Purpose To show the necessity of distinguishing several patterns of drug prescribing that may lead to co-medication. It is demonstrated how these different patterns can be investigated using large databases containing pharmacy data or reimbursement data. Methods Two examples illustrate how the particular pattern of co-medication studied will influence the reported proportion of patients having co-medication, the use of antidepressants among people using anticonvulsants, and the use of antihistamines among people receiving penicillines. Results Depending on definition and period considered, the percentage of anticonvulsant users co-medicated with antidepressants ranged from 5.8% (95%CI 5.0%, 6.8%) to 14.5% (95%CI 13.2%, 15.9%) in 2000. Comparing 2002 with 2000, the ratio of proportions ranged from 1.3 to 2.1. The percentage of people who received penicillines and were co-medicated with antihistamines ranged from 0.5% (95%CI 0.4%, 0.6%) to 9.7% (95%CI 9.3%, 10.2%) in 2000. Comparing 2002 with 2000, the ratio of proportions ranged from 1.2 to 1.6. Conclusion The co-medication patterns investigated yielded clinical as well as statistically significant different estimates. The estimates differed up to a factor 2.5 for the drugs usually prescribed for long periods, and a factor 12 for drugs prescribed for short periods. Hence, we propose to distinguish the patterns 'co-prescribing', 'concomitant medication,' and 'possibly concurrent medication.' The research question determines the co-medication pattern of interest, and the drug and disease under study determine the time window. Copyright (c) 2006 John Wiley & Sons, Ltd.