Pulmonary hypertension among children with in-hospital cardiac arrest: a multicenter study

Abstract

Aims:

To determine the prevalence of pulmonary hypertension (PH) among children with in-hospital cardiac arrest (IHCA) and its association with survival.

Methods:

Children (<18 years) admitted to ICUs participating in the Virtual Pediatric Systems multicenter registry between January 2011 and December 2017 who had an IHCA during their hospitalization were included. Patients were classified by whether they had a documented diagnosis of PH at the time of IHCA. Clinical characteristics were compared between patients with and without PH. After propensity score matching, conditional logistic regression within the matched cohort determined the association between PH and survival to hospital discharge.

Results:

Of 18,575 children with IHCA during the study period, 1,590 (8.6%) had a pre-arrest diagnosis of PH. Patients with PH were more likely to be 29 days to 2 years of age, female, Black/African American, and American Indian/Alaskan Native, and to be treated in a cardiac ICU or mixed PICU/cardiac ICU. At ICU admission, PH patients had a lower probability of death as determined by the Pediatric Index of Mortality 2 (PIM-2) score. Patients with PH were more likely to be receiving inhaled nitric oxide (13.0% vs. 2.1%; p<0.001). Propensity score matching successfully matched 1,302 PH patients with 3,604 non-PH patients. Patients with PH were less likely to survive to hospital discharge (aOR 0.83; 95% CI: 0.72–0.95; p=0.01) than non-PH patients.

Conclusions:

In this large multicenter study, 8.6% of children with IHCA had pre-existing documented PH. These children were less likely to survive to hospital discharge than those without PH.

INTRODUCTION

Among adults who suffer an in-hospital cardiac arrest (IHCA), pre-existing pulmonary hypertension (PH) is associated with dismal outcomes.¹ Laboratory data indicate that PH with elevated pulmonary vascular resistance (PVR) during cardiopulmonary resuscitation (CPR) decreases pulmonary blood flow, cardiac output, and coronary and cerebral perfusion.^2–4 Hospitalized children with PH are at high risk of mortality and IHCA.^5–8 Thus, expert consensus postulates that PH-associated pediatric cardiac arrest portends poor outcomes,^9,10 but published data on pediatric IHCA associated with PH are limited.

In a recent small, retrospective study, 35% of children with IHCAs in the intensive care unit (ICU) had echocardiographic evidence of PH prior to cardiac arrest.¹¹ However, this study was limited to a single institution and did not identify an association between PH and cardiac arrest outcomes. Therefore, the objectives of the current study were to leverage a large, multicenter database to: 1) describe the prevalence of PH among children with IHCA and 2) compare survival outcomes between IHCA patients with and without pre-existing PH in a propensity matched cohort.

MATERIALS AND METHODS

Data Source

This study utilized the Virtual Pediatric Systems (VPS, LLC; Los Angeles, California) Registry, which contains data from diverse institutions providing care for critically ill children in North America. The web-based registry, which was developed in a partnership between the Children’s Hospital Association and the Children’s Hospital of Los Angeles, provides a prospective, observational cohort of consecutive pediatric ICU admissions. Data are collected and entered by trained individuals at each ICU site using standardized clinical definitions and quality control systems. The VPS staff performs extensive quality validation prior to the release of data to enhance reliability. The inter-rater reliability of the VPS registry data elements is consistently greater than 95%.¹² The Institutional Review Board (IRB) at the Children’s Hospital of Philadelphia deemed that this study was not human subjects research (IRB 18–015164).

Patient Population and Clinical Definitions

Cardiac arrest was defined within the VPS registry as any event characterized by either pulselessness or critically compromised perfusion treated with external chest compressions and/or defibrillation.^13,14 This definition was utilized by staff at each site to enter patients into the database with pertinent International Classification of Diseases, Ninth and Tenth Revisions (ICD-9; ICD-10) codes, which were also used to describe other patient characteristics. For this study, the VPS registry was queried for children <18 years of age who were admitted to a participating ICU between January 1, 2011 and December 31, 2017 and had a cardiac arrest during their hospitalization. The data pertaining to these patients was then provided by VPS to the study team. Only data regarding index cardiac arrests (the first IHCA occurring during the hospitalization) were included. Patients with incomplete data regarding age or sex were excluded.

Data on patient demographics, severity of illness, pre-existing conditions, ICU interventions, and outcomes were collected. Demographic information included age, sex, and race/ethnicity. Severity of illness was determined by calculation of the Pediatric Index of Mortality 2 (PIM-2) score at ICU admission.¹⁵ Individual components of the PIM-2 score were entered by data abstractors at each site and the score and corresponding risk of mortality were automatically calculated within the database and then provided to the study team. Pediatric Cerebral Performance Category (PCPC) scores were determined as a measure of pre-hospitalization neurodevelopmental status.¹⁶ Specific pre-existing conditions present prior to the time of IHCA and the category of the primary hospital diagnosis were reported, as were ICU-level interventions in place at the time of IHCA (e.g., arterial catheter, invasive mechanical ventilation). All patients were categorized according to whether they already had an active diagnosis of PH (ICD-9 codes 416.0 or 461.8A) at the time of IHCA. Intensive care unit size (number of beds) and type (pediatric [PICU], pediatric cardiac [CICU], or mixed) were collected. The primary outcome was survival to hospital discharge.

Statistical Analyses

Descriptive statistics and unadjusted univariate analyses:

Categorical data were described as frequencies and percentages and compared between patients with and without PH using Chi-square test or Fisher’s exact test. Continuous data were described as medians with interquartile ranges (IQR) and compared using Wilcoxon rank-sum test. Categorical data included whether inhaled nitric oxide (iNO) was being administered at the time of IHCA. This field was included due to its relevance as a treatment for PH. However, some sites did not report iNO exposure, as it is an optional field in the VPS registry. To optimize accuracy, iNO use was calculated only among ICUs reporting iNO use in at least one patient during the study period.

Matching of patients with and without PH:

Matching was performed in order to facilitate evaluation of survival outcomes between patients with and without PH. Propensity score matching was utilized to match a PH patient with up to four non-PH patients. The matching was stratified within each unit type (PICU, CICU, mixed). Exact matching was required for age group, baseline PCPC score, and ICU size category. Other demographic variables, pre-existing conditions, and interventions in place were incorporated into the propensity score unless pre-matching mean difference was ≤0.02. In addition to exact matching methods, the matching algorithm utilized a nearest neighbor with specified caliper (caliper=0.1) method. Interaction terms were included in the matching models to improve balance and avoid excessive exclusion of PH cases. Matching was performed using matchit, an R package (R Foundation for Statistical Computing, Vienna, Austria). The variables used for PSM and the results of the matching balance before and after PSM are detailed for each unit type in Supplemental Tables 1–3.

Logistic regression:

Within the propensity-matched cohort, logistic regression with conditioning for matched cases/controls was performed with survival to hospital discharge as the outcome and PH diagnosis as the predictor variable. The adjusted odds ratios of survival to hospital discharge in the PH group were reported, and the predicted probabilities of survival to discharge were reported in each of the two groups. Due to the optional nature of PCPC within the registry, neurological outcome at hospital discharge was not assessed.

RESULTS

Overview and PH prevalence:

During the 7-year study period, 19,199 patients had IHCAs out of 982,626 ICU admissions (1.95%). After applying all inclusion/exclusion criteria, 18,575 children in 164 ICUs were included in the study cohort with 1,590 (8.6%) having a pre-arrest diagnosis of PH (Figure 1).

Figure 1. Study flow diagram.

ICU = intensive care unit, IHCA = in-hospital cardiac arrest, PH = pulmonary hypertension.

Comparisons between patients with and without PH:

Patient demographics and ICU characteristics for the overall study cohort and for patients with and without PH are reported in Table 1. Patients with PH were more likely to be 29 days to 2 years of age, female, Black / African American, and American Indian / Alaskan Native and were less likely to have a normal baseline PCPC. Patients with PH were also more likely to be treated in a CICU or mixed PICU/CICU. Clinical characteristics are reported in Table 2. Patients with PH were more likely to have primary hospital diagnoses in the Respiratory and Cardiovascular categories and less likely to have primary diagnoses in the Trauma, Oncologic, or Neurologic categories. At the time of ICU admission, patients with PH had lower PIM-2 scores (−3.1 [−4.2, −1.7] vs. −1.7 [−3.2, −0.2]; p<0.001) with a corresponding lower PIM-2-determined probability of death (4.2 [1.4, 15.9]% vs. 15.5 [4.0, 46.1]%; p<0.001). The PH group had higher prevalence of many pre-existing conditions and more interventions in place at the time of IHCA, including documented iNO administration (13.0% vs. 2.1%; p<0.001) among children treated in ICUs reporting iNO data.

Table 1.

Patient and ICU Characteristics.

	Overall (N =18,575)	PH (N = 1,590)	No PH (N = 16,985)	P-value
Age				<0.001
≤28 days	1,860 (10.0%)	173 (10.9%)	1,687 (9.9%)	0.25
29 days to <2 years	8,299 (44.7%)	977 (61.5%)	7,322 (43.1%)	<0.001
2 years to <12 years	5,369 (28.9%)	328 (20.6%)	5,041 (29.7%)	<0.001
≥ 12 years	3,047 (16.4%)	112 (7.0%)	2935 (17.3 %)	<0.001
Sex				<0.001
Female	7,955 (42.8%)	782 (49.2%)	7,173 (42.2%)	<0.001
Male	10,620 (57.2%)	808 (50.8%)	9,812 (57.8%)	<0.001
Race / Ethnicity				<0.001
White	7,093 (38.2%)	585 (36.8%)	6,508 (38.3%)	0.24
Black or African American	3,362 (18.1%)	322 (20.2%)	3,040 (17.9%)	0.02
Hispanic or Latino	2,345 (12.6%)	210 (13.2%)	2,135 (12.6%)	0.49
Asian / Indian / Pacific Islander	558 (3.0%)	49 (3.1%)	509 (3.0%)	0.91
American Indian or Alaskan Native	206 (1.1%)	36 (2.3%)	170 (1.0%)	<0.001
Other / Mixed	1,083 (5.8%)	99 (6.2%)	984 (5.8%)	0.52
Unknown or Not Reported	3,928 (21.2%)	289 (18.2%)	3,639 (21.4%)	0.003
Baseline PCPC Score				<0.001
1 - Normal	3,337 (18.0%)	186 (11.7%)	3,151 (18.5%)	<0.001
2 - Mild disability	694 (3.3%)	119 (7.5%)	575 (3.4%)	<0.001
3 - Moderate disability	606 (3.3%)	124 (7.8%)	482 (2.8%)	<0.001
4 - Severe disability	381 (2.0%)	42 (2.6%)	339 (2.0%)	0.10
5 - Coma / Vegetative State	17 (0.1%)	2 (0.1%)	15 (0.1%)	0.65
Missing	13,540 (72.9%)	1,117 (70.3%)	12,423 (73.1%)	0.01
Unit Type				<0.001
PICU	9,109 (49.0%)	590 (37.1%)	8,519 (50.1%)	<0.001
CICU	1,777 (9.6%)	255 (16.0%)	1,522 (9.0%)	<0.001
Mixed	7,689 (41.4%)	745 (46.9%)	6,944 (40.9%)	<0.001
Unit Size				<0.001
<15 beds	1,799 (9.7%)	109 (6.9%)	1,690 (10.0%)	<0.001
15–20 beds	4,572 (24.6%)	438 (27.5%)	4,134 (24.3%)	0.005
21–26 beds	4,787 (25.8%)	382 (24.0%)	4,405 (25.9%)	0.10
>26 beds	7,417 (39.9%)	661 (41.6%)	6,756 (39.7%)	0.17

PH = pulmonary hypertension; PCPC = Pediatric Cerebral Performance Category; PICU = pediatric intensive care unit; CICU = pediatric cardiac intensive care unit.

Table 2.

Pre-Arrest Clinical Characteristics.

	Overall (N = 18,575)	PH (N = 1,590)	No PH (N = 16,985)	P-value
Primary hospital diagnosis category
Trauma	2,331 (12.6%)	14 (0.9%)	2,317 (13.6%)	<0.001
Respiratory	3,638 (19.6%)	462 (29.1%)	3,176 (18.7%)	<0.001
Oncologic	363 (2.0%)	12 (0.8%)	351 (2.1%)	<0.001
Endocrinologic	57 (0.3%)	1 (0.1%)	56 (0.3%)	0.09
Neurologic	1,368 (7.4%)	35 (2.2%)	1,333 (7.9%)	<0001
Cardiovascular	6,356 (34.2%)	784 (49.3%)	5,572 (32.8%)	<0.001
Other	4,455 (24.0%)	280 (17.6%)	4,175 (24.6%)	<0.001
Pre-existing conditions (at time of arrest)
Chronic lung disease	2,762 (14.9%)	678 (42.6%)	2,084 (12.3%)	<0.001
Acute respiratory insufficiency	13,226 (71.2%)	1,263 (79.4%)	11,963 (70.4%)	<0.001
Respiratory infection	3,200 (17.2%)	436 (27.4%)	2,764 (16.3%)	<0.001
Congenital heart disease	5,104 (27.5%)	1,007 (63.3%)	4,097 (24.1%)	<0.001
Heart failure / Cardiomyopathy / Myocarditis	2,313 (12.4%)	484 (30.4%)	1,829 (10.1%)	<0.001
Pulmonary embolism	98 (0.5%)	14 (0.9%)	84 (0.5%)	0.04
Shock	3,038 (16.4%)	256 (16.1%)	2,782 (16.3%)	0.77
Sepsis / Bacteremia	2,578 (13.9%)	280 (17.6%)	2,298 (13.5%)	<0.001
Trauma / Burn	2,461 (13.2%)	96 (6.0%)	2,365 (13.9%)	<0.001
Drowning / Near-drowning	948 (5.1%)	7 (0.4%)	941 (5.5%)	<0.001
Renal failure	2,229 (12.0%)	246 (15.5%)	1,983 (11.7%)	<0.001
Malignancy / Bone marrow transplant	1,154 (6.2%)	53 (3.3%)	1,101 (6.5%)	<0.001
Hepatic dysfunction	1,143 (6.1%)	102 (6.4%)	1,041 (6.1%)	0.65
Seizure disorder	3,177 (17.1%)	141 (8.9%)	3,036 (17.9%)	<0.001
Stroke	1,180 (6.3%)	142 (8.9%)	1,038 (6.1%)	<0.001
Other CNS dysfunction	5,919 (31.9%)	354 (22.3%)	5,565 (32.8%)	<0.001
PIM-2 Score	−1.8 [−3.3, −0.3]	−3.1 [−4.2, −1.7]	−1.7 [−3.2, −0.2]	<0.001
PIM-2 Probability of Death	14.8 [3.7, 41.6]	4.2 [1.4, 15.9]	15.5 [4.0, 46.1]	<0.001
Interventions in place at time of arrest
Central venous catheter	5,721 (30.8%)	679 (42.7%)	5,042 (29.7%)	<0.00
Arterial catheter	3,280 (17.7%)	376 (23.7%)	2,904 (17.1%)	<0.001
Invasive mechanical ventilation	10,983 (59.1%)	908 (57.1%)	10,075 (59.3%)	0.09
Advanced mode of ventilation (HFOV, jet ventilation, liquid ventilation)	113 (0.6%)	23 (1.4%)	90 (0.5%)	<0.001
Conventional mechanical ventilation	10,870 (58.5%)	885 (55.7%)	9,985 (58.8%)	0.02
Non-invasive ventilation (CPAP/BiPAP)	343 (1.9%)	53 (3.3%)	290 (1.7%)	<0.001
High-flow nasal cannula	438 (2.4%)	86 (5.4%)	352 (2.1%)	<0.001
Inhaled nitric oxide (among reporting ICUs)	354 / 11,052 (3.2%)	145 / 1,113 (13.0%)	209 / 9,939 (2.1%)	<0.001
Extracorporeal membrane oxygenation	302 (1.6%)	43 (2.7%)	259 (1.5%)	<0.001
Cardiac surgery during hospitalization	2,873 (15.5%)	413 (26.1%)	2,460 (14.5%)	<0.001
Renal replacement therapy	41 (0.2%)	7 (0.4%)	34 (0.2%)	0.05

PH = pulmonary hypertension; CNS = central nervous system; PIM-2 = Pediatric Index of Mortality 2; HFOV = high-frequency oscillatory ventilation; CPAP = continuous positive airway pressure; BiPAP = bilevel positive airway pressure; ICU = intensive care unit.

Propensity matching:

Of the 1,590 patients with PH, PSM led to the successful matching of 1,302 (81.9%) to 3,604 patients without PH. After matching, patient and event characteristics were similar between groups within all three ICU types (Supplemental Tables 1–3).

Outcomes within the matched cohort:

In conditional multivariable LR analysis of the matched cohort, patients with PH were less likely to survive to hospital discharge: aOR 0.83 (95%CI: 0.72–0.95; p=0.01) with a predicted survival rate of 59.1% (56.5–61.8%) in the PH group and 61.6% (60.0–63.2%) in the non-PH group.

DISCUSSION

In this large, multicenter registry study of more than 18,000 children with IHCA, 1,590 (8.6%) had a documented pre-existing diagnosis of PH, demonstrating that children with PH represent a sizable sub-population of children requiring CPR. These children had lower odds of survival to hospital discharge than children without PH after propensity matching and controlling for residual confounding factors. This study represents, to the best of our knowledge, the first dedicated multicenter report of the prevalence and outcomes of PH-associated pediatric cardiac arrests and provides an important epidemiological foundation to begin understanding this understudied problem.

Resuscitation of children with PH is recognized as a unique challenge in pediatric advanced life support guidelines as these children may benefit from therapies outside of standard algorithmic resuscitation care.^10,17,18 Thus, the 8.6% prevalence of documented PH among children with IHCAs in this large cohort is an important finding as it represents a substantial number of patients for whom standard resuscitation therapies may be less than optimal. These patients were younger than those without PH – an unsurprising finding given the common association of pediatric PH with congenital heart disease and chronic lung disease,¹⁹ which were present in 63% and 43% of PH patients in this study, respectively. The observed prevalence of documented PH was considerably lower than that the 35% reported in our single center cohort study of pediatric IHCA.¹¹ However, our prior study used echocardiographic criteria to identify PH, and only 11 of the 20 patients with PH by echocardiography had a documented PH history. Together, these findings suggest that many patients without a history of PH develop alterations in PVR as a result of their acute illness and that PH physiology may be clinically underrecognized and underreported.²⁰ Thus, the true prevalence of PH at the time of IHCA is presumably higher than commonly recognized clinically.

After controlling for documented confounders with propensity score matching and logistic regression, children with PH were less likely to survive to hospital discharge (aOR 0.83; p=0.01) as compared to IHCA patients without PH. This large-scale epidemiologic evidence supports expert opinion regarding this population of critically ill children.^10,18,21 We postulate that some children with pre-existing PH have pulmonary hypertensive crises due to acute pulmonary vasoconstriction from acute hypoxia, acidosis, and/or noxious stimuli leading to severe hypotensive shock and resultant cardiac arrest. In addition, children with pre-existing PH and right ventricular (RV) dysfunction are at risk of less acute, progressive deterioration from the physiologic disturbances that accompany common critical illnesses, such as sepsis, pneumonia, or metabolic derangements. In either situation, the increased PVR in these patients during the low-flow state of CPR can limit pulmonary blood flow, left heart preload, cardiac output, and coronary perfusion, and thus result in worse outcomes.^2–4 Some patients with pre-arrest PH and RV hypertrophy or RV dilation may also need higher RV perfusion pressures during CPR to provide adequate myocardial oxygenation for successful resuscitation.^22–25

Given this high-risk pathophysiology, it is perhaps surprising that the magnitude of the difference in survival rates is not greater. The predicted survival to discharge rate among these children with IHCA and pre-existing PH was 59.1%, in contrast to the 6% survival reported among adults with ICHA and pre-existing PH.¹ As children with PH are known to have a high likelihood of suffering a cardiac arrest during their hospitalization,^{5–7,10,26–28} clinicians may monitor their clinical status more closely and provide CPR more promptly at a time when successful resuscitation is more likely. For example, CPR is recommended and commonly provided to children with bradycardia and poor perfusion prior to true pulselessness.¹⁷ Since bradycardia and poor perfusion was the initial CPR rhythm in 50% of children with PH and IHCA in our single-center study,¹¹ it may be that the early provision of CPR mitigates the severity of these events, and thus prevents poor outcomes in some patients. Notably, CPR for this rhythm is consistently associated with more favorable outcomes than CPR for other rhythms in several multicenter studies.^29–31 We could not evaluate the relative frequency of CPR for bradycardia and poor perfusion among IHCA patients with PH, as initial rhythms at time of CPR are not available in the VPS registry.

In animal studies of cardiac arrest associated with elevated PVR, intra-arrest iNO therapy improves hemodynamics during CPR and results in better outcomes.^2–4,32 Thus, children with PH-associated cardiac arrest likely benefit from peri- and intra-arrest therapies specifically directed toward reducing PVR and supporting the failing RV, such as iNO, inhaled prostaglandins, intravenous vasodilators, and early provision of extracorporeal life support.^{1,10,27,28,33,34} The VPS registry has limited medication data, but iNO is coded as a procedure and data regarding its use were available for analysis at some sites. Of the 1,113 patients with PH in ICUs reporting iNO data, 145 (13.0%) were reported to be receiving iNO at the time of IHCA. In contrast, 80% of children with PH received iNO at the time of IHCA in our previously reported single center study.¹¹ Data were not granular enough to determine how frequently iNO was started as a rescue therapy during CPR or whether institution of iNO during CPR affected outcomes. Nevertheless, promising pre-clinical findings² suggest a role for future prospective clinical studies of targeted pulmonary vasodilator therapies during CPR for patients with PH as a means of improving outcomes.

This study’s limitations are largely related to the nature of the dataset itself. First, inaccurate or incomplete data entry related to diagnosis codes or other data elements could have resulted in misclassification of patients. In particular, failure to either diagnose PH or to appropriately enter diagnosis codes for PH patients could have resulted in underestimation of PH prevalence in this study. However, the VPS Registry is large, well-established, undergoes quarterly audits to assure data quality, and has high inter-rater reliability. Second, details regarding cardiac arrest characteristics and the resuscitation care provided were not available. Information pertaining to CPR duration and the immediate outcome of CPR (e.g., return of spontaneous circulation) would be particularly valuable. Third, our outcome assessment was limited to survival to hospital discharge due to a limited number of patients in the dataset having available PCPC data at discharge. Future investigations should include neurological and neurofunctional outcomes to more comprehensively define the morbidity associated with PH-associated IHCA. Fourth, although we performed propensity score matching and subsequently employed conditional logistic regression to address residual confounding, we cannot draw definitive causal associations as this is a retrospective, observational study and unmeasured confounding is possible.

CONCLUSIONS

In this large, retrospective study, 8.6% of children with IHCAs had a documented diagnosis of PH. These patients had lower rates of survival to hospital discharge than children with IHCA without documented PH.