DO PATIENT CHARACTERISTICS EXPLAIN PRACTICE VARIABILITY IN THE DIAGNOSIS AND TREATMENT OF FEBRILE INFANTS? DA Bergman, RH Pantell, A Lin, M Mayer, R Olshen, R Wasserman. Pediatrics, Stanford University, Stanford, CA; Health Research and Policy, Stanford University, Stanford, CA ; Pediatric Research in Office Settings (PROS), American Academy of Pediatrics; and Pediatrics, University of California, San Francisco, San Francisco, CA. Presented as a platform presentation at the 2000 Pediatric Academic Societies? Annual Meeting.

BACKGROUND: Significant variability in clinical practice has been documented for a wide variety of conditions both across physicians and between regions. Yet few studies exist in pediatrics that document whether this variability is appropriate (i.e. explained by differences in clinical severity) or is accounted for by differences in practice type, practitioner characteristics or region. Understanding the reasons for this variability will be important in identifying opportunities to improve care.

OBJECTIVE: To understand the basis for clinical practice variability in the diagnosis and treatment of febrile infants by examining the explanatory power of patient, practice/practitioner and geographic variables.

DESIGN/METHODS: Data was analyzed from a prospective cohort of 2170 febrile infants seen by 579 pediatricians located in 220 practices. To develop a summary score for treatment intensity we performed a principal components analysis (PCA) on five dichotomous clinical variables: hospitalization, lumbar puncture, urinalysis, blood work, and initial antibiotic administration. Our first principal component explained 56% of the overall variability in the five treatment variables of interest. We considered 13 potential predictors which were divided into 3 categories: 1) patient variables, 2) practitioner/practice characteristics and 3) geographic location. The summary score developed from the PCA was then regressed on these three classes of explanatory variables.

RESULTS: The selected 3 categories explained > 30% of the variance R2 = .337 (p< .0001). The patient variables of age, temperature and clinical appearance explained 28.9% of this variance. Practice/practitioner variable explained < 2% of the variance R2 = .017 (p< .0001), and geographic region explained < 3% of the variance, R2 = .028 (p<.0001). While the non-clinical variables were statistically significant their explanatory power was minimal.

CONCLUSION: Our results showed that > 90% of the variability in clinical practice that can be explained in our model is accounted for by individual patient characteristics. This suggests that differences in clinical presentation underlies much of the variability in the treatment of febrile infants. It is reassuring to know that for this common and potentially serious condition, the type of care you receive is dependent on clinical factors and not on the pediatrician you choose or in what region you reside.