In claims database research, when counting the number of patients (number of visits), development of a disease definition that combines one or more conditions is necessary. We illustrated to what extent the calculated number of patients is affected by differences in conditions in the disease definitions, using HIV and syphilis as examples. Disease definitions which appear clinically plausible sometimes do not represent the actual situation. To avoid such pitfalls, a sensitivity analysis is important with changing conditions. Although clinical research using claims databases is growing, there is a potential risk of producing lots of studies whose results will not be replicated later. Researchers are encouraged to conduct critical appraisal of disease definitions.