The Cost of Uncertainty: Operational Consequences of Eliminating Routine DAT Screening

A healthy infant is born at 39 weeks and 2 days via spontaneous vaginal delivery. The birthparent’s blood type is O positive. Cord blood studies were not obtained. At 24 hours of life, the infant’s transcutaneous bilirubin (TcB) level is 8.8 mg/dL. The phototherapy threshold is 12.8 mg/dL without neurotoxicity risk factors and 10.5 mg/dL if risk factors are present (Figures 1 and 2).1,2The family requests discharge home. The nurse calls and asks, “What would you like to do?” Do you interpret TcB using the non-neurotoxicity curve and discharge with routine follow-up? Do you assume possible hemolysis and obtain a serum bilirubin because the infant’s blood type and direct antiglobulin test (DAT) are unknown and birthparent has O blood group? Do you delay discharge to send cord blood studies? If cord blood is unavailable, do you draw blood from the infant to run a DAT?In 2024, there were just over 3.6 million live births in the United States,3 with hyperbilirubinemia noted as the most frequent reason for rehospitalization in the first 2 weeks of life.4 This statistic makes the clinical scenario described above one that pediatric practitioners encounter frequently. Although “hazardous hyperbilirubinemia” has been described in the literature at markedly higher bilirubin levels,5 clinicians more often encounter cases in which well appearing infants have bilirubin levels between or close to treatment thresholds in the setting of unclear risk stratification. It is in these routine, lower-risk decision points that uncertainty emerges and practice variation becomes most apparent.When the 2022 American Academy of Pediatrics (AAP) hyperbilirubinemia guideline was published,1 it was welcomed by newborn and pediatric hospital medicine clinicians. Like all clinical practice guidelines intended to standardize care, reduce unnecessary interventions, and improve patient-centered outcomes, this guideline aimed to safely reduce excess laboratory testing and phototherapy, minimize maternal-infant separation, support breastfeeding, and improve the efficiency of care. The phototherapy thresholds were increased based on evidence that kernicterus occurs at higher bilirubin levels than previously thought. Early postimplementation data suggest a nearly 50% reduction in phototherapy initiation, with a total decline in phototherapy from 3.9% to 2.1% of all neonates and corresponding reductions in maternal-infant separation.6Our institution participated in the AAP’s Learning and Implementing Guidelines for Hyperbilirubinemia Treatment initiative. We revised order sets, redesigned workflows, retrained staff, and built a detailed clinical pathway to standardize care across units and practice groups. One major change in our newborn nursery was the elimination of routine cord blood type and DAT screening for infants born to a birthparent with blood group O. This practice, although permitted under both the 2004 and 2022 guidelines, was not implemented at our hospital until the most recent update. Blood bank projections were estimated at approximately $1 million in cost savings annually. It appeared to be a clear win: fewer tests, lower costs, and equivalent outcomes.Implementing these changes required substantial cross-departmental coordination and formal policy revisions across labor and delivery, the newborn nursery, neonatal intensive care unit, pediatric inpatient services, pediatric intensive care unit, and the blood bank. However, despite multidisciplinary buy-in and meticulous planning, we encountered familiar challenges after implementation: uncertainty in neurotoxicity risk assessment and wide variation in practice.Per the 2022 guidelines, neurotoxicity risk is defined as risk of hemolysis (most commonly immune mediated owing to the A, B, and O blood group system [ABO] or Rh incompatibility or other maternal antibodies), albumin levels less than 3 g/dL, glucose-6-phosphate dehydrogenase (G6PD) deficiency, sepsis, or recent clinical instability. Sepsis and instability can often be identified clinically, but determining hemolysis, G6PD deficiency, or hypoalbuminemia requires laboratory testing. Although the 2022 hyperbilirubinemia guideline explicitly recommends performing a DAT on babies born to birthparents whose antibody findings are positive or unknown, guidance on checking a DAT on babies born to birthparents with O blood group continues to be optional if “bilirubin surveillance and risk assessment follows this clinical practice guideline and appropriate follow-up after discharge is arranged.”1Guidance regarding albumin is even less definitive. Although hypoalbuminemia (<3 g/dL) is listed as a neurotoxicity risk factor, albumin measurement is recommended only when bilirubin approaches the escalation-of-care threshold (within 2 mg/dL of the exchange transfusion level). This likely reflects the perception that hypoalbuminemia is rare in well newborns. However, one study that looked at neonates at 35 weeks gestation or older noted that 26.7% of infants had a serum albumin level less than 3.0 g/dL.7 By comparison, ABO incompatibility occurs in approximately 11% of pregnancies, with only 2.5% of all newborns demonstrating positive DAT results owing to anti-A or anti-B antibodies.8 By deferring albumin measurement until late in the clinical course, clinicians may fail to identify a common but underrecognized neurotoxicity risk factor.In practice, this left us confronting a fundamental dilemma: how should clinicians interpret risk when key elements of risk stratification are unknown? Without available data (unknown cord blood studies in a birthparent with O blood group or unknown albumin level in the neonate), the approach to guideline implementation came to a critical juncture—should the clinician assume the “worst-case scenario” (a sick bias assuming a bilirubin neurotoxicity risk factor) or the “best-case scenario” (a wellness bias assuming that no neurotoxicity risk factor is present)? Furthermore, if one adopts a “sick bias” approach and presumes hemolysis in the absence of a DAT result, should we not also presume hypoalbuminemia? To our knowledge, few pediatricians routinely assume low albumin levels in otherwise well newborns. Applying a worst-case assumption for ABO incompatibility but not for albumin levels is neither intuitive nor universally accepted. If this principle is self-evident, one might expect it to be explicitly recommended in the 2022 guidelines.In the absence of a DAT result for babies born to birthparents with O blood group, decisions about whether to run the DAT and adjust treatment accordingly were left largely to the attending practitioner. The impact was significant variation in practice, even within the same practicing groups. Some pediatricians chose to send cord blood studies for any baby born to a birthparent with O blood group. Others sent them if the TcB result was between thresholds for treatment (or was predicted to end up between thresholds based on trending previous results). Still, others only sent cord blood studies if the TcB level showed an alarming rate of rise or if the baby reached phototherapy threshold on the non-neurotoxicity risk curve. Some practitioners applied the non-neurotoxicity curve for all babies at discharge unless a positive DAT result was known.In addition to variation in practice, we discovered many unintended downstream consequences. Our hospital laboratory, after performing validation testing, determined that cord blood specimens became unreliable for DAT interpretation after 48 hours. Beyond that window, a new blood sample drawn directly from the infant was required, which might have been avoided had cord blood testing been performed at birth, and this often led to understandable frustration for families. In addition, the blood volume required was not insignificant, and turnaround times were often prolonged. At best, this delayed discharge. At worst, it delayed initiation of phototherapy for infants whose bilirubin levels were rising.These challenges were magnified in the outpatient setting. If an infant born to a birthparent with blood group O presented after discharge with a concerning bilirubin level and no prior DAT result, obtaining timely testing proved difficult. Unlike some institutions with streamlined outpatient blood bank services, our laboratory requires witnessed specimen collection for any blood bank studies, which meant blood sample collections by home health teams (one of our most used services for newborns after discharge) were rejected. Alternative community laboratories could not provide stat DAT testing, and reported turnaround times were highly variable (in some instances taking nearly 48 hours to communicate results to the outpatient physician). Even after results became available, additional time was needed to contact families and initiate treatment, further contributing to an already protracted process. In effect, the absence of readily available results complicated implementation of the very guidelines intended to standardize and optimize care.Ultimately, the operational burdens of additional blood draws, delayed decision-making, outpatient inefficiencies, and persistent clinical ambiguity offset the cost savings from eliminating routine cord blood testing. This realization led us to reinstate routine DAT testing for infants born a birthparent with blood group O, a move that encountered minimal resistance from our blood bank and was fully supported by all our units caring for neonates. More importantly, our experience reinforced a broader lesson: in theory, clinical guidelines may streamline decision-making but their success depends on the accessibility of the data required to apply them. As a group, we have not yet standardized an approach to albumin measurement and await further guidance from the AAP. However, the DAT, despite its cost, proved to be an essential anchor for consistent, timely, and equitable hyperbilirubinemia management in our setting. Our experience taught us that rather than defaulting to a “worst-case” or “best-case” interpretive bias—approaching care with pessimism or optimism—clinical decision-making is best served when uncertainty is minimized and management reflects both available data and real-world logistic constraints.