 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Sudhakar Ezhuthachan, MD, DCH, FAAP
Christine Newman, MS, RNC, CNNP
Henry Ford Health System,
Detroit, MI 48202
M. Jeffrey Maisels, MD, FAAP
William Beaumont Hospital,
Royal Oak, MI 48073
Marcia A. Testa, MPH, PhD
Department of Biostatistics,
Harvard School of Public Health,
Boston, MA 02115
To the Editor.—
We agree with our colleagues from the Northern California Kaiser Permanente Medical Care Program (NC-KPMCP)1 that observational studies such as ours at the Henry Ford Health System (HFHS)2 and theirs3,4 are subject to competing explanations for findings. None of these studies were randomized, double-blind, controlled studies. In fact, a benchmarking model is intended to promote discussion of competing explanations. The model is used to ask: If a peer organization seems to have better outcomes, could this be due to a difference in practices, or is it spurious? Also, if the former could be true, what are our peers doing differently, and could we improve our outcomes by adopting their practices? If practice differences seem plausible as an explanation of outcome differences, then one may try adopting the practices of the benchmark organization and then reexamine one’s results. Similar to most quality-improvement activities, benchmarking uses a dialog among peers and an iterative approach to clarify relationships between practices and outcomes. We appreciate that Newman et al are interested in promoting such dialog.
Our colleagues from NC-KPMCP first question our definitions of hyperbilirubinemia and severe hyperbilirubinemia. They argue that pediatric health care providers cannot influence the low levels of hyperbilirubinemia that we selected, because they are below those at which the American Academy of Pediatrics (AAP) 1994 guideline recommends phototherapy.5 Apparently, they see phototherapy as the only means to arrest increasing hyperbilirubinemia. However, Maisels and Kring6 and several other authors7–11 have drawn attention to breastfeeding with inadequate intake as an important contributory cause for severe hyperbilirubinemia and the possibility of preventing hyperbilirubinemia by supportive management during the establishment of breastfeeding. From the beginning of our study we were aware that the HFHS guideline for management of hyperbilirubinemia was lenient in its recommendations for phototherapy while it strongly emphasized early screening for hyperbilirubinemia and management of breastfeeding with inadequate intake. We therefore chose definitions of outcome that could be responsive to these guideline recommendations. Methodologic considerations also favored choosing these definitions. Outcomes such as total serum bilirubin (TSB) 
25 mg/dL occur too rarely to permit modeling in a data set with several explanatory variables and just over 5000 cases.
Newman et al comment that the level of exclusive breastfeeding at HFHS is less than at NC-KPMCP. NC-KPMCP reported 66% exclusive and 11% partial breastfeeding, whereas we reported 30% exclusive and 25% partial breastfeeding.2,4 This imbalance is not surprising, because the NC-KPMCP population is 53% white and 19% black, whereas the HFHS population is 67% black and 14% white,2,4 and black mothers have often been reported as less likely to breastfeed. That is why we adjusted for race, breastfeeding, and other risk factors in comparing outcomes. However, we suspect that comparisons between our 2 studies using the feeding variables are complicated by methodologic differences. For our article, the variables describing feeding practices came from the nursing discharge logs and reflected the mother’s expressed intention about feeding at the time of her discharge. Because these data were regularly used to chart progress for the HFHS participation in the Baby-Friendly Hospital initiative, nurses were aware of the importance of recording these data accurately by 3 categories: exclusive breastfeeding, partial breastfeeding, and exclusive formula feeding. In the NC-KPMCP article,4 it seems that data categorizing breastfeeding may have been collected in part by retrospective record review and in part by interview of the mothers 2 to 3 years after the birth event. If patterns of feeding classified as exclusive breastfeeding at NC-KPMCP were classified as partial breastfeeding at HFHS, we could expect to see the differences in patterns of feeding described above.
We agree with Newman et al that promoting breastfeeding might increase the risk for hyperbilirubinemia. However, with excellent lactation help, this may not occur.7 The HFHS had a vigorous program for promoting breastfeeding but also, as our article reports, had a vigorous program for overcoming difficulties in establishing breastfeeding. There are tantalizing glimpses of the competing explanation of breastfeeding with inadequate intake in the article by Newman et al.4 At the time of readmission, all 66 cases readmitted for phototherapy had lost an average of 6.7% of birth weight, and 10 of 66 had lost >12% of birth weight. Also, those newborns who became severely hyperbilirubinemic were more likely to receive early follow-up as recommended by the AAP guideline,5 "primarily because of an almost 20% greater unscheduled outpatient visit rate in the 72 hours following an initial stay of less than 48 hours."4 Could mothers who encountered difficulties in breastfeeding have sought these unscheduled visits?
There are 2 other features of the HFHS guideline that could have influenced the outcomes we observed. First, the HFHS guideline emphasizes closer attention to newborns of gestational age of 36 to 37 weeks or with a birth weight of 2250 to 2700 g, including intervention with phototherapy at lower age-specific TSB levels than for more mature newborns. The 1994 AAP guideline did not address newborns of <37 completed weeks of gestation.12 Newman et al speculate that pediatricians in their study did not select 36- to 37-week-old newborns for more intensive follow-up and testing than more mature newborns.3
Second, as we reported, the HFHS practices universal screening with transcutaneous bilirubin (TcB) measurement, with a TSB drawn at a specific trigger value of TcB. Based on these screening results and other risk factors, discharge is delayed for higher-risk newborns or proceeds with a home care nurse referral that may include a specific request for a TSB to be drawn but always permits the home care nurse to draw a TSB. Our article provided evidence that HFHS successfully focuses attention on risk factors: at least 1 postdischarge TSB was drawn for 9 of the 13 newborns of 35 to 36 weeks’ gestation who were white and fully or partially breastfed (ie, high risk).
Newman et al propose as a competing explanation for the differences in outcomes between HFHS and NC-KPMCP that it results from case mix. Specifically, as we reported in Table 3,2 we adjusted for the main effect of each case mix variable 1 at a time and not simultaneously. We did this because we used the tabulated data directly from the Newman et al article, which did not contain cross-tabulations, and also because our models were limited by the number of cases in the sample. The example Newman et al cite in their letter is that "the difference in the rate of hyperbilirubinemia between white newborns from the NC-KPMCP (1.7%) and HFHS (1.2%) reported in Table 3 of the Chou et al article can be explained by the greater proportion of exclusive breastfeeding in the NC-KPMCP (66% vs 30%)."1 This statement implies that if controlling for race yields a relative risk (RR) of 0.5 and controlling for breastfeeding also yields an RR of 0.5, then the total risk reduction if the 2 variables were entered simultaneously would somehow reverse this direction substantially toward 1 (no difference). However, this type of reversal would require an extremely strong breastfeeding-by-race interaction. That is, the impact of race would essentially have to alter the relationship between breastfeeding and outcomes so that it would become protective or less risky for one category of race as compared with another. This is the only way that one could maintain an RR of 0.5 for both factors individually but when entered simultaneously produce an RR close to 1. We observed no such interaction effects in our multivariate analyses,2 Newman et al report no such interaction effects in their multivariate analyses,3,4 and we know of no reports of such interaction effects in the literature to date. If one assumes a constant marginal risk over each level of race and breastfeeding within the HFHS and NC-KPMCP groups, the RR for the between-group effects should be similar whether controlling for each individually or simultaneously.
We do not find the Newman et al competing explanation of case mix plausible for any of the risk factors, in part because of the differences in classifying breastfeeding described above. In addition, the main effects for all risk factors were strong and consistently favored better outcomes at the HFHS. That is, for each risk group and for each level of risk within each risk group, the HFHS newborns had less hyperbilirubinemia than newborns at NC-KPMCP. In the example Newman et al select, race and feeding, these main effects are the strongest.
The second competing explanation offered by Newman et al is that interlaboratory variability in measurements of TSB could explain the differences in hyperbilirubinemia between the HFHS and NC-KPMCP. Laboratory standardization is a problem for benchmarking and also for management of newborns, since a sequence of TSBs done to guide clinical decisions may be done in different laboratories.13 In November 2001, we sponsored a meeting of our Making Advances Against Jaundice in Newborn Care (MAJIC) Consortium with a representative of the College of American Pathologists (CAP) to discuss the problem. In January 2002, CAP began to include within its laboratory standardization program an improved proficiency testing sample for neonatal bilirubin of 19.5 mg/dL human serum. The CAP program tests both accuracy and interlaboratory variation. Accuracy evaluates how well a given laboratory’s result on the proficiency test sample matches the result obtained in a CAP reference laboratory. Interlaboratory coefficients of variation test how well all participating laboratories agree on their results for the proficiency test sample. Accurate laboratories vary little from one another, because their results cluster closely around those of the reference laboratory. In their original study, Newman et al reported that 6 of their 11 hospitals were CAP-accredited, which suggests that they are more accurate than the average laboratory. The HFHS laboratory is also CAP-accredited and also likely to be more accurate than the average laboratory. Thus, we think it unlikely that interlaboratory variation explains the 3-fold differences in percentage of newborns experiencing severe hyperbilirubinemia between the NC-KPMCP and HFHS populations. However, any hospital engaging in benchmarking regarding hyperbilirubinemia should certainly explore this issue as one of the competing explanations for differences from the benchmark.
Differences in practice remain among the competing explanations for differences in the incidence of hyperbilirubinemia. Indeed, Newman et al reported interhospital differences in the incidence of hyperbilirubinemia that were not explained by biological risk factors and which they speculate were caused by differences in practices.3 Thus, we see a clear agenda ahead. Why not improve our capacity for benchmarking while at the same time exploring adoption of apparently better practices in the pursuit of improvement? We could disseminate more standardized methods for documenting feeding practices. The CAP and AAP could collaborate to improve laboratory standardization for neonatal TSB levels. Meanwhile, pediatric care providers could improve their screening for newborns at high risk, provide close follow-up to those at risk, and improve management of breastfeeding with inadequate intake. We hope that dialog, exploratory analysis, and benchmarking for quality improvement will flourish!
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
