We conducted a case-control study to elucidate associations between pneumonia in elderly individuals and 23-valent pneumococcal polysaccharide vaccine (PPSV23) and seasonal influenza vaccine (influenza vaccine). Here, we examined selection of controls in our study using an analytic epidemiology approach. The study period was from October 1, 2009 through September 30, 2014. Cases comprised >65-year-old patients newly diagnosed with pneumonia. For every case with pneumonia, two patients with other diseases (one respiratory medicine, one non-respiratory medicine) who were sex-, age-, visit date- and visit hospital-matched were selected as controls. Odds ratios (ORs) and 95% confidence intervals (Cls) of vaccination for pneumonia were calculated using conditional logistic regression model. Similar analyses were also conducted based on the clinical department of controls. Analysis was conducted in 234 cases and 438 controls. Effectiveness of pneumococcal vaccination or influenza vaccination against pneumonia was not detected. Proportions of either vaccination in controls were greater among respiratory medicine (pneumococcal vaccine, 38%; influenza vaccine, 55%) than among non-respiratory medicine (23%; 48%). Analysis using controls restricted to respiratory medicine showed marginally significant effectiveness of pneumococcal vaccination (OR, 0.59; 95%Cl, 0.34-1.03; P = 0.064) and influenza vaccination (0.64; 0.40-1.04; 0.072). However, this effectiveness might have been overestimated by selection bias of controls, as pneumonia cases are not necessarily respiratory medicine patients. In the analysis using controls restricted to non-respiratory medicine, OR of pneumococcal vaccination for pneumonia was close to 1, presumably because the proportion of pneumococcal vaccination was higher in cases than in controls. Because pneumococcal vaccine was not routinely administered during the study period, differences in recommendations of vaccination by physician in different clinical departments might have greatly affected vaccination proportions. When we select controls, we should consider the background factors (underlying diseases, clinical department, etc.) which affect physicians' recommendation of vaccination. (C) 2017 The Authors. Published by Elsevier Ltd.