Background. Accreditors hold hospitals accountable for harm from serious newborn hyperbilirubinemia, yet standards for evaluating performance in prevention are lacking.
Objective. We confirmed prognostic variables for newborn hyperbilirubinemia and developed a benchmarking model for self-evaluation of hyperbilirubinemia management.
Methods. We conducted a 3-year prospective cohort study in the Henry Ford Health System (HFHS) on 5507 healthy newborns of ≥35 weeks’ gestational age. HFHS follows a rigorous protocol for hyperbilirubinemia management. Defining hyperbilirubinemia as age-specific levels of total serum bilirubin exceeding American Academy of Pediatrics criteria for considering phototherapy and severe hyperbilirubinemia as total serum bilirubin ≥20 mg/dL, we used logistic and Poisson regressions to determine predictors and estimate parameters for a benchmarking model. We compared incidence rates for severe hyperbilirubinemia from HFHS to aggregate data from 11 hospitals reported to have less rigorous management.
Results. Newborns were 52.9% black, 14.4% white, 24.3% Latino, and 2.4% Asian; 30% were exclusively and 28% partially breastfed. Regression analyses revealed associations for hyperbilirubinemia and severe hyperbilirubinemia with black mothers (negative) and exclusive or partial breastfeeding and younger gestational age (positive). Male newborns and older mothers were also associated with severe hyperbilirubinemia. For all 5 variables, we found a lower risk for severe hyperbilirubinemia at HFHS than in the comparison hospital group. To compare hospitals, we developed a benchmarking model for incidence of hyperbilirubinemia adjusting for race, feeding method, and gestational age.
Conclusions. Hospitals with access to newborns’ inpatient and postdischarge data can use our benchmarking model to compare their management of hyperbilirubinemia with a reference population that received rigorous care.
Neonatal jaundice is estimated to occur in 60% of term newborns in the first week of life,1 and ∼2% reach total serum bilirubin (TSB) levels of ≥20 mg/dL.2 The TSB normally rises over the first 3 to 5 days and then declines. Therefore, it is important that interpretation of bilirubin levels is based on the infant’s age in hours.3 In rare instances, the TSB reaches levels that can cause kernicterus, a condition characterized by bilirubin staining of neurons and neuronal necrosis involving primarily the basal ganglia of the brain and manifested in athetoid cerebral palsy, hearing loss, dental dysplasia, and paralysis of upward gaze.4
Kernicterus was rarely seen in the decades after the introduction of exchange transfusion and phototherapy for the treatment of hyperbilirubinemia, but recent reports suggest that it is occurring again even in apparently healthy newborns.5,,6 In their analysis of 61 patients reported to the kernicterus registry who required hospital readmission and/or medical intervention by 7 days of age, Johnson et al6 noted that nonclinical factors such as early discharge without prompt medical follow-up were changing the risk profiles for kernicterus.
Insufficient physician awareness of the risk of kernicterus coupled with changes in the delivery of health care have resulted in a new risk model for neonatal hyperbilirubinemia. Although prematurity and medical complications are still primary risk factors, early discharge, and breastfeeding with inadequate intake have increased the risk of hyperbilirubinemia in apparently healthy newborns.7–9 The risk of hyperbilirubinemia derives from 2 major categories: the physiologic status of the newborn at birth (eg, lower gestational age, medical complications, and genetic determinants) and health care delivery and management practices (eg, early discharge, monitoring for jaundice, breastfeeding promotion, home visits, and prompt treatment with phototherapy when indicated). The second of these categories can be influenced directly by changes in monitoring and practice management to reduce risk.
Intensive monitoring and preventive intervention, guided by awareness of the common risk factors, are important components of this process. Newman et al2 conducted a retrospective cohort study among 50 000 full-term, healthy newborns in 11 hospitals and found that birth hospital, gestational age of 36 to 38 weeks, male gender, Asian race, and older maternal age were positively associated with higher risk for hyperbilirubinemia (as defined by TSB levels ≥20 mg/dL). Their study also found that the level of screening for hyperbilirubinemia and the incidence of hyperbilirubinemia varied greatly among hospitals. Moreover, hospitals varied in their frequency of testing newborns for hyperbilirubinemia between 17% and 52%, whereas the prevalence of hyperbilirubinemia ranged from 0.9% to 3.4%.2 They also noted that these interhospital differences in hyperbilirubinemia rates might be caused by variations in practice patterns.
The impact of health care practice patterns on the risk of hyperbilirubinemia suggests a mechanism for intervention. However, currently there are no validated methods for health care organizations (HCOs) to benchmark their rates of incidence of hyperbilirubinemia against a standard, making it difficult to initiate and develop quality-improvement programs. Because patient characteristics vary from hospital to hospital, levels of hyperbilirubinemia must be adjusted for benchmarking purposes. Differences such as race, gestational age, and feeding behavior among different hospital systems do not allow for direct comparisons of incidence rates without controlling for these variables known to be related to serum bilirubin levels.
In April 2001, the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) recommended that “organizations review their current patient care processes with regard to the identification and management of hyperbilirubinemia in newborns.”10 Because hyperbilirubinemia often appears first after early discharge from the birth hospitalization, to comply with this recommendation hospitals must monitor the incidence of postdischarge hyperbilirubinemia. This is easiest for hospitals that are parts of vertically integrated HCOs and hospitals with laboratories that perform postdischarge TSBs for newborns born in their facility. HCOs would also be aided in complying with the JCAHO recommendation if they can ascertain how well they are doing in preventing hyperbilirubinemia compared with hospitals that have developed “best practices” in management of hyperbilirubinemia. The primary purpose of our study is to develop a simple benchmarking model based on estimating the impact of risk factors for hyperbilirubinemia and to propose the use of this model for comparison purposes among HCOs with varying population characteristics.
Definition of Neonatal Hyperbilirubinemia
We adopted 1 of 2 approaches for defining hyperbilirubinemia for our analyses as cited below. 1) We classified newborns as hyperbilirubinemic according to the age-specific TSB levels used to recommend consideration of phototherapy by the 1994 American Academy of Pediatrics (AAP) Clinical Practice Parameter for Management of Newborn Hyperbilirubinemia.1 We defined a newborn 25 to 48 hours old as hyperbilirubinemic if the TSB measurement was ≥12 mg/dL, at 49 to 72 hours if the TSB measurement was ≥15 mg/dL, and at 73+ hours if the TSB measurement was ≥17 mg/dL. Using these criteria, we then classified newborns as hyperbilirubinemic if their TSB met or exceeded age-specific AAP criteria. 2) Severe hyperbilirubinemia was defined as TSB ≥20 mg/dL in the first 30 days of life.
We conducted a prospective cohort study at the Henry Ford Health System (HFHS), which includes the Henry Ford Hospital (HFH), Henry Ford Medical Group, Henry Ford Home Care, Henry Ford Information Services, Henry Ford Corporate Data Stores (CDS), and the Health Alliance Plan. We studied healthy newborns born ≥35 weeks’ gestation and delivered at HFH. We excluded newborns with birth weight <2000 g, those who stayed ≥3 days in an intensive care nursery, and those with TSB ≥10 mg/dL in the first 24 hours of life.
TSB measurements were performed by using the Roche Diagnostics Corporation, Hitachi 917 Analyzer. Transcutaneous bilirubin measurements were made by using a Minolta Bilirubinometer JM 102. However, as is common with so-called point-of-care testing, the results obtained from the transcutaneous bilirubinometer were only recorded in the infant’s chart but not entered in the HFHS databases. Only laboratory-performed results were available in the CDS database at HFH. A calibration study (n = 134) at HFH indicated that for all newborns with bilirubinometer readings of ≤14 (n = 28), TSB did not exceed a value of 6 mg/dL. Because all newborns at HFH are screened with a bilirubinometer at 24 hours of age, newborns with no recorded TSB results are assumed to have had a TSB ≤6 mg/dL at 24 hours of age.
HFHS Guidelines for Management of Newborn Hyperbilirubinemia
The Division of Neonatology at HFH is 1 of 20 “high-performing front-line clinical units” recently selected for study of microsystems in health care.11 “High-performing” was defined as “best-quality, best-value clinical units.”11 The Division of Neonatology has had in place for almost 2 decades (18 years) both in-hospital and postdischarge practice guidelines for the management of hyperbilirubinemia for newborns in the normal newborn nurseries and of >35 weeks’ gestation. An important aid that captures some but not all recommendations of the guidelines is a laminated color-coded pocket card (created in 1994) that is widely distributed and used by HFHS caregivers (see Fig 1). These guidelines were developed before the publication of the AAP’s practice parameter in 19941 and differ from it in some respects. Key features can be summarized as follows.
Screening and detection of hyperbilirubinemia:
Screen all newborns in the newborn nurseries transcutaneously with a Minolta Bilirubinometer JM 102 at 24 hours of age.
If the bilirubinometer index is ≥14, obtain TSB when the blood sample is drawn at 24 hours of age for the State Newborn Screen for Inborn Errors of Metabolism.
Interpret TSB results by hour of age by using the graph shown in Fig 1, which recommends responses ranging from heightened monitoring, repeating the TSB, to instituting phototherapy or, ultimately, exchange transfusion. (Note that the HFH guideline is less aggressive for recommending phototherapy than the 1994 AAP practice parameter.)1
Work up for hemolysis those newborns that reach TSB levels where Fig 1 indicates that phototherapy must be considered. (Those found to be hemolyzing exit this guideline.)
Promote frequent breastfeeding (every 2 hours), provide lactation support, and assess adequacy of feeding.
Monitoring and discharge planning:
Follow more closely and intervene at lower levels for newborns of 36 to 37 weeks’ gestation or <2250 to 2700 g of birth weight (unless small for gestational age).
Delay discharge until risk for increasing hyperbilirubinemia and/or loss to follow-up has been assessed.
Order home care nurse follow-up within 1 to 2 days of discharge from the hospital for all newborns discharged from the newborn nurseries at ≤48 hours if delivered vaginally and ≤72 hours if delivered by caesarian section, with serum bilirubin preordered for those with TSB >8 mg/dL at 24 hours of age. Often if TSB >8 mg/dL at 24 hours of age, a repeat TSB is drawn in 6 to 8 hours to determine rate of rise.
Home care nurse follow-up:
Assess newborns visually for jaundice, and use judgment to draw blood and order TSB if it has not been ordered already but seems warranted (ie, home care nurses are trained in assessment of jaundice and authorized to order TSBs).
Evaluate the newborn’s cry, suck, and activity.
For breastfed newborns, assess adequacy of feeding and promote breastfeeding. If feeding appears inadequate, supplement with expressed breast milk or formula if necessary, never water.
Contact the on-call neonatologist or neonatal nurse practitioner by cell phone (available 24 hours/day) to discuss any management issues and ensure appropriate care.
If phototherapy is indicated according to Fig 1, admit the newborn and use at least a fiber-optic blanket and overhead lights or 2 overhead halogen lights (physician order only).
Do not interrupt breastfeeding, but if inadequate, consider supplementation with expressed breast milk or formula.
We obtained birth hospitalization data for each mother and newborn and postdischarge data for the newborn. Birth hospitalization data included maternal age and race, newborn birth weight, admission and discharge dates, and TSB dates and times with results. Postdischarge data covered a period up to 30 days of life and included TSB times and results, dates for home care, office and emergency room (ER) visits, and dates for readmissions together with any associated diagnoses. HFH compiled 3 separate data files (birth hospitalization data, TSB data, and postdischarge visit and readmission data) on a monthly basis for analysis. These data came from the HFHS CDS and corresponded to births from November 1, 1998 through August 31, 2001. To supplement the birth hospitalization data, a research assistant manually abstracted 6 variables from the nursing logs in the newborn nurseries including birth time, discharge time, method of newborn feeding at discharge as recorded in the nursery discharge log, gestational age, delivery type, and whether the newborn had a hospital stay of ≥3 days in the neonatal intensive care unit.
The research assistant also abstracted home care visit data from computer printouts covering the time period of November 1998 to March 2001. Because complete postdischarge visit data are available only during this time window, we restricted our analyses of postdischarge care (ie, home care, office, or ER visit or readmission) to this period.
We used Microsoft Access for database management and SAS and S-Plus for analyses.
Access to Care: HFHS Follow-up
Approximately 33% of newborns delivered at HFH do not have insurance coverage for postdischarge care at HFHS. Nevertheless, the HFH Division of Neonatology provides follow-up of newborns according to guidelines for the first 15 days of life irrespective of insurance coverage. We were concerned that those without coverage for postdischarge care at HFHS might get some or all their postdischarge care elsewhere. Because access to care could impact both the observation of bilirubin testing and subsequent outcomes, we stratified our analysis by selecting newborns known to have had an HFHS postdischarge office or ER visit or readmission after 15 days of age. We are more likely to obtain complete postdischarge follow-up data from birth onward for this group of newborns, because they have demonstrated an ongoing propensity to seek health care at the HFHS. We coded these newborns as the HFHS follow-up (HFHSF) group and all others as the non-HFHSF group. We used the HFHSF group born from November 1998 to August 2001 to make comparisons to another study and generate the final benchmarking model, because this allows us to minimize the likelihood of loss of information on postdischarge TSBs or on newborns with hyperbilirubinemia.
The HFH and Harvard School of Public Health Institutional Review Boards for the protection of human subjects approved this study.
We calculated descriptive statistics exploring the underlying distributions of the data and assessed associations among categorical variables by using the χ2 test.
We performed multiple logistic regression to select variables predictive of hyperbilirubinemia as defined previously: the variables we considered included maternal race, newborn’s gender, feeding type, gestational age, birth weight, maternal age, and HFHSF versus non-HFHSF group, as defined above. For feeding type, we collapsed the categories of exclusive and partial breastfeeding (versus formula feeding), because we found no significant differences in the risks of developing hyperbilirubinemia and severe hyperbilirubinemia between exclusively and partially breastfed infants. We examined the model using the Hosmer-Lemeshow goodness-of-fit test and assessed the overall predictive ability using c statistics. We used Poisson regression analysis to model the probability of TSB ≥20 mg/dL because this is a rare occurrence.
In addition, we used the data results from the study conducted by Newman et al2 as a basis for evaluating the validity of our model estimates. Using the HFHSF group only and after controlling for gestational age, newborn gender, maternal race, maternal age, and feeding method, we compared the incidence of hyperbilirubinemia at HFH to the study published by Newman et al2 in 1999 by using summary data obtained from the publication. One important risk factor missing from the Newman et al study is the proportion of newborns that were breastfed. However, in a study published in 2000 using cases and controls derived from the 1999 study, Newman et al12 reported that 65% of controls (TSB <25 mg/dL) and 90% of cases (TSB ≥25 mg/dL) were exclusively breastfed, and 11% of controls and 5% of cases were partially breastfed. Because Newman et al used the same cohort of births in both studies, we made the assumption that the exclusive and partial breastfeeding rates in the 1999 study were equivalent to those in the 2000 study. We therefore used exclusive and partial breastfeeding rates from the 2000 study12 to adjust for feeding type when we compare the HFH results to those of Newman et al. That is, we assumed that, in the 1999 Newman et al study population, the exclusive breastfeeding rate was 65% (partial breastfeeding rate 11%) for newborns with TSB <20 mg/dL and exclusive breastfeeding 90% (partial breastfeeding 5%) for those with TSB ≥20 mg/dL.
Finally, using the 3498 newborns in the HFHSF group only, we used logistic regression to generate a simplified prediction model for hyperbilirubinemia as defined by the age-specific AAP criteria. We included in the model 3 commonly recorded and significant prognostic variables: maternal race, gestational age, and feeding type. For feeding type, we again collapsed the categories of exclusive and partial breastfeeding (versus formula feeding) because of insignificant differences in the risk of developing hyperbilirubinemia between the 2 groups. To demonstrate an application of the benchmarking model to compare hospital performance on preventing hyperbilirubinemia, we invented hypothetical data about numbers of newborns born at an imaginary hospital by 3 risk factors: exclusive and partial breastfeeding versus formula feeding, gestational age, and maternal race. In this hypothetical hospital, we assumed that 3526 newborns were born in 1 year, and of those, 300 (8.5%) had observed values for TSB exceeding AAP age-specific criteria. We purposefully posited a high number of observed events for demonstration purposes because we assumed that most HCOs are less tightly managed than HFH. We computed the expected occurrences of hyperbilirubinemia as defined by the age-specific AAP criteria for each risk group in this hypothetical hospital by using the prediction model obtained from our regression analysis, ie, these are the occurrences of hyperbilirubinemia to be expected if the hypothetical HCO has an equivalent rate of testing for hyperbilirubinemia as HFH. We then computed an observed events/expected events (O/E) ratio by using the hypothetical observed occurrences of hyperbilirubinemia divided by the occurrences of hyperbilirubinemia expected according to the prediction model to demonstrate how health care providers can adjust for significant risk factors to benchmark their performance against the HFHS follow-up standard.
Demographic Characteristics of the Sample
A total of 6748 newborns were born at HFH from November 1, 1998 to August 31, 2001. Of these newborns, 5579 met the inclusion criteria cited above. Of these, an additional 72 were excluded for reasons such as invalid medical record number, incorrect gestational age, missing neonatal intensive care unit status, adoption, foster placement, and home birth. The total evaluable sample size consisted of 5507 mother-newborn paired participants.
As shown in Table 1, just over half of the study newborns were female, >50% of the study mothers were black (African American), 24% were Latino (Hispanic), 14% were white (Caucasian), and 2% were Asian. Approximately 43% of mothers used formula feeding only, and only 30% were exclusively breastfeeding. Of the study newborns, 13% had a gestational age of <38 weeks. The sample consisted of relatively young mothers (mean age ± standard deviation [SD], 26.21 ± 6.15 years). The mean birth weight was 3340 ± 513.49 g, and the mean gestational age was 39.1 ± 1.4 weeks.
By the end of the first 24 hours of life, 8% (416/5507) of newborns had been discharged; by 48 hours of life, 61% (3353/5507) had been discharged; and by 72 hours of life, 86% (4714/5507) had been discharged. The mean length of stay/age of the newborn at time of hospital discharge was 41.6 ± 20.5 hours for vaginal delivery and 74.1 ± 8.9 hours for caesarian delivery.
Of 5507 newborns born during this period, 36.5% (2009/5507) of newborns delivered at HFH did not have an office or ER visit or readmission after 15 days of age through HFHS (ie, they are in the non-HFHSF group). Nearly 50% of the non-HFHSF group had mothers who were Hispanic, and many of these newborns were followed at a community health and social services center where Spanish-speaking Henry Ford Medical Group providers are available. The remaining 63.5% (3498/5507) of newborns were delivered at HFH and received at least one office or ER visit or readmission at HFHS after 15 days of age (ie, they are in the HFHSF group).
Of these 3498 HFHSF group newborns, 2735 were born during the time period of November 1998 to March 2001, for which we have complete postdischarge visit data including home care visits. Of these newborns, 49% (1335/2735) received some type of follow-up within the first week of life. This number increased to 67% (122/187) for HFHSF group newborns at higher risk, ie, exclusive or partial breastfeeding, white mother, and discharged ≤48 hours after vaginal delivery and ≤72 hours after caesarian section.
Serum Bilirubin Testing
Overall, the proportion of newborns born at HFH between November 1998 and August 2001 who received at least one TSB test at any time was 57.44% (3163/5507). The proportion of all newborns receiving a postdischarge TSB test was 12% (647/5507). If we consider only the HFHSF group newborns, 13% (460/3498) had at least one postdischarge TSB. Among HFHSF group newborns with both white mothers and exclusively or partially breastfed, this number increased to 19% (59/306). Of HFHSF group newborns who had white mothers and were exclusively or partially breastfed and also born at 35 to 36 weeks’ gestation, 69% (9/13) had at least one postdischarge TSB. In other words, higher rates of testing occurred among higher-risk newborns.
Incidence of Hyperbilirubinemia
A maximum TSB level ≥20 mg/dL was identified in 0.6% (21/3498) of all HFHSF newborns and TSB levels meeting or exceeding age-specific AAP criteria in 2.74% (96/3498). A high proportion of HFHSF newborns with hyperbilirubinemia were detected after discharge as defined either by maximum observed TSB ≥20 mg/dL, 95% (20/21), or by AAP criteria, 61% (59/96).
Identification of Prognostic Variables
Logistic Regression Modeling
For this analysis, we collapsed the categories of exclusive and partial breastfeeding because we found no significant differences in the risks of developing hyperbilirubinemia and severe hyperbilirubinemia between exclusively and partially breastfed infants (see Table 2 legend).
As shown in Table 2, logistic regression revealed that the risk of developing a maximum observed TSB ≥20 mg/dL was positively associated with lower gestational age (gestational age: 35–36 weeks), exclusive/partial breastfeeding, male gender and older maternal age, and negatively associated with the mother being black. The “deviation from fit” of the logistic regression model to hyperbilirubinemia as defined by maximum TSB ≥20 mg/dL was not statistically significant (Hosmer-Lemeshow goodness-of-fit test, P = .312). The overall ability of the logistic model (including all variables listed in Table 2) to predict this degree of hyperbilirubinemia was satisfactory (c = 0.789).
Predictors of maximum observed TSB meeting or exceeding age-specific AAP criteria were generally similar to predictors of TSB ≥20 mg/dL, but male gender and older maternal age became statistically nonsignificant. In addition, a positive association was found with access to HFHS care: HFHSF newborns were 89% more likely to have hyperbilirubinemia detected according to age-specific AAP criteria than non-HFHSF newborns. In general, the results indicated that the mother being black was the most significant and stable predictor across the 2 outcome variables.
We also analyzed the data using Poisson regression to estimate relative risk for severe hyperbilirubinemia of various factors, eg, newborn gender, gestational age, and feeding type after allowing for maternal race. As with the logistic regression analysis, Poisson regression revealed that the incidence of severe hyperbilirubinemia (TSB ≥20 mg/dL) was associated negatively with the mother being black but positively with exclusive and partial breastfeeding, lower gestational age, and male gender.
Comparison With the Study of Newman et al
Using the HFHSF group only, we compared bilirubin testing rates with published data from the Newman et al study.2 Testing rates were significantly higher for HFHSF newborns than those in the Newman et al study (62% vs 27%). In addition, the overall crude incidence rate of severe hyperbilirubinemia, as defined by maximum TSB ≥20 mg/dL, was approximately threefold lower in HFHSF newborns than in the Newman et al study (0.6% vs 2%). HFHSF newborns also had a lower incidence of severe hyperbilirubinemia for every risk factor group (see Table 3). For example, the frequency of severe hyperbilirubinemia was 0.4% in HFHSF newborns of black mothers versus 1% for newborns of black mothers in the Newman et al study.
We recognized that the HFHSF and Newman et al groups differed substantially with regard to the distribution of gestational age (P < .0001), maternal age (P < .0001), race (P < .0001), and newborn’s gender (P = .018). As explained earlier, we used data from the second study of Newman et al12 to assume that the exclusive breastfeeding rate in the first study2 was 65% and the partial breastfeeding rate was 11% for newborns with TSB <20 mg/dL, and the exclusive breastfeeding rate was 90% and partial breastfeeding 5% for those with TSB ≥20 mg/dL. Only 30% of the HFHSF newborns were breastfed exclusively, with an additional 25% partially breastfed. The HFHSF group had a younger gestational age, a younger maternal age, a greater proportion of female newborns, a lower proportion of exclusively breastfed newborns, a higher proportion of partially breastfed newborns, and a higher proportion of black mothers. Therefore, we controlled for these variables in Poisson regression analyses to determine if HFHSF newborns still had a lower risk compared with the Newman et al study after controlling for the main effects. The results of Poisson regression analyses (relative risk = 0.26–0.49 for HFHSF group) confirmed that the incidence of severe hyperbilirubinemia was still statistically significantly lower as compared with the Newman et al study population after adjusting for gestational age, gender, race, feeding type, or maternal age (see Table 3).
For this analysis also we collapsed the categories of exclusively and partially breastfed because we had found no significant differences in the risks of developing hyperbilirubinemia and severe hyperbilirubinemia between exclusive and partial breastfeeding infants. The middle 5 columns of Table 4 show the prediction probabilities resulting from our regression analysis of the HFHSF group. For example, the predicted probability of having hyperbilirubinemia as defined by age-specific AAP criteria is 0.181 for a newborn with a white mother, 36 weeks’ gestational age, and exclusively or partially breastfed. That is, a newborn of this kind has an estimated risk of 18% for having a value exceeding age-specific AAP criteria. In the right 5 columns of Table 4 we illustrate how these prediction probabilities can be applied to data from a hypothetical hospital to benchmark the performance of this hospital against that of HFHS. The prediction model takes the specific risks for maternal race, gestational age, and feeding type observed in the HFHSF group (shown in the middle 5 columns of Table 4) and multiplies them by the numbers of newborns in each matching category as posited for the hypothetical hospital. For example, (0.18 × 63) = 11.3 is the expected number of newborns with white mothers, 36 weeks’ gestational age, and exclusively or partially breastfed whose highest serum bilirubin will exceed the age-specific AAP criteria. This computation is repeated to create each of the 50 cells in the right 5 columns of Table 4. As shown in Table 5, the expected numbers of newborns with hyperbilirubinemia for all 50 cells in the right 5 columns of Table 4 are then summed to give the total number of expected events (167.8). That is, 167.8 newborns with TSB values exceeding age-specific AAP criteria would have been expected in the hypothetical hospital if it had an equivalent rate of testing for hyperbilirubinemia as HFH. Because we have posited 300 occurrences of hyperbilirubinemia in this hypothetical hospital, the O/E ratio is 300/167.8 = 1.8. In our hypothetical hospital, then, the observed rate of hyperbilirubinemia is 1.8 times higher than in the HFHSF group after adjusting for race, gestational age, and feeding type. This indicates that the hypothetical hospital is much less successful in preventing hyperbilirubinemia than HFH and might be able to improve care for its newborns by adhering closely to a protocol such as that used at HFH. A simple tool for using the benchmarking model is available at http://www.hsph.harvard.edu/facres/plmr.html.
A better understanding of the factors that predispose to hyperbilirubinemia is needed to design systems of care that will prevent newborns from developing levels that put them at risk for kernicterus. In our logistic regression analysis, we found that maternal race, gestational age, and exclusive or partial breastfeeding are the most significant risk factors for both hyperbilirubinemia as defined by age-specific AAP criteria and for severe hyperbilirubinemia defined as TSB ≥20 mg/dL. Our finding of an association of hyperbilirubinemia with previously identified factors such as maternal race, newborn gender, gestational and maternal ages,2 and feeding method13 attests to the validity of our study. In addition to these clinical factors, we found that the HFHSF variable was a significant predictor for hyperbilirubinemia as defined by the age-specific AAP criteria. The results indicated that the HFHSF group was 89% more likely to have hyperbilirubinemia detected than the non-HFHSF group. These results suggest that better access to care and continuity lead to higher rates for detection of jaundice.
Our comparison to the Newman et al data was made after controlling for gestational age, sex, maternal race, feeding type, and maternal age. We did not include other important prognostic variables identified by the later study of Newman et al12 such as family history of jaundice, bruising, and cephalohematoma (Newman’s odds ratios: 6, 4.0, and 3.3, respectively) because they were not included in the 1999 Newman et al study that we used for our comparison and we did not collect them as part of our study design. It also was not possible to examine the interaction effects of the predictor variables because we only had access to the tables in the original publication from Newman et al. After controlling for the main effects, the results indicated that HFHSF newborns, whose care was governed by a rigorous bilirubin screening, follow-up, and treatment program, were less likely to have severe hyperbilirubinemia than the newborns in the Newman et al study, whose care was given in hospitals that, on average, used less-rigorous protocols.2 We speculate that the significant differences shown in these analyses are caused by differences in the process of care.
Our benchmarking model grew out of our awareness that differences in the characteristics or case-mix variables among different HCOs hinder direct comparison. By controlling for those variables known to be related to higher TSB levels, comparisons can be made that allow HCOs an opportunity to benchmark their performance against a standard. The rigorous protocol for detection of hyperbilirubinemia at HFH affords the opportunity to use data from this HCO as a benchmark for good clinical management.
Potentially, by including several variables that the 2000 Newman et al study found to predict severe hyperbilirubinemia, namely family history of jaundice, bruising, and cephalohematoma, we could refine our benchmarking model. However, similar to many clinical terms, these variables are not typically defined in a standardized way. HCOs interested in using our model may not collect these variables in the same way as in the Newman et al study. For instance, how many and how large must bruises be to qualify for coding as “bruising?” Should the family history of jaundice include only prior pregnancies of the mother and father concerned, and if so, how should we code primiparae? Also, in our benchmarking model we collapsed exclusive and partial breastfeeding into one category because of insignificant differences in the risk of developing hyperbilirubinemia between the 2 groups. We believe that this makes our model more useful to other HCOs, because the definition of exclusive versus partial breastfeeding on discharge is also not standard across HCOs. Is it based on the mother’s reported intent or by the feeding given in the hospital? Does even one feeding of formula justify removing the label of exclusive breastfeeding? The distinction between only formula feeding versus exclusive or partial breastfeeding seems more likely to be standardizable across HCOs, because mothers who did not breastfeed in the hospital are unlikely to begin doing so after discharge.
It is important to point out that this model has been developed under the assumption of equivalent rates of testing; that is, any hospital using the model would have to be confident that evaluation and testing of newborns for hyperbilirubinemia both pre- and postdischarge were at least as good as those reported in this study. Under the assumption of equivalent evaluation and testing, an O/E ratio >1 indicates that the hospital is managing hyperbilirubinemia worse than the benchmarked reference of high quality. An O/E ratio <1 indicates better management of hyperbilirubinemia than the reference standard. If a lower ratio is obtained simply because one hospital has less evaluation and testing of newborns and so misses more cases than HFHS, the assumption is violated.
We offer our methods to help HCOs evaluate whether their programs are yielding satisfactory hyperbilirubinemia monitoring and management outcomes. By standardizing rates of occurrence of hyperbilirubinemia for 3 risk factors (ie, gestational age, maternal race, and feeding type) and assuming equal rates of testing, any observed differences should primarily be caused by either practice patterns or random measurement error. We believe that our methods are innovative and will ultimately prove to be valuable to organizations that desire to benchmark their rates of hyperbilirubinemia with a known reference standard of high performance.
Currently, many HCOs that wish to adopt this approach might find that the required postdischarge data are unavailable to them. However, the JCAHO Sentinel Event Alert requires HCOs to assume responsibility for preventing kernicterus even if the newborn leaves the hospital during the critical risk period of the first few days of life. As HCOs accept this challenge, our benchmarking model will permit them to compare their management of hyperbilirubinemia with a reference population that received care according to a strictly managed protocol.
We found that, by using logistic regression analysis, shorter gestational age, exclusive or partial breastfeeding, and maternal race (not black) are associated with a higher incidence of neonatal hyperbilirubinemia, a result that is consistent with many previously published studies.2,13 We further demonstrated that newborns who received their follow-up care at HFHS had a lower overall risk of severe hyperbilirubinemia as compared with those HCOs in the Newman et al study. Although only a randomized study can prove a causal association between rigorous in-hospital and postdischarge management and lower rates of hyperbilirubinemia, our results provide a reference rate for other similar organizations desiring to benchmark their performance regarding implementation of their practice guidelines for managing hyperbilirubinemia in newborns.
APPENDIX I: KEY POINTS
The text shown below appears on the reverse of the “Laminated Pocket Card” shown in Fig 1.
Near-term infants (36–37 weeks/2250–2700 g, unless small for gestational age) need closer follow-up and lower thresholds for repeat bilirubin/phototherapy.
For Breastfeeding mothers, ensure frequency/adequacy of breastfeeding (number of stools, wet diapers, etc). Supplement with formula if in doubt. The combination of borderline gestation, breastfeeding, and a first-time breastfeeding mother often increases the risk of hyperbilirubinemia. These infants must be followed closely.
During home/clinic visits, question/evaluate the infant’s behavior (cry, suck, activity, etc).
Infants whose serum bilirubin levels are >8 mg/dL at 24 hours of age need a follow-up evaluation and/or a repeat bilirubin drawn at 48 to 72 hours of age.
In those categories where phototherapy is mandated or considered, a work-up for hemolysis, ABO incompatibility, glucose-6-phosphate dehydrogenase deficiency, etc must be considered.
For phototherapy (yellow zone), at least double phototherapy must be instituted (biliblanket and overhead or 2 overheads). Effective spectral irradiance is ∼20 μW/cm2/nm.
Intensive phototherapy (orange zone) generally requires double or triple phototherapy and hydration if the infant is feeding poorly. Supplement breast feeding if indicated, but do not discontinue breastfeeding. If bilirubin fails to drop a few mg/dL in a few hours, then exchange.
Repeat bilirubin or phototherapy (pink zone). Breast feeding issues are often a consideration. Consider supplementation if increased frequency of feedings is unsuccessful. If the infant is feeding poorly, continue breastfeeding, supplement with formula, and start phototherapy.
APPENDIX II: DEFINITIONS
Hyperbilirubinemia is defined as an age-specific TSB in the first 30 days of life greater than or equal to the category used by the AAP for consideration of phototherapy in its 1994 Practice Parameter for Management of Hyperbilirubinemia in Healthy Term Newborns,1 i.e., TSB ≥12 mg/dL at 25 to 48 hours of age, TSB ≥15 mg/dL at 49 to 72 hours of age, or TSB ≥17 mg/dL at 72+ hours of age.
Severe hyperbilirubinemia is defined as TSB ≥20 mg/dL in the first 30 days of life.
Management of hyperbilirubinemia is the set of practices designed to detect hyperbilirubinemia early in order to prevent or treat severe hyperbilirubinemia before it can cause kernicterus.
This study was supported by the Agency for Healthcare Research and Quality (grant R01 HS09782, Making Advances Against Jaundice in Newborn Care).
We acknowledge the assistance and support of our colleagues Khushnam Sidhwa, Ann G. Lawthers, DSc, and Laura Peterson, BSN, SM, for work in establishing the database. We also thank Dr. Savitiri Kumar, MD, and the participating staff of the Division of Neonatology of HFHS for help and support.
- Received October 28, 2002.
- Accepted April 7, 2003.
- Reprint requests to (R.H.P.) Harvard School of Public Health, 677 Huntington Ave, Boston, MA 02115. E-mail:
THE EXERCISE BUSINESS
“There are more than 17 000 health clubs in America today. One out of every eight adult Americans is a member; the total number of members is 32.8 million. The median amount of money coming into a health club in 2002 was $5 116 000. And these clubs are just a fragment of the exercise business … an entire industry and mystique has built up around shoes alone. The National Sporting Goods Association reports that in 2001 the sales of footwear for sports reached $13.3 billion.”
Kolata G. Ultimate Fitness. Farrar Straus Giroux; 2003
Submitted by Student
- ↵American Academy of Pediatrics, Provisional Committee for Quality Improvement. Practice parameter: management of hyperbilirubinemia in the healthy term newborn. Pediatrics.1994;94 :558– 565
- ↵Bhutani VK, Johnson LH, Sivieri EM. Predictive ability of a predischarge hour-specific serum bilirubin for subsequent significant hyperbilirubinemia in healthy term and near-term newborns. Pediatrics.1999;103 :6– 14
- ↵Maisels MJ. Jaundice. In: Avery GB, Fletcher MA, MacDonald MG, eds. Neonatology: Pathophysiology and Management of the Newborn. Philadelphia, PA: JB Lippincott, Co; 1999:765–819
- ↵American Academy of Pediatrics, Subcommittee on Neonatal Hyperbilirubinemia. Neonatal jaundice and kernicterus. Pediatrics.2001;108 :763– 765
- ↵Maisels MJ, Kring E. Length of stay, jaundice, and hospital readmission. Pediatrics.1998;101 :995– 998
- ↵Joint Commission on Accreditation of Healthcare Organizations. Sentinel event alert, Issue 18. Available at: http://www.jcaho.org/about+us/news+letters/sentinel+event+alert/sentinel+event+alert+index.htm. Accessed October 28, 2003
- ↵Stevenson DK, Fanaroff A, Maisels MJ, et al. Prediction of hyperbilirubinemia in near-term and term infants. Pediatrics.2001;108 :31– 39
- Copyright © 2003 by the American Academy of Pediatrics