CONTROLLING FOR CONFOUNDING BY INDICATION FOR TREATMENT - ARE ADMINISTRATIVE DATA EQUIVALENT TO CLINICAL-DATA

There has been controversy about whether confounding by indication for treatment-that is, owing to physicians' conscious efforts to base tre atment decisions on patients' pretreatment prognoses-makes nonrandomiz ed, observational comparisons of treatments invalid. Some now believe evidence from studies of practice variation means that physicians' tre atment decisions have little relationship to patients' prognostic clin ical characteristics. They therefore believe that patients who receive different treatments should vary little in their baseline prognoses, and multivariable statistical methods should easily be able to adjust for any resultant confounding, even when analyses are restricted to ad ministrative rather than clinical data. The objective of this study is to determine whether adjusting for variables found in administrative data sets produces the same results as does adjusting for clinical var iables. Data were reanalyzed from a previously enrolled prospective se quential cohort of 227 hospitalized patients with suspected bacteremia who had blood cultures. The treatment under study was aminoglycoside therapy given empirically, that is, before blood culture results were known. The outcome of interest was death during hospitalization. Univa riable analyses suggest that empiric aminoglycoside therapy had a posi tive association with mortality, by univariable logistic regression, o dds ratio (OR) = 3.1 (95% confidence interval = [1.6, 5.8]). Few admin istrative variables had univariable associations with aminoglycoside u se or death. Multivariable analyses that controlled for them still sug gest that aminoglycosides increased mortality; for example, in one mod el, adjusted OR = 3.2 (1.6, 6.5). Many clinical variables were strongl y associated with aminoglycoside use or death. Analyses that controlle d for them suggested that empiric aminoglycosides did not increase mor tality; for example, in one model, adjusted OR = 1.2 (0.55, 2.7.) Resu lts of adjustment for confounding using administrative data disagreed with the results of adjustment using clinical data. It is concluded th at nonrandomized, observational outcome studies that fail to control f or prognostic differences between patients receiving different treatme nts may not always be valid.