The author replies
The authors correctly point out that we were unable to account for statin use outside of the hospital and point to the likelihood that the majority of patients using statin would be under chronic treatment. The authors further suggest conducting a sensitivity analysis in patients receiving other oral medications throughout the duration of hospitalization. I agree with this assessment, and point to the sensitivity analysis performed with the addition medications, many of which would be assumed to be used outside of the acute care setting. This subanalysis included patients using zolpriden (statin group 8.6%; nonstatin 8.3%), promethazine (3.1%; 3.2%), metoclopramide (12.7%; 12.9%), lorazepam (29.8%; 31.4%), famotidine (16.9%; 18.4%), and diphenhydramine (7.9%; 8.3%). The persistence in the association with a reduction in delirium with the use of statin in this sensitivity analysis with the inclusion of these medications points to the likelihood that there are healthy users in both the statin and nonstatin groups.
The authors also imply that the statin users may have had a lower severity of illness than the nonstatin users. I point to the inclusion of the Deyo-Charleson Comorbidity Index (DCCI) (4) score in the propensity match and its use in the ICU (5). Furthermore, our sensitivity analysis with both low and high risk DCCI subanalysis provides further evidence that the severity of illness was not a confounder in the analysis.
Following the suggestions of Shrank et al (3), the collection of detailed information on home medication use or the use of preventive services would have provided further adjustment of the healthy user effect as a confounder in the observational study. We appreciate the authors raising this issue and will strive to include more objective measures of the healthy user effect where possible.