In Reply We appreciate the comments of Edidin and colleagues as they touch on key elements of our study,1 and their previous article2 provides insightful contrast to our own.
We included the “preprocedure subgroup” analysis to illustrate that a substantial proportion of the augmented group (29%) had a markedly lower risk of complications compared with controls despite being “theoretically” equivalent—both groups had the same treatment during this time (no augmentation), and we controlled for baseline characteristics, including Quan comorbidity scores, prior inpatient admissions, and chronic pulmonary disease, among others, using traditional multivariate models. Edidin et al are right to be concerned that these traditional multivariate models might not adequately account for acute differences in health at the time, such as patients needing emergent care. We agree. There are many other clinical details available in real-time that are not evident in billing claims data. The entire clinical picture at presentation, past and present, will influence therapeutic decisions as well as eventual patient outcomes. We suggest selection bias is the unmeasured factor allowing 2 “theoretically” equivalent groups to have such different outcomes.
McCullough BJ, Deyo RA, Jarvik JG. Treatment of Osteoporotic Vertebral Fractures—Reply. JAMA Intern Med. 2014;174(4):642-643. doi:10.1001/jamainternmed.2013.13481