Fellow: Lea Bornstein and Amelia Castro
Article: Chang, Pearl and Newman, Thomas. A Simpler Prediction Rule for Rebound Hyperbilirubinemia. Pediatrics, 2019 (144) e20183712; DOI: https://doi.org/10.1542/peds.2018-3712
Summary:
Authors used a retrospective cohort of over 7,000 >ex 35 week neonates from the Kaiser Permanente Northern California Database who required phototherapy in the first two weeks of life to analyze which babies developed rebound hyperbilirubinemia within 72 hours of phototherapy termination. They initially developed and internally validated a 3-variable formula to determine rebound risk. In this study, they further simplified the rule to 2 variables, a) gestational age and b) the difference between the phototherapy threshold at initiation and the bilirubin level at termination. Both the initial 3-variable model and the newer 2-variable model were able to “discriminate events,” that is, the models were able to separate which babies would or would not have rebound hyperbilirubinemia. “Discrimination” was measured by an Area Under Receiver Operating Curve Characteristic, which effectively combines sensitivity and specificity (3 variable: 0.887, CI 0.864-0.910, 2 variable: 0.877, CI 0.856-0.899, max of 1). The model was also “calibrated” well; that is, the model could predict the probability of rebound hyperbilirubinemia close to the actual observed frequency in the validation dataset.
Key Strengths:
-Introduces evidence-based guidance for when to discontinue phototherapy, an end-point that previously had very little guidance
-Cohort is large and diverse
-Risk of rebound independent of access to follow up. This is important since access to follow is variable across the country
-Simplicity: utilizes 2 variables (do not need to adjust for race, type of feeding, DAT status, birth weight
Limitations:
-Cohort is derived from a large integrated care database and babies had access to home phototherapy (though percentage of babies who were prescribed this is low)
-Babies of Asian descent over represented in this cohort compared to general US population
-Impact of G6PD not evaluated
Main Takeaway:
In this article, the authors used a large database to develop a model to predict likelihood of rebound hyperbilirubinemia with good sensitivity, specificity, and model fit based on a combination of two factors: a) gestational age and b) difference between bilirubin level at phototherapy termination to the threshold cutoff at phototherapy initiation.
The model can be used to calculate the predicted probability of rebound bilirubinemia for a given infant. Additionally, they reported the following guidelines:
For an “acceptable” rebound risk of 2.5%, for babies >/= 38 weeks of gestation, phototherapy can be terminated at 2 mg/dL below the initiation threshold.
For the same rebound risk of 2.5%, for babies <38 weeks gestation, phototherapy can be terminated at 5.5 mg/dL below the initiation threshold.
Impact on Practice:
Utilizing these guidelines could significantly decrease the length of stay for many infants undergoing phototherapy. Not only could this potentially represent decreased cost to the family and more efficient utilization of resources for the hospital, but also decrease the disruption to establishing breastfeeding and negative impacts on parental bonding. To illustrate: this paper notes that there were 3023 infants who had >/ 2 TSB measure after the start of phototherapy. 1025 (34% of them) had a predicted rebound risk of <4% at the penultimate TSB measurement. These babies were continued on inpatient phototherapy for another average 23 +/- 9 hours. Having the guidance to be able to essentially eliminate what amounts to an extra hospital day for a significant number of babies represents the potential for a large impact.
Article: Chang, Pearl and Newman, Thomas. A Simpler Prediction Rule for Rebound Hyperbilirubinemia. Pediatrics, 2019 (144) e20183712; DOI: https://doi.org/10.1542/peds.2018-3712
Summary:
Authors used a retrospective cohort of over 7,000 >ex 35 week neonates from the Kaiser Permanente Northern California Database who required phototherapy in the first two weeks of life to analyze which babies developed rebound hyperbilirubinemia within 72 hours of phototherapy termination. They initially developed and internally validated a 3-variable formula to determine rebound risk. In this study, they further simplified the rule to 2 variables, a) gestational age and b) the difference between the phototherapy threshold at initiation and the bilirubin level at termination. Both the initial 3-variable model and the newer 2-variable model were able to “discriminate events,” that is, the models were able to separate which babies would or would not have rebound hyperbilirubinemia. “Discrimination” was measured by an Area Under Receiver Operating Curve Characteristic, which effectively combines sensitivity and specificity (3 variable: 0.887, CI 0.864-0.910, 2 variable: 0.877, CI 0.856-0.899, max of 1). The model was also “calibrated” well; that is, the model could predict the probability of rebound hyperbilirubinemia close to the actual observed frequency in the validation dataset.
Key Strengths:
-Introduces evidence-based guidance for when to discontinue phototherapy, an end-point that previously had very little guidance
-Cohort is large and diverse
-Risk of rebound independent of access to follow up. This is important since access to follow is variable across the country
-Simplicity: utilizes 2 variables (do not need to adjust for race, type of feeding, DAT status, birth weight
Limitations:
-Cohort is derived from a large integrated care database and babies had access to home phototherapy (though percentage of babies who were prescribed this is low)
-Babies of Asian descent over represented in this cohort compared to general US population
-Impact of G6PD not evaluated
Main Takeaway:
In this article, the authors used a large database to develop a model to predict likelihood of rebound hyperbilirubinemia with good sensitivity, specificity, and model fit based on a combination of two factors: a) gestational age and b) difference between bilirubin level at phototherapy termination to the threshold cutoff at phototherapy initiation.
The model can be used to calculate the predicted probability of rebound bilirubinemia for a given infant. Additionally, they reported the following guidelines:
For an “acceptable” rebound risk of 2.5%, for babies >/= 38 weeks of gestation, phototherapy can be terminated at 2 mg/dL below the initiation threshold.
For the same rebound risk of 2.5%, for babies <38 weeks gestation, phototherapy can be terminated at 5.5 mg/dL below the initiation threshold.
Impact on Practice:
Utilizing these guidelines could significantly decrease the length of stay for many infants undergoing phototherapy. Not only could this potentially represent decreased cost to the family and more efficient utilization of resources for the hospital, but also decrease the disruption to establishing breastfeeding and negative impacts on parental bonding. To illustrate: this paper notes that there were 3023 infants who had >/ 2 TSB measure after the start of phototherapy. 1025 (34% of them) had a predicted rebound risk of <4% at the penultimate TSB measurement. These babies were continued on inpatient phototherapy for another average 23 +/- 9 hours. Having the guidance to be able to essentially eliminate what amounts to an extra hospital day for a significant number of babies represents the potential for a large impact.