抄録
The purpose of this study is to examine the possibility of using the data file of computerized health insurance billing system maintained by our University Hospital for postmarketingsurveillance (PMS) for adverse drug reactions (ADRs).
We selected 12 widely used drugs now requiring special attention because of their potentially serious ADRs and certain diseases that might relate to the ADRs. The coverage period is January 1982-November 1988. The total number of the patients covered was 109, 088.
First, the ratios of patients with the diseases to patient groups of the following two kinds were determined: (A) all the patients who took the drugs; (B) all the patients who did not take the drugs. Then, fourfold tables were prepared. The association between the drug and the disease was decided to be significant when the odds ratio computed from the tables was larger than 1.0 and X2 value was larger than 3.84.
Second, focusing on the cases having the said significant associations, we tried to set the control group (C) composed of patients having taken none of the drugs who matched with group (A) patients in terms of the following four factors: consultation clinic, sex, age, and in- or out- or in-/out-patient.
Finally, we could choose from group (A) those who 1 to 1 matched to group (C) [group (A) ']. In this, random sampling [1/2 from group (A)] was required for two of the drugs. As a result, group (A) ' and (C) were judged to be comparable as to the four factors.
Then, the ratios of patients with the diseases to group (A) ' and group (C) were determined for comparison with the said method, using Fisher's exact test instead of x2 test because the numbers less than 5 were included in the cells of the fourfold tables.
There were 44 drug-disease combination cases (8 drugs) which showed significant associations when groups (A) and (C) were compared; 3 cases (3 drugs) out of the 44 also showed significant associations when (A) ' and (C) were compared.
In conclusion, it was suggested that such a data file of our hospital might be applied to PMS for ADRs. Some issues, however, have remained unsolved. For example, it is not certain that all of the disease names recorded in the data file are scientific ones because they were given for the health insurance billing purpose. Therefore, it must be ascertained in the near future whether or not our proposal can be effectively applied.
