Pulse Oximetry Screening Has Not Changed Timing of Diagnosis or Mortality of Critical Congenital Heart Disease

Abstract

Objectives

This study evaluates the effectiveness of mandatory pulse oximetry screening. The objective is to evaluate whether mandatory pulse oximetry testing had decreased the late CCHD diagnosis rate and reduced mortality in neonatal subjects.

Design

This was a single center, retrospective cohort study comparing the timing of diagnosis of CCHD between neonates undergoing cardiac surgery in 2009–2010, prior to mandatory pulse oximetry screening, and neonates in 2015–2016, after mandatory pulse oximetry screening was instituted. Follow-up was for 1 year. We defined CCHD as lesions requiring surgical correction within 30 days of life. Exclusions included: pacemaker insertions, vascular ring divisions, closure of patent ductus arteriosus, arterial cutdown, or ECMO cannulation without structural heart disease as the sole procedure, or if subjects were born at home. Infants diagnosed prior to discharge from birth hospital were defined as early postnatal; late postnatal subjects were diagnosed after birth hospital discharge. In-hospital mortality and 1-year mortality were measured.

Results

A total of 527 neonates were included; 251 (47.6%) comprised the pre-mandatory pulse oximetry screening cohort (2009–2010). Only 3.6% of the 2009–2010 cohort and 4.3% of the 2015–2016 cohort were diagnosed late (p=0.66). One-year mortality decreased during the study period (17.2% in 2009–2010 versus 10.5% in 2015–2016, P= 0.03). There were no deaths in the late CCHD diagnosis groups.

Conclusions

Mandatory pulse oximetry screening legislation has not changed the late postnatal diagnosis rate at our institution. Mortality for neonatal critical congenital heart disease has declined, but this decline is not attributable to mandatory pulse oximetry screening.

Introduction

Infants with critical congenital heart disease (CCHD) are at risk of death if not recognized and treated early [1,2]. In 2011, The Health and Human Services Secretary recommended that pulse oximetry screening for CCHD be added to the recommended mandatory screening panel, to identify at-risk neonates prior to index hospital discharge [1]. In response, mandatory pulse oximetry screening legislation was put into effect in New Jersey in 2011 and in Pennsylvania in 2014 [3].

Prior literature has shown that the pre-operative factors, including acidosis, occur less frequently in prenatally compared to postnatally diagnosed subjects [4]. Early postnatal diagnosis is also beneficial, as mortality increases for infants with CCHD diagnosed after discharge from birth hospital [2,4]. Meta-analysis prior to mandatory pulse oximetry implementation estimated that screening would have a sensitivity of 69.6% and specificity of 99.9%, based on the results of 10 studies representing 123,846 patients [5]. Utilizing infant birth/death linkage data, investigators recently described decreased CCHD mortality after statewide implementation of mandatory pulse oximetry screening [6]. This registry study was limited by its inability to account for the timing of CCHD diagnosis. In addition, the study could not directly account for advances in CCHD management. Recent single center data reporting on the effect of mandatory pulse oximetry screening of 77,148 newborns were considerably less promising; this study demonstrated that screening had low sensitivity (14.3%) for CCHD detection [7]. Additionally, a multicenter study of 138,176 live births in the United Kingdom showed there was no statistically significant effect of pulse oximetry screening on post discharge diagnosis of CCHD [8]. Given the discrepancy in the reported sensitivity of mandatory pulse oximetry screening, study of more thorough clinical data is necessary to distinguish the effects of the screen on late CCHD diagnosis rate and mortality.

Our study defines neonates diagnosed prior to discharge from birth hospital as early postnatal; late postnatal neonates were diagnosed after birth hospital discharge. We hypothesize that readily available fetal echocardiography and mandatory pulse oximetry are associated with reduced late post-natal CCHD diagnosis rate and that less frequent late CCHD diagnosis is associated with decreased mortality. By comparing late diagnosis rates between a pre-screening and post-screening cohort, we can determine whether mandatory screening has reduced late CCHD diagnosis and mortality.

Materials and Methods

Critical congenital heart disease was defined as defects that required surgical intervention prior to 30 days of life. This age of less than 30 days was chosen to focus on infants with ductal dependent congenital heart lesions, who are a major target of pulse oximetry screening and are at high risk of mortality if not identified quickly. A historical cohort of neonates (<30 days of age at the time of surgery) undergoing cardiac surgery from 2009–2010, prior to mandatory pulse oximetry screening, was compared to a more contemporary cohort from 2015–2016, after pulse oximetry screening was mandated. Neonates were retrospectively identified from the cardiothoracic surgical database maintained by the Children’s Hospital of Philadelphia (CHOP). The CHOP cardiothoracic surgical database was queried for cardiac diagnosis, surgical procedure and STAT mortality category, antenatal diagnosis, age and weight at surgery, in-hospital mortality, and 1-year mortality. All subjects underwent surgery at CHOP. Information on prenatal diagnosis rate, in-hospital mortality, and 1-year mortality are routinely tracked and included in the database, and there was no missing data on mortality. Medical records for all patients with postnatal diagnosis were reviewed to determine timing of diagnosis. Neonates diagnosed prior to discharge from birth hospital were defined as early postnatal; late postnatal neonates were diagnosed after birth hospital discharge to home. If neonates were transferred to CHOP from the birth hospital for further testing, and subsequently diagnosed at CHOP, they were included in the early postnatal group.

Neonates were excluded if the sole operative procedure was: pacemaker implantation, vascular ring division, closure of patent ductus arteriosus, arterial cutdown, or ECMO cannulation, provided the infant undergoing these procedures did not have major structural congenital heart defects. Subjects born at home were excluded (Figure 1).

Figure 1.

Flow Diagram of Patient Identification, Inclusion and Exclusion.

Patient characteristics, such as age at surgery and weight at surgery were compared using Chi-square analysis. Chi-square analysis was also used to compare diagnosis timing, STAT category, in-hospital mortality, and 1-year mortality. In-hospital mortality and 1-year mortality were compared between the cohorts using ordinal multivariable regression modeling after adjusting for weight, age, and STAT category.

Results

A total of 527 neonates were included, of which 251 comprised the 2009–2010 cohort (47.6%). Characteristics of the two cohorts are shown in table 1. The 2009–2010 cohort was younger at age of surgery (mean 6 days versus 7 days, p=0.04). There was no difference in the weight at surgery. Most subjects were diagnosed prenatally (76.9% in 2009–2010 vs. 77.9% in 2015–2016, p= 0.78). Of the postnatally diagnosed infants, 19.5% of the 2009–2010 cohort were diagnosed early postnatally vs. 17.8% in the 2015–2016 cohort (p=0.60). The late diagnosis rate in both cohorts was low (3.6% versus 4.3%, p=0.66). No neonates were identified by pulse oximetry screen in the 2009–2010 cohort. All postnatally diagnosed infants in the 2015–2016 cohort were born in states that mandated pulse oximetry screening. When using STAT mortality categories to assess surgical complexity between the two groups, cohort 1 had a greater proportion subjects undergoing STAT category 1 surgery (5.2% vs 1.5%, p=0.02) and fewer STAT 4 operations (34.7% vs 49.3%, p= 0.001) compared to cohort 2 (Table 2). There were no other differences in surgical complexity between the cohorts.

Table 1.

Baseline Characteristics of the Two Cohorts.

Characteristic	Cohort 1	Cohort 2	P-value
n	251	276
Mean Age (d)	6	7	0.041
Median age	4	5
Mean weight (kg)	3.15	3.12	0.454

Table 2. STAT Category of Each Cohort.

Each cohort has been subdivided by number of neonates in each STAT category.

Stat Score	Cohort 1 N (%)	Cohort 2 N (%)	P-value
1	13 (5.2)	4 (1.5)	0.024
2	33 (13.2)	36 (13.0)	0.930
3	48 (19.1)	41 (14.9)	0.171
4	86 (34.7)	136 (49.3)	0.001
5	68 (27.1)	59 (21.4)	0.107

Mandatory pulse oximetry screening identified a small number of subjects with critical congenital heart disease; three subjects were diagnosed with transposition of the great arteries, four with total anomalous pulmonary venous return, one with aortic coarctation and one with interrupted aortic arch. The low rate of late post-natal diagnosis from the 2009–2010 cohort suggest that without mandatory pulse oximetry screening, most subjects would have been identified by history, vital sign changes and physical exam prior to being discharged from the hospital. Within the latter cohort, one subject with truncus arteriosus and 10 subjects with aortic coarctation were discharged from birth hospital after pulse oximetry screening was negative.

In-hospital mortality was similar (Table 3). There was a significant decline in 1-year mortality, from 17.1% in the 2009–2010 cohort to 10.5% in the 2015–2016 cohort (p=0.02). This decline in 1-year mortality occurred exclusively in the a diagnosed group, from 20.2% in the 2009–2010 cohort to 11.6% in the 2015–2016 cohort (p=0.03). No late postnatally diagnosed subjects died in either cohort. We identified a 1-year mortality reduction in subjects undergoing the Norwood procedure from 32.8% to 21.1%, although it was not statistically significant (p=0.16). Ordinal multivariable regression modelling was used to adjust for diagnosis timing, STAT category, age and weight at surgery. Using this model, there was no statistically significant difference in adjusted in-hospital mortality between the two cohorts (OR 0.572, 95% CI 0.306, 1.067, p=0.08). However, there was a statistically significant decline in adjusted 1-year mortality in the later cohort (OR 0.489, 95% CI 0.286, 0.837 p=0.01). Subgroup analysis identified decreased adjusted 1-year mortality in prenatally diagnosed subjects in the later cohort (OR 0.511, 95% CI 0.284, 0.918 p=0.03).

Table 3. Timing of Diagnosis of Critical Congenital Heart Disease.

The cohorts were divided into sub-groups based on the timing of diagnosis of critical congenital heart disease. In-hospital mortality and 1-year mortality were analyzed by cohort and by subgroups within the cohorts.

	Cohort 1 (2009–2010)	Cohort 2 (2015–2016)	P value
	N (%)	N (%)
Prenatal Group	193 (76.9)	215 (77.9)	0.78
In-hospital mortality	27 (14.0)	20 (9.3)	0.13
1-year mortality	39 (20.2)	25 (11.6)	0.02
Early Postnatal Group	49 (19.5)	49 (17.8)	0.60
In-hospital mortality	2 (4.1)	2 (4.1)	>0.99
1-year mortality	4 (8.2)	4 (8.2)	>0.99
Late Postnatal Group	9 (3.6)	12 (4.3)	0.66
In-hospital mortality	0 (0)	0 (0)	>0.99
1-year mortality	0 (0)	0 (0)	>0.99
Total	251 (100)	276 (100)
In-hospital mortality	29 (11.6)	22 (8.0)	0.16
1-year mortality	43 (17.1)	29 (10.5)	0.03

Discussion

While mandatory pulse oximetry screening identified multiple cases of critical congenital heart disease, the rate of late postnatal CCHD diagnosis did not change over the study period. As no late postnatally diagnosed subjects died in either cohort, we were unable to evaluate the effect of pulse oximetry screening on mortality; this may be limited by our low number of late postnatally diagnosed subjects in either cohort. Subjects with more complex CCHD were more likely to be identified prenatally (Figure 2). The mortality decrease in the 2015–2016 cohort is unrelated to mandatory pulse oximetry screening in our database.

Figure 2. Analysis of timing of Diagnosis by Disease Severity.

For each STAT category, the percentage of neonates diagnosed prenatally, early postnatally, and late postnatally are shown.

Ostensibly, mandatory pulse oximetry screening seeks to reduce the late CCHD diagnosis rate and thereby limit to sequelae of late diagnosis, including cardiogenic shock and mortality [1]. Prior to mandatory pulse oximetry screening, the late CCHD diagnosis rate at our center was already quite low at 3.6% and did not significantly change over the study period. If mandatory pulse oximetry screening were identifying infants who were previously dying prior to surgical correction, and hence were being missed in the 2009–2010 cohort, our study would have expected to find an increase in the number of early postnatal diagnosis patients in the 2015–2016 cohort. In reality, our study saw no change in the number of early postnatal diagnosis patients between the two cohorts. Given the limited target population size and that no late diagnosis mortality was identified in the earlier cohort, we identified no potential mortality benefit from mandatory pulse oximetry screening in our database.

Our data stand in contrast to Abouk et al. who described that mandatory pulse oximetry screening was associated with a 33.4% decline in CCHD mortality using registry birth/death linkage data [6]. This discrepancy in mortality benefit after mandatory pulse oximetry screening may be explained by several differences in study design. The Abouk study used administrative data, which are inferior to clinical registry data for CCHD case ascertainment when evaluating congenital heart disease outcomes [9]. The administrative registry Abouk used cannot differentiate between method and timing of CCHD diagnosis. A strong methodologic difference in difference approach was employed to try to account for these limitations. However, the crux of the analysis relies on the interpretation of the deaths in states that mandated pulse oximetry screening before and after implementation. In the Abouk study, the total reported annual death rate due to CCHD at >24 hours and <6 months of age over the 8 states of interest ranged from 26–41 deaths per year, without insight into why the subjects died [6]. By contrast, our study offers an in-depth evaluation of 43 deaths due to CCHD that happened over two years in the pre-pulse oximetry screening era, and found that none of these subjects were likely to have benefited from pulse oximetry screening because 41 of them were diagnosed prenatally and the other two were diagnosed prior to discharge from their birth hospital. If the practice change that led to the decreased mortality in our study was unevenly implemented across the country, it may have biased the results of Abouk’s study, causing an overestimation of pulse oximetry’s effects on mortality from CCHD. Other studies evaluating pulse oximetry screening have found a limited impact from pulse oximetry screening. Diller and colleagues identified that mandatory pulse oximetry screening has 14.3% sensitivity for CCHD detection [7]. Johnson et al found that of the 112 infants born with critical congenital heart disease in that nursery, 111 were diagnosed prenatally, and the one patient who was diagnosed postnatally was missed by pulse oximetry screening [10]. In an analysis of a decade long congenital anomalies survey of 138,176 live births in the northern United Kingdom, pulse oximetry did not statistically significantly impact mortality [8]. In the face of these negative studies, it is worth investigating alternate causes for the decline in mortality due to CCHD reported by Abouk et al.

We identified a significant decline in 1-year mortality among subjects who were diagnosed prenatally. As we identified no mortality benefit attributable to mandatory pulse oximetry screening in our population, the improvement in 1-year mortality is likely due to an unmeasured practice change. During the study period, many centers adopted an infant single ventricle monitoring program. Dedicated infant single ventricle monitoring is associated with improved 1-year transplant free survival after stage 1 palliation [11]. Out of these efforts, the National Pediatric Cardiology Quality Improvement Collaborative (NPC-QIC) developed [12]. NPC-QIC was established prior to mandatory pulse oximetry screening, but membership increased over the study period and now encompasses 55 congenital heart surgery centers [12]. Over the study period, NPC-QIC identified reduced interstage mortality after single ventricle palliation, with most improvement occurring in 2014 [12]. In December 2010, we established a dedicated single ventricle monitoring program at our center, contemporaneous with the identified improvement in 1-year mortality, suggesting that at least some of the identified mortality benefit may be due to the monitoring program. While we can hypothesize as to why 1-year mortality declined in our subset, we identify no relationship between mandatory pulse oximetry screening and lower mortality.

The main reason for falsely negative pulse oximetry screens was the failure of mandatory pulse oximetry to identify critical coarctation of the aorta. In the 2015–2016 cohort, there were eleven subjects not identified by pulse oximetry screening, 10 of whom had critical coarctation (91.0%); all subjects were in states mandating pulse oximetry screening. This observation is corroborated by a recent study evaluating the performance of mandatory pulse oximetry screening on 77,148 neonates. That study, by Diller et al., found that of the 6 false negative screens, 4 subjects had critical coarctation of the aorta [7]. Pulse oximetry would not detect coarctation if the ductus arteriosus has closed and or there were no right-to-left shunt; this may explain why neonates with critical coarctation were missed by the pulse oximetry screen [13,14]. The results of our study highlight this clinically important limitation of pulse oximetry, and emphasize meticulous newborn physical examination to detect critical coarctation prior to discharging newborns from the nursery.

Limitations

Our study has several important limitations. This is a retrospective single center analysis of existing data, limiting identification of unmeasured covariates. Moreover, our study has relatively low sample size and a low postnatal diagnosis rate in both cohorts, limiting assessment of existing covariates and impacting our power to detect true differences. Specifically, close to 80% of the patients were prenatally diagnosed, leaving only 20% of the sample as potential beneficiaries of pulse oximetry screening. The Philadelphia area has a high rate of prenatal detection of CCHD, which likely decreased the impact of mandatory pulse oximetry screening and may limit the extrapolation of our study nationwide [15]. With higher prenatal diagnosis rates, there are fewer patients with critical congenital heart disease identified by pulse oximetry screening.. As patients with more severe congenital heart disease have higher mortality, this further limits the potential mortality benefit of pulse oximetry screening in areas with high prenatal diagnosis rates. This limits the applicability of the study’s findings to areas of the country with lower prenatal CCHD detection rates. While this is a limitation, it also demonstrates the ability of prenatal ultrasound to detect most CCHD. Because of the case ascertainment method, infants who died of critical congenital heart disease without undergoing surgery would be missed, although this number is likely very small. By limiting the analysis to patients with critical congenital heart disease, we were not able to measure potential secondary benefits to pulse oximetry screening, such as detection of sepsis or non-critical congenital heart disease. There were small differences in STAT categories, age, and weight at surgery, but multivariate regression models accounting for these differences still demonstrated a statistically significant decline in 1-year mortality between the cohorts.

Conclusions

Mandatory pulse oximetry screening legislation did not change the late CCHD postnatal diagnosis rate in our cohort. The vast majority of infants with CCHD were diagnosed prenatally. Of those diagnosed postnatally, most were diagnosed early postnatally. There was a statistically significant decline in 1-year mortality from critical congenital heart disease from 2009–2010 to 2015–2016, but this decline is not attributable to pulse oximetry screening in our population.

Abbreviations

CCHD

Critical Congenital Heart Disease

ECMO

Extracorporeal Membrane Oxygenation

NPC-QIC

National Pediatric Cardiology Quality Improvement Collaborative

CHOP

Children’s Hospital of Philadelphia