Beyond “Asian”: Specific East and Southeast Asian Races or Ethnicities Associated With Jaundice Readmission

Methods

We performed a case-control study with 653 infants who were born into the XXX Children’s Hospital well-baby nursery during from 2014 to 2016. The case patients were 105 infants who were born into the well-baby nursery and later readmitted for hyperbilirubinemia within 30 days of discharge. To be considered a case patient, it was required that an infant had a primary diagnosis of hyperbilirubinemia and received phototherapy on readmission. Infants readmitted for direct hyperbilirubinemia were not included in the analysis. Each case patient was matched with 5 control patients born during the same week as a case patient infant. Because there were more than 5 potential control patients for each case, the control patients were first selected on the basis of a random number generator. The control patients were 548 infants who were admitted to the well-baby nursery and not readmitted for hyperbilirubinemia within 30 days of discharge. For 4 instances, case patients were made control patients when during data abstraction, it was determined that the patient did not have a primary diagnosis of hyperbilirubinemia and did not receive phototherapy on readmission.

We abstracted the following variables from the electronic health record of each infant: history of readmission for hyperbilirubinemia, Coombs positivity, gestational age, sex, maternal parity, feeding method, and race or ethnicity. Variables were abstracted twice, by separate, independent abstractors to confirm the data. Discrepancies were checked by a third party and resolved by consensus and re-abstracting the necessary data.

Readmission for indirect hyperbilirubinemia was the dependent variable. Coombs positivity, gestational age, sex, maternal parity, feeding method, and race or ethnicity were the independent variables. Coombs status was routinely determined during the initial admission for infants that had an O blood type or were Rh negative. Gestational age, sex, maternal parity, and feeding method were determined on the basis of documentation during the initial admission note.

Parent-reported race or ethnicity was abstracted from electronic medical records of the California Newborn Screening Test, which is performed by state law for each infant born in California at least 24 hours after birth. The screening test form has a semistructured question for the parent(s) to describe the races and/or ethnicities of their infant.⁶ The form states, “Race/Ethnicity: Fill all that apply.” The responses are “White,” “Hispanic,” “Black,” “Chinese,” “Japanese,” “Korean,” “Vietnamese,” “Filipino,” “Cambodian,” “Laotian (Laos),” “Other S.E. Asian,” “Middle Eastern,” “Hawaiian,” “Samoan,” “Asian-East Indian,” “Guamanian,” “Native American,” and “Other (Specify).” If the parent(s) select Other (Specify), then there is an open-ended space that includes up to 11 characters to describe the race or ethnicity.

Because there were 45 unique combinations of parent responses to this question, for our analyses, we consolidated the specific races and ethnicities into the following groups, as determined by the National Institutes of Health definitions of racial and ethnic categories⁷: Black, Far East Asian, Hispanic, Native American, Pacific Islander, Southeast Asian, South Asian, White, and Unknown. In Fig 1, we outline the consolidation process and the grouping of each of the 45 individual categories. The following self-reported races or ethnicities were placed into the Unknown group: Asian, “Multiracial,” and those responses that were left unanswered by the parent(s). We included Asian in this group because this parent description is ambiguous and spans multiple categories such as Far East Asian, South Asian, and Southeast Asian. Infants identified as multiple races or ethnicities (up to 2) were included in both categories.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Consolidation of parent-reported races and ethnicities. NA, Not applicable

We performed bivariate analyses (logistic regression) to establish which variables are associated with newborn hospital readmissions for hyperbilirubinemia. We then performed multivariable logistic regression with all variables that were significant in the bivariate analyses to determine factors associated with newborn hospital readmissions for hyperbilirubinemia.

With a total of over 100 readmission cases, we had sufficient power to analyze ∼5 to 10 independent variables for our final multivariable logistic regression.⁸ We used Stata (Version 14.0; Stata Corp, College Station, TX) to perform all analyses. The study was approved by the XXX Committee on Human Research (Institutional Review Board).