients undergoing mechanical circulatory support via VAD and/or ECMO (VAD only 9/35 [26%], ECMO only 22/35 [63%], VAD, and ECMO 4/35 [11%]).

The most severe morbidity, cardiac arrest, occurred in 10% of ADHF admissions (N=152). No differences were noted between CHD and non-CHD subgroups for cardiac arrest incidence nor for cardiac arrest duration. E-CPR was utilized in over a third of cardiac arrests (35%, N=53/152).

Disparate utilization rates were noted between ECMO and VAD. While ECMO was utilized in 12% of the overall cohort (N=173), only 4% (N=65) underwent VAD implantation during the CICU admission. Of the 65 ADHF admissions in which a VAD was implanted, 24 (36.9%) also underwent heart transplantation during the same hospitalization. Patients without CHD were more likely to be both listed for heart transplantation (83/642 [13%] versus 76/852 [9%]; P=0.013) and undergo heart transplantation during the same hospitalization. ECMO support and VAD implantation were also more commonly used by patients without CHD than their CHD counterparts (ECMO 89/642 [14%] versus 84/852 [10%]; P=0.017 and VAD 51/642 [8%] versus 14/852 [2%]; P<0.001).

The CICU mortality rate for the overall ADHF cohort was 15% (N=231). Variation in mortality across the participating PC⁴ centers is displayed in Figure 2. Mortality in the CICU ranges from 0% (single center with n=6 total admissions entered into PC⁴ at the time of data abstraction) to as high as 25% of CICU admissions per center. In contrast to CICU-level outcomes, hospitalization-level morbidity and mortality are described in Table IV in the Data Supplement. Hospital mortality reached 19% for the overall ADHF cohort and similar to CICU mortality rates was much higher for patients with CHD.

We evaluated the association of patient characteristics, coexisting high-risk diagnoses, therapies received, and complications with CICU mortality solely in patients who did not undergo cardiothoracic surgical procedures during their CICU admission (n=1184; Table 4). Presence of CHD, extracardiac anomalies, neonatal age at presentation, preexisting high-risk diagnoses (CPR as reason for admission; any cardiomyopathy including dilated, hypertrophic, and restrictive; pulmonary hypertension), peak vasoactive inotrope score within 2 hours of admission, mechanical ventilation, ECMO, and key complications were some of the important univariate risk factors for CICU mortality.

Although a smaller cohort, patients who underwent cardiothoracic surgical procedures (N=310) during ADHF admissions were analyzed for univariate associations with mortality with results presented as Table V in the Data Supplement. Heart transplants were performed for a total of 66 admissions. Overall mortality was 10% and of those who died, 2 (6%) had undergone transplantation during CICU admission. Cardiac arrest occurred in 11% of encounters with ECPR utilized in 43% of arrests.

Several factors persisted as independent risk factors for CICU mortality for the larger cohort of nonsurgical patients after multivariable logistic regression (Table 5). Independent risk factors associated with CICU mortality included age <30 days, presence of CHD, any use of vasoactive infusion, ventricular tachycardia, mechanical ventilation, sepsis, pulmonary hypertension, use of ECMO, and cardiac arrest. When classified by CPR duration categories, admissions experiencing cardiac arrest of >20 minutes duration were found to have higher odds of CICU mortality than cardiac arrest of <20 minutes and all other variables included in the multivariate model.

In our exploratory model to test interactions between CHD, age, vasoactive infusions, mechanical ventilation, and ECMO, only the interaction between age and CHD was statistically significant (P=0.002). Although all age groups interacted with CHD to increase the risk of mortality compared with non-CHD, the strongest effect was the interaction between CHD and neonates; neonates with CHD and ADHF had a 16-fold greater odds of mortality than the reference group (children >1 year old, no CHD; adjusted OR 16.2 [95% CI, 7.2–36.4], P<0.001; Table VI in the Data Supplement).