Combining clinical risk factors with serum bilirubin levels to predict hyperbilirubinemia in newborns
Article Outline
Newman TB, Liljestrand P, Escobar GJ. Arch Pediatr Adolesc Med 2005;159:113-9.
Context It is a challenge to predict which infants will need early follow-up for hyperbilirubinemia. The present authors previously developed a model using elements from the history and physical (exclusive breastfeeding, bruising, race, cephalhematoma, maternal age, sex, jaundice in previous sibling, and gestational age) to predict which infants would have a total serum bilirubin (TSB) of 25 mg/dL or higher.
Objectives (1) To validate the previously reported risk index for predicting TSB levels of 25 mg/dL or higher in a separate population of infants; and (2) to combine a subset of this index with TSB levels measured at less than 48 hours to predict subsequent development of TSB levels of 20 mg/dL or higher.
Design Nested case-control study using electronic and paper records (study 1). Retrospective cohort study using electronic records only (study 2).
Setting Northern California Kaiser Permanente hospitals.
Participants Subjects for both studies were newborns weighing ≥2000 g born at a gestational age ≥36 weeks. The validation study included 67 cases born between 1997-1998 who developed TSB levels of 25 mg/dL or higher at less than 30 days and 208 randomly-selected control subjects. Subjects for study 2 included 5706 newborns born between 1995-1996, who were discharged from the hospital and had a TSB level measured at less than 48 hours.
Main Outcome Measure Performance of the risk index, measured as the area under the receiver operating characteristic (ROC) curve.
Results The risk index was shown to be similar in the validation group and the derivation group (area under the ROC curve=0.83 vs 0.84, respectively). Of the 5706 newborns with TSB levels measured before 48 hours, 270 (4.7%) developed a TSB level of 20 mg/dL or higher. Of these, 254 (94%) had a TSB level at the 75th percentile or higher at less than 48 hours. Application of the risk index improved prediction over using TSB level alone, largely owing to the effect of gestational age. For example, for those infants with a TSB level at the 95th percentile or higher at less than 48 hours, the risk increased from 9% for newborns born at 40 weeks or more gestation to 42% for those born at 36 weeks.
Conclusions Clinical risk factors significantly improve prediction of subsequent hyperbilirubinemia compared with early TSB levels alone, especially in those with early TSB levels above the 75th percentile.
Comment Newman et al. statistically link the diagnostic performance of two tests: pre-discharge bilirubin (TSB) risk zone assignment and a clinical risk factor index to predict excessive hyperbilirubinemia (TSB ≥20 mg/dL and ≥25 mg/dL). Both predictive tests, as independent variables, intend to provide surrogate measures for the bilirubin load: a composite of bilirubin production plus enterohepatic circulation minus bilirubin elimination. The latter is often a “wild card” with a paucity of surrogate indices other than TSB. Validation for the discriminative ability of a combined approach would reinforce the time-honored dictums for clinical practice: relevant history, astute physical examination, and objective verification. The study design is limited by choices for outcome (dependent) and predictors (independent) variables and population sample to develop safer and validated predictive rules that achieve a sensitivity of 100% and a positivity criterion with the most optimal specificity. Some challenges to achieving that goal are listed here. First, based on perceived risks of neurotoxicity, well infants with TSB between 18 and 25 mg/dL (beyond age 72 hours) are in need of hospital-based intensive phototherapy and preparation for exchange transfusion;1 thus, an outcome variable should correspond to safer thresholds of hyperbilirubinemia (such as those recommended for initiation of home phototherapy). Second, clinical history and physical signs, as predictive factors, are limited by lack of standardized definitions, vagaries of charting, and propensity to observer error. Alternatively, Keren et al.2 report practice-based objective factors: use of vacuum assistance and oxytocin induction in lieu of the frequently unrecognized or unrecorded cephalhematoma and bruising. Observer-independent factors identified by procedures (e.g., birthweight, vacuum assistance) allow for generalized results and avoid factors with inequity, such as “sibling with jaundice” as applied to the first-born infant, or “racial identity” as applied to a multiracial infant. In employing bilirubin screening and/or clinical risk measurements, measurement errors can arise and also affect predictive performance. Third, because simple interventions or aggressive management styles are likely to influence a pre-discharge hyperbilirubinemia experience, lack of consistent implementation and enforcement of pre-discharge guidelines during the course of the study is likely to influence the natural history of jaundice, and thereby the predictive performance of these tests. Actually, these two tests have a natural dynamic relationship rather than a static one as implied by the proposed linkage. Fourth, restrictions of infants to this study sample, a retrospective study with selective pre- and post-discharge TSB measurements for infants, who were presumably tested upon clinician's recognition of jaundice, requires verification and introduces a spectrum bias: a form of selection bias. The sample also excludes a group of infants with very early onset of hyperbilirubinemia who were probably well managed at the study institutions, but who may not be identified at other birthing facilities. Finally, in such a select jaundiced population, the data may erroneously suggest a better performance, and therefore a higher area under the ROC curve. Prospective studies, regardless of jaundice recognition, would be useful to demonstrate the value of universally generated predictive rules, not merely to identify an infant who is at risk for severe hyperbilirubinemia (TSB >95th percentile), but to identify accurately and unambiguously, one who is not. Thus, infants with TSB <40th percentile, who have no (or minimal) clinical risk factors, may need a less rigorous follow-up than those who are suspected to be at any risk for severe hyperbilirubinemia. The challenge is to predict which infant needs early follow-up and ensure that a practicing clinician has no decision regret.
References
PII: S0022-3476(05)00377-X
doi:10.1016/j.jpeds.2005.04.048
© 2005 Elsevier Inc. All rights reserved.
