ORIGINAL REPORT Using prescription claims data for drugs available over-the-counter (OTC) y Marianne Ulcickas Yood DSc, MPH 1,2,3 * , Ulka B. Campbell 3 , Kenneth J. Rothman 4 , Susan S. Jick 4,5 , Janet Lang 4 , Karen E. Wells 2 , Hershel Jick 5,6 and Christine C. Johnson 2 1 Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT 2 Josephine Ford Cancer Center, Henry Ford Health System, Detroit, MI 3 EpiSource, LLC, Hamden, CT 4 Department of Epidemiology, Boston University School of Public Health, Boston, MA 5 Boston Collaborative Drug Surveillance Program, Lexington, MA 6 Department of Medicine, Boston University School of Medicine, Boston, MA SUMMARY Purpose Many pharmacoepidemiologic studies use automated prescription claims data to estimate the association between exposure and disease. One limitation of automated data, when studying drugs that are also available via retail, is that over-the-counter (OTC) exposure is missed. The purpose of this study is to quantify the effect of misclassification of OTC use in research that uses prescription claims data as the sole source of exposure information. Methods We conducted a sensitivity analysis in the context of studies of non-steroidal anti-inflammatory drugs (NSAIDs) and colorectal cancer. The following factors were widely varied to examine the impact on the validity of the effect estimate for NSAIDs and colorectal cancer: (1) the overall prevalence of NSAID exposure in the population, (2) the proportion of NSAID exposure due to OTC use (the prevalence of missed NSAID exposure in studies of prescription claims), and (3) the true risk ratio (RR true ). We graphed the RR that would be observed (RR observed ) as a function of overall prevalence of NSAID use and the prevalence of NSAID use that is OTC exposure. Results We found that when the true RR ranges from 0.25 to 0.75, missing OTC drug exposure is not a large source of bias in those situations in which the overall prevalence of drug use is relatively low (less than 35%) and the proportion of drug use that is OTC exposure is as high as 80%. Conclusion Results from our sensitivity analysis indicate that, in many circumstances, prescription claims data can give valid estimates of association even though some of the drugs are available OTC. Copyright # 2007 John Wiley & Sons, Ltd. key words — prescription; claims; non-steroidal anti-inflammatory drugs (NSAIDs); over-the-counter (OTC); misclassification Received 22 February 2006; Revised 26 April 2007; Accepted 27 May 2007 INTRODUCTION The availability of administrative data for pharma- coepidemiologic research has enabled investigators to conduct relatively quick and efficient studies addressing important questions about the effects of prescription drugs. When the drug of interest is also pharmacoepidemiology and drug safety 2007; 16: 961–968 Published online 26 July 2007 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/pds.1454 *Correspondence to: M. Ulcickas Yood, 275 Blue Trail, Suite 3, Hamden, CT 06518, USA. E-mail: muyood@muyood.com y All authors have made a substantial contribution to the manuscript, and have read and approved the final version. This manuscript has not been previously published in whole or part. The authors have not identified any conflicts of interest, financial or otherwise. Copyright # 2007 John Wiley & Sons, Ltd.