The Effect of Readmission for Phototherapy on Early Breast Milk Feeding

This study was a continuation of the Late Adverse Impact of Getting Hyperbilirubinemia or Phototherapy (LIGHT) study,^4,5  approved by the UCSF (#10-04918) and Kaiser Permanente (#1270417-19) institutional review boards.

We designed a retrospective cohort study with electronic health record data using a target trial emulation framework^3,6,7  to guide the eligibility criteria and study design. The initial cohort was newborns born at ≥35 weeks’ gestation at 16 KPNC hospitals from 2013 through 2017 (N = 191 004; Fig 1). We excluded 122 infants for missing data on length of birth hospitalization stay.

We categorized infants into neurotoxicity risk groups on the basis of gestational age and direct antiglobulin test results, as previously described,^3,8  and compared their TSB levels with the 2004 American Academy of Pediatrics (AAP) phototherapy threshold⁹  for their risk group and postnatal age. Infants were eligible for inclusion if they had a qualifying outpatient TSB level ranging from 1 mg/dL below to 2.9 mg/dL above the AAP threshold with no previously higher TSB at <15 days of age. Within this TSB range, readmission for phototherapy would be reasonable but not required.

Data from electronic medical records contained information on phototherapy, breast milk feeding, and relevant covariates (see Covariates, below). The primary analysis included only those with complete covariate and outcome data. Two-month feeding data were available on 7855 infants (84.3%). Complete data were available on 7729 infants (82.9%).

The exposure of interest was inpatient phototherapy during readmission at <15 days of age. We identified inpatient phototherapy from the presence of a phototherapy nursing flowsheet or both an International Classification of Diseases, Ninth/Tenth Revision procedure code and an order for phototherapy.

We controlled for each infant’s first outpatient TSB in the qualifying range using the difference from the AAP threshold in categories of 1.0 mg/dL.^5,8  We also controlled for age at the first qualifying TSB level. We selected other confounding variables on the basis of a directed acyclic graph and availability in the electronic medical record. These included: sex, gestational age, delivery mode (spontaneous vaginal, cesarean, or assisted vaginal), maternal age (<25, 25–34, or ≥35 years), average number of formula feedings per day during the birth hospitalization, hospital and year of birth, and birth hospitalization phototherapy. Self-reported race or ethnicity was also included because known racial and ethnic differences in both phototherapy treatment¹⁰  and breastfeeding exist.¹¹ 

Our outcomes were exclusive breast milk feeding and any breast milk feeding, as reported by caregivers at the 2-month well-child visit (completed between 45 and 89 days of age). Questions asked about breast milk feeding rather than breastfeeding, so some mothers who reported breast milk feeding may have been using pumped breast milk rather than breastfeeding.

We estimated adjusted risk ratios (RR) using Poisson models including the above covariates for the outcomes of (1) exclusive breast milk feeding and (2) any breast milk feeding. Analyses were conducted in Stata 15.1 (College Station, TX). To quantify absolute risk differences, we used Stata’s margins command to estimate the average treatment effect on the treated (ATET). The ATET is an estimate of the difference between the observed breast milk feeding rates among infants who were readmitted for phototherapy and what their breast milk feeding rates would have been had they not been readmitted for phototherapy. Among infants who were readmitted for phototherapy, we also estimated adjusted risk ratios using Poisson models (including the above covariates) for the exposure of the average number of formula feedings per day during readmission with the same outcomes.

We assessed missing data by comparing participants with and without complete 2-month feeding data because few infants were missing only covariate data. We re-estimated the primary models, imputing missing covariate and outcome data using multiple imputation by chained equations with 200 imputations. We included all exposure, covariate, and outcome variables in the imputation models, as well as breast milk feeding data from the 4- and 6-month well-child visits. Of those missing 2-month feeding data, 50.5% had either 4- or 6-month feeding available. We used the mimrgns package¹²  to calculate the ATET.

Approximately one-quarter (26.5%) of the 7729 infants in the cohort were readmitted for phototherapy. Those who had received birth hospitalization phototherapy and those who had higher TSB levels relative to the threshold were more likely to be readmitted (Table 1). Infants readmitted for phototherapy were more likely to be of non-White race or ethnicity, delivered vaginally, younger at qualifying TSB level, fed less formula during their birth hospitalization, and born in earlier years. They were less likely to have received home phototherapy.

Infants readmitted for phototherapy were less likely to be fed breast milk only at the 2-month visit (42.9%) than those who were not (48.5%; Table 2). Readmission for phototherapy remained associated with a slightly lower likelihood of exclusive breast milk feeding in adjusted models (adjusted RR, 0.90; 95% confidence interval [CI], 0.84 to 0.96). Among infants readmitted for phototherapy, the proportion being fed breast milk exclusively at the 2-month visit was estimated to be 5 percentage points lower than it would have been if they had not been readmitted for phototherapy (ATET, −5.0%; 95% CI, −7.9% to −2.1%). Any breast milk feeding was similar among those readmitted for phototherapy (81.2%) and those not (82.2%). Readmission for phototherapy remained independent of any breast milk feeding at the 2-month visit in adjusted models (adjusted RR, 1.00; 95% CI, 0.97 to 1.02; ATET, −0.4%; 95% CI, −2.6% to 1.9%).

Among readmitted infants, an increasing number of formula feedings per day was associated with a dose-response decrease in both exclusive breast milk feeding and any breast milk feeding at 2 months. Adjusted RRs (compared with no formula) ranged from 0.91 (95% CI, 0.84 to 0.99) for 2 or fewer formula feedings per day for any breast milk feeding at 2 months to 0.36 (95% CI, 0.29 to 0.43) for ≥6 formula feedings per day for exclusive breastfeeding at 2 months.

Characteristics of patients with complete 2-month feeding data were similar to those of patients missing 2-month feeding data, although the distributions of race or ethnicity, maternal age, and year of birth differed slightly between groups (Supplemental Table 3). Results from multiple imputation analyses including all eligible patients (Supplemental Table 4) were similar to those from primary analyses.

We found that readmission for phototherapy was not associated with a decrease in any breast milk feeding but was associated with a small decrease in exclusive breast milk feeding.

This contrasts with the results of our previous study, in which birth hospitalization phototherapy was not associated with a reduction in exclusive breast milk feeding but was associated with a small increase in any breast milk feeding. One possible explanation for this difference is that readmission phototherapy generally takes place on the pediatric ward in this hospital system, where lactation support may be less available than in the mother–baby unit or newborn nursery.

It is possible there is residual confounding by unmeasured factors (eg, lack of outpatient lactation support or increased infant weight loss) that would make an infant more likely to be readmitted for phototherapy and less likely to be exclusively breastfed. These might affect exclusive breast milk feeding more than any breast milk feeding, which would be consistent with our results. Additionally, feeding problems are more likely to contribute to jaundice as infants age. Mothers of infants readmitted for phototherapy may have had lower breast milk supply (eg, due to obesity),¹³  which could contribute to both readmission for phototherapy and decreased breast milk feeding at 2 months.

It should be noted that, although we found a slight decrement in exclusive breastfeeding for readmission phototherapy and not for birth hospitalization phototherapy, this does not imply that providing subthreshold phototherapy¹⁴  is desirable to prevent readmission.

Because KPNC has a strong commitment to breastfeeding throughout its system¹⁵  these results may not be generalizable to health care settings with less lactation support. At KPNC there is usually an opportunity for mothers to stay with their infants on the pediatrics floor during readmission for phototherapy. In settings where infants are separated from their mothers, phototherapy may have a more negative effect on breast milk feeding than what we found in our cohort.

Historically, formula has been used to prevent and treat breast milk jaundice. The medical team may influence formula use for jaundiced infants, especially during inpatient admissions. With increasing average daily formula use during readmission, we found a dose-response decrease in both exclusive and any breast milk feeding at 2 months, suggesting a possible adverse effect of such use.

Our findings suggest that there may be an opportunity to provide increased lactation support to mothers of infants readmitted for phototherapy, but that in a system that supports lactation, any deleterious effect of readmission for phototherapy on breast milk feeding is likely small.

The authors thank Eileen Walsh, RN, who assisted with data extraction at the Kaiser Permanente Northern California Division of Research, and David Glidden, PhD, who served as a statistical advisor in the University of California, San Francisco Department of Epidemiology & Biostatistics.