Variation in prescriber responses to drug safety alert messages: disease, symptom, or red herring?

I n this issue of JMCP, Feifer and James provide significant food for thought in their examination of geographic variation in the rate of automated drug safety alerts as well as prescriber response to such alerts. 1 Findings from this study suggest that the major influences on this variation are uncertain, as member demographics explained a relatively small proportion of the observed variance in alerting rates. The authors conclude that outreach and education are necessary to improve prescriber behavior where warranted. When broken down into the study's constituent parts, however , both measures of variation are linked to much broader considerations regarding the state of American health care practice. The rates of " alerting events " in this study varied considerably by state, ranging from 78 to 240 per 1,000 members. However, is this variation really surprising? After all, small-area variation in health care utilization and clinical practice is a well-documented phenomenon; 2,3 why should awareness of potential drug safety concerns be subject to any less variation? As an example, consider one of the clinical situations that Feifer and James present in Table 1 of their study report—the prescribing of stimulants for children with attention deficit hyperactivity disorder (ADHD) and significant cardiac structural or rhythm abnormalities. This is indeed a significant concern, but its calculus involves more than simple variation in prescriber awareness of the clinical problem. Indeed, the documented geographic variation in both the prevalence of diagnosed ADHD and receipt of ADHD medication is significant, with 2-to 3-fold variation reported in each measure by state of residence. 4 While there are no published data on variation in the prevalence of co-diagnosed ADHD and structural/rhythm abnormalities in children, it is not unreasonable to imagine this prevalence to be (a) relatively low; and (b) distributed in some disproportionate fashion regionally, perhaps in relation to the location of advanced pediatric cardiology services and/or children's hospitals. Variability in inappropriate prescribing is therefore likely to be a result of a complex interaction of disease epidemiology, practice variation, the likelihood of a given clinician encountering a relatively rare safety issue in practice, and many other concerns. As a clue for the development and targeting of a programmatic intervention to change prescribing behavior, it is also a red herring. Why? The answer can perhaps be found in the examination of Feifer and James' second key measure, the rate of successful therapy change per alerting event. The …