Impact of the 2022 AAP Guidelines on Neonatal Hyperbilirubinemia Admissions: A PHIS Study Free

Neonatal jaundice is the most common reason for hospitalization in the first 2 weeks of life.¹ Hospitalization for phototherapy contributes to parental stress, as the initiation of phototherapy can separate the mother from the infant, disrupt breastfeeding, and create concern for illness in parents.^2,3 In addition, there are potential short-term negative side effects of phototherapy, including circadian rhythm disturbance, as well as potential long-term adverse effects, including small potential associations with seizures, leukemia, and allergic conditions.^4–7 Bilirubin encephalopathy has decreased since universal predischarge screening for newborns was recommended by the American Academy of Pediatrics (AAP) in 2004.^8–10 Rates of kernicterus are extremely low in term and late preterm infants without hemolysis, estimated at 1.9/100 000.^8–10 Acute bilirubin encephalopathy is the neurologic manifestation of bilirubin toxicity, which, if untreated, can lead to chronic bilirubin encephalopathy, characterized by cerebral palsy with movement disorders, hearing loss, and gaze palsies.⁹ Although kernicterus is a pathologic diagnosis in which bilirubin is deposited in the basal ganglia and other regions of the brain, the term kernicterus is often used to represent the clinical diagnosis of acute bilirubin encephalopathy.¹¹ 

The 2022 AAP Clinical Practice Guideline Revision: Management of Hyperbilirubinemia in the Newborn Infant 35 or More Weeks of Gestation recommended increased thresholds for newborn phototherapy to decrease unnecessary exposure.¹² Following guideline publication, Sarathy et al observed a reduction of phototherapy by 46% during birth hospitalization in their network of 8 hospitals.¹³ Single institutional and multicenter quality improvement efforts are further striving to decrease the use of phototherapy, decrease hospitalization for phototherapy, and decrease bilirubin measurements.¹³ To date, there has not been a study to determine whether hospitalizations for phototherapy have decreased nationally and to evaluate for potential changes in post-guideline length of stay (LOS) and kernicterus diagnosis. Our objective was to determine changes in hospitalization of infants with newborn jaundice at the national level in the year following the publication of the guidelines.

We used the Pediatric Health Information System (PHIS) database to identify a cohort of infants aged 2 to 14 days who were hospitalized after discharge from their initial birth hospitalization between August 1, 2021, and August 31, 2023. This timeframe was chosen to include the 12 months before and after the publication of Clinical Practice Guideline Revision: Management of Hyperbilirubinemia in the Newborn Infant 35 or More Weeks of Gestation on August 5, 2022. We excluded hospitalized infants with birth gestation of fewer than 35 weeks because the 2022 guidelines do not apply to this population. We also excluded infants who were transferred from another facility to minimize bias, as the diagnoses and procedures performed prior to transfer could not be verified. We did not include data from birth hospitalizations since most occur outside participating children’s hospitals. Figure 1 presents the inclusion and exclusion criteria for this study.

The PHIS contains inpatient, emergency department, ambulatory surgery, and observation encounter data from pediatric tertiary-care hospitals in the United States that are affiliated with the Children’s Hospital Association. Participating hospitals provide demographic data, diagnoses, and data on procedures, as well as resource utilization data, such as medications, imaging, and laboratory tests performed during the encounter. Data from 49 hospitals were submitted to the PHIS. We included 42 hospitals with complete data during the study period. This study was considered nonhuman subject research and was therefore exempted from review by the children’s hospital institutional review board.

Discharge diagnosis codes related to newborn jaundice from the International Classification of Diseases, Tenth Revision (ICD-10) were used to identify infants (Supplemental Table 1). Clinical and demographic characteristics of interest included age (in days) at admission; gestational age (in weeks); sex; race; ethnicity; Child Opportunity Index (COI) category; use of intensive care services during hospitalization; severity of illness; and indication of jaundice, phototherapy, or other primary diagnoses, indicating specific additional illnesses that may contribute to hyperbilirubinemia or prolonged hospitalization (Table 1, Supplemental Table 2). Although we acknowledge that race is a social construct and not genetically or biologically based and that there are limitations to maternally reported race and ethnicity as it appears in PHIS, we chose to include race and ethnicity in our final model due to the known associated outcome disparities in newborns.¹⁴ 

Among infants with a diagnosis of jaundice or a billing code indicating the use of phototherapy, we also identified the use of intravenous immunoglobulin (IVIG), exchange transfusion, and kernicterus. The term kernicterus (ICD-10 codes P57.8 and P57.9) was used to detect patients with acute and chronic bilirubin encephalopathy, as there are no ICD-10 codes for either acute or chronic bilirubin encephalopathy or bilirubin-induced neurologic dysfunction.

COI uses census data to create a composite of 44 indicators into a single metric that ranks neighborhoods in the United States on a scale of 1 to 100, where lower scores indicate fewer resources for children’s healthy growth and development.¹⁵ Normalized scores can be grouped into the following 5 categories based on quintiles: “very low,” “low,” “moderate,” “high,” and “very high.” The COI was applied to the PHIS at the zip code level. The severity of illness was assessed using the hospitalization resource intensity scores for kids (H-RISK), a pediatric-specific case mix index measure.¹⁶ 

The primary outcome was the probability of hospitalization following birth hospitalization for jaundice or phototherapy. ICD-10 diagnoses related to newborn jaundice were used to identify infants with jaundice (Supplemental Table 1). The PHIS billing and procedure codes were used to define the use of phototherapy during hospitalization. Secondary outcomes included median LOS and 14-day readmission for hyperbilirubinemia.

We used a modified Delphi process to identify infants with a primary diagnosis indicating additional comorbidities that may contribute to increased LOS, such as diseases associated with jaundice or conditions that may lead to gastrointestinal or hepatic disease, as well as conditions that might be associated with increased case complexity. We purposefully excluded dehydration because jaundice from suboptimal intake, so called “breastfeeding jaundice,” is a common physiologic process and may not lead to admission or to prolonged LOS.¹⁷ Two physician authors (AJ and JS) reviewed all primary diagnoses that were not jaundice and independently classified each diagnosis code as a condition indicating comorbidities. A third nonphysician author (TR) compared the classification results and moderated the discussion to arrive at consensus criteria. Supplemental Table 2 presents the final list of comorbid conditions in our cohort that likely contribute to jaundice.

Categorical data were summarized as frequencies and percentages. Continuous data were summarized using medians and IQRs for nonnormally distributed variables or mean and SE for normally distributed variables. Categorical demographics and clinical characteristics pre- and post-guideline were compared using the χ² test for association. We compared nonnormal continuous variables pre- vs post-guideline using the Wilcoxon rank-sum test and normally distributed continuous variables using the Student’s t test. For LOS and 14-day readmission among infants with jaundice, we performed a further subanalysis by excluding children whose primary diagnosis was not jaundice, indicating additional illnesses likely to cause jaundice (Supplemental Table 2).

We modeled the adjusted probability of hospitalization for jaundice or phototherapy pre- vs post- guideline using a generalized linear mixed model (GLMM) assuming a binomial distribution and a logit link. We modeled LOS among infants with jaundice or phototherapy using a log-linear GLMM. The covariates in both models included gestational age, sex, COI category, and intensive care unit (ICU) utilization. We included a random hospital effect to account for clustering of infants within the same hospital. Modeled results are presented as adjusted probabilities and adjusted odds ratios (aORs) with 95% CIs for jaundice/phototherapy hospitalization and adjusted ratio of means with 95% CI for LOS.

To confirm any changes in the probability of hospitalization for jaundice occurring at or near the guideline release, we assessed changes in both the unadjusted and adjusted probabilities of hospitalization over 24 months using an interrupted time series (ITS) approach. We performed this for both infants with any indication of jaundice and infants without a primary diagnosis code indicating illness that may contribute to increased LOS (Supplemental Table 2). All ITS models include parameters for a pre-guideline slope to account for any change during the pre-guideline period, a parameter to accommodate an immediate change in the intercept at the point of interruption (ie, a step parameter), and a post-guideline slope parameter to account for any change during the post-guideline period. The adjusted ITS models included gestational age, sex, COI category, and ICU utilization as covariates.

Given the rarity of the diagnosis, we performed a post-hoc analysis of kernicterus data over 5 years, including data between 2018 and 2022, to confirm the findings in the 1-year pre- vs 1-year post-guideline analysis.

All analyses were performed using SAS, version 9.4. Statistical significance was set at P < .05.

We identified 44 755 patients from 42 hospitals between August 1, 2021, and August 31, 2023, who were 2 to 14 days of age at admission. A total of 16 112 patients were excluded because their gestational age was less than 35 weeks. After excluding transfers to the hospital, 21 194 patients met our inclusion criteria, with 11 162 and 10 032 encounters in the pre-guideline and post-guideline cohorts, respectively (Figure 1). Of these, 11 906 infants (56.2%) had a diagnosis of jaundice and/or received phototherapy. Table 1 shows the characteristics of the cohort pre- and post-guidelines. Age, race, ethnicity, and COI remained unchanged (Table 1). Among infants with jaundice or phototherapy, administration of IVIG (1.3% and 1.5%; P = .527) and exchange transfusion (0.9% and 1.0%; P = .658) were similar in both pre- and post-guideline cohorts.

There was a significant decrease in hospitalizations with jaundice or phototherapy post-guideline (58.8% pre-guideline vs 53.2% post-guideline; P < .001). Furthermore, among jaundice/phototherapy encounters, there was a decrease in the percentage of encounters with a primary diagnosis of jaundice (76.9% pre-guideline vs 70.8% post-guideline; P < .001) (Table 1) with a concomitant increase in hospitalizations with other comorbidities (Supplemental Table 2; 15.0% pre-guideline vs 19.3% post-guideline; P < .001). Infants in the post-guideline cohort also appeared to have a higher severity of illness, as indicated by a higher H-RISK score (P = .009) and higher rates of ICU utilization (P < .001). After adjusting for differences in patient characteristics pre- vs post-guideline, GLMM results demonstrated a decrease in the probability of hospitalization for jaundice or phototherapy post-guideline (aOR [95% CI]: 0.80 [0.75,0.84]; P < .001) (Table 2).

More infants were admitted with a diagnosis of kernicterus in the 12-month post-guideline (n = 8) compared with the previous year (n = 3) (P = .092). Given the small number of annual cases, we further analyzed the incidence of kernicterus by including an additional 3 years of data prior to our study period. When annual kernicterus cases for the 4 years prior to the guideline (August 2018–2022) were compared to the post-guideline year (August 2022–2023), there was no significant change in the incidence of kernicterus (P = .138) (Supplemental Table 3).

Among all infants with jaundice or those that received phototherapy, the median LOS slightly increased from 29 hours pre-guideline to 32 hours post-guideline (P < .001) (Table 1). However, after excluding infants who had other diagnoses that could contribute to increased LOS (Supplemental Table 2), the LOS was unchanged from pre-guideline. There was no difference between the pre- vs post- guideline 14-day readmission rate for all encounters with jaundice (0.7% pre-guideline and 0.7% post-guideline; P = .557).

When we analyzed the data using an ITS, adjusting for potential measured confounders, there were no significant trends either pre-guideline (P = .080) or post-guideline (P = .673). However, there was an immediate post-guideline decrease in the hospitalization rate for infants with a primary diagnosis of jaundice, from 59.6% to 52.0%, beginning in the month following guideline publication (Figure 2; P < .001). When infants with additional comorbid diagnoses (Supplemental Table 2) were excluded, there was a slight upward trend in the 12 months pre-guideline (P = .047) and a similar immediate decrease (P < .001), which continued to decrease over time during the 12 months post-guideline (P = .030).

This retrospective multicenter study found a significant reduction in the number of infants admitted for jaundice in the first 2 weeks of life, immediately following the publication of the 2022 AAP clinical practice guideline revision.¹² Jaundice is a common newborn diagnosis leading to hospitalization in the postnatal period,¹ and an aim of the 2022 guideline was to safely reduce unnecessary phototherapy and excess hospitalization. Hospital admission creates parental stress, may interfere with bonding and exclusive breastfeeding, and gives parents the impression that their infant is ill.^18,19 Fewer unnecessary hospitalizations for newborn jaundice would translate to decreased stress on families, time away from home during the critical postnatal bonding period, and reduced health care costs for families and society.

The initial decrease in admissions began 1 month after guideline publication. This suggests that these guidelines were implemented quickly in the 42 children’s hospitals we studied, which is remarkable as it often stated that it takes 17 years for widespread practice changes in medicine.²⁰ The quick adoption of this guideline was likely enhanced by both the widespread implementation of the 2004 guideline as well as efforts of local and national quality improvement projects. For example, the Learning and Implementing Guidelines for Hyperbilirubinemia Treatment (LIGHT) project, conducted by the AAP, was launched within a month of guideline publication. Online physician decision support tools (eg, Bilitool) were also updated within a month of guideline publication, further supporting these quality improvement initiatives and contributing to guideline utilization in practices that may not have initiated implementation. We suspect timely updates in these decision support tools as well as strategically timed quality improvement initiatives such as LIGHT were key in implementing swift practice change nationwide.

The infants hospitalized in our post-guideline cohort also had higher ICU utilization and higher H-RISK scores, indicating a higher severity of illness. This cohort may appear more ill because it has a smaller proportion of well newborns hospitalized with hyperbilirubinemia due to higher phototherapy thresholds. Furthermore, the higher ICU utilization may have been driven by the “escalation of care” recommendations of the 2022 guideline, which provides clear guidance about when to admit to the ICU.¹² 

The median LOS for all encounters increased by 3 hours (P < .001) post-guideline but was unchanged when infants with comorbidities were excluded. This suggests that these infants with comorbidities disproportionately contribute to the LOS. The hospitalizations of infants with comorbidities may be least likely to be reduced by the higher phototherapy thresholds. In other words, due to a higher severity of illness, these infants are more likely to require hospitalization. By design, the guidelines sought to prevent unnecessary admissions and phototherapy while ensuring that infants who need inpatient phototherapy and additional medical care still receive it. Our study suggests that these infants who are “sicker” are driving hospitalizations for jaundice in children’s hospitals post-guidelines.

There was no significant change in the number of infants receiving exchange transfusion or IVIG upon admission, which we would expect to observe if more infants were admitted for the treatment of acute bilirubin encephalopathy. The number of kernicterus cases were higher in our post-guideline period (8 cases vs 3 cases pre-guideline) but did not reach statistical significance (P = .092). Because cases increased post-guideline, we further investigated cases of kernicterus between the 4-year pre-guideline and 1-year post-guideline period. There was no significant difference (P = .138) (Supplemental Table 3). Because kernicterus is a complex and lifelong condition that is preventable, it is imperative that future studies continue to monitor kernicterus rates over time.

The strengths of our investigation include the large multicenter sample size from across the United States and the ability to track admission timing and associated diagnoses. Despite these strengths, this study has several limitations. The PHIS database does not contain data on most birth hospitalizations. Therefore, this study focused on nonbirth admissions and did not pertain to infants diagnosed during their initial birth hospitalization and those transferred to children’s hospitals for escalation of care. The data were subject to potential coding errors because the diagnosis codes were acquired from the billing codes. Another limitation is that the PHIS database includes only freestanding children’s hospitals, limiting inferences from rural areas, where admission for hyperbilirubinemia treatment is more likely to occur in community hospital settings. Furthermore, the study period was only 12 months post-guideline. Some hospitals may still have been in the process of implementing the guidelines during our study period. Longitudinal follow-up of admission rates for hyperbilirubinemia treatment and bilirubin-associated diagnoses, including kernicterus, is needed.

Using the PHIS database, we found that the rate of hospitalization for newborn jaundice declined in the 12 months following the release of the 2022 AAP hyperbilirubinemia guidelines. There was an immediate decrease in all hospitalized infants, with further evidence of a continued decline in hospitalized infants diagnosed with jaundice or those receiving phototherapy without additional comorbidities. This suggests that the guidelines were implemented quickly in the centers included in the study and resulted in fewer hospitalizations without a significant change in IVIG or exchange transfusions. Although the median LOS for all encounters increased slightly post-guideline, once infants with additional comorbidities were excluded, the average LOS remained unchanged. There was no significant increase in kernicterus post-guideline compared with either the preceding year or the 4-year period preceding guideline publication. Admission rates in infants with hyperbilirubinemia and hyperbilirubinemia-associated outcomes, including kernicterus, should be monitored longitudinally.