Objective : To examine whether drug treatment for hypertension prevents the first occurrence of cerebral hemorrhage, a population-based case-control study was conducted in Hirara-city, Okinawa Prefecture, Japan.
Methods : Cases were 36 hypertensive persons, 45 to 84 years of age, who had experienced cerebral hemorrhage from 1991 to 1994. Twenty of them were ascertained to be hypertensive by residential mass health examinations before the occurrence of cerebral hemorrhage, and other cases were ascertained by stroke register. Controls were 158 hypertensive persons unaffected by stroke and other life-threatening diseases, who were matched with their corresponding case for sex, age and mean blood pressure. All controls were ascertained as hypertensive by health examinations in 1991. Data on antihypertensive treatment, during the two years before the occurrence for cases and between April 1992 and March 1994 for controls, were collected by claims for medical care cost and medical records. Conditional logistic regression was employed for matched analyses.
Results : Regarding treatment, 16 (44.4%) cases and 36 (23.5%) controls did not receive antihypertensive drug treatment, and 13 (36.1%) cases and 30 (19.6%) controls had interrupted treatment. Compared with continuously treated persons, untreated persons and interrupted persons had odds ratios of 6.27 (95% confidence interval : 2.21 to 17.8) and 4.94 (1.79 to 13.6) for cerebral hemorrhage, respectively. Cases were prescribed antihypertensive drugs for fewer months than controls. By subgroup analyses limited to the cases ascertained as hypertensive by mass health examinations and their matched controls, the same relation was observed.
Conclusion : Among hypertensive persons, non-treatment and interruption of antihypertensive drug treatment were associated with a higher risk of cerebral hemorrhage. Long-term continuous treatment with antihypertensive drug (s) prevents the first occurrence of cerebral hemorrhage.

View full abstract

Objective : To evaluate the necessity to complement the information obtained from pharmacists in the community pharmacy by that from the hospital pharmacy in Prescription-Event Monitoring in Japan (J-PEM) by using data in a J-PEM pilot study.
Methods : For each patient, two questionnaires were sent to the prescribing doctor and the pharmacist who registered the patient ID code in the pilot study. If the patient ID code was registered by the pharmacist in the community pharmacy and if a pharmacist inside the hospital where the prescription was issued was willing to co-operate, a third questionnaire for the same patient was sent to the pharmacist in the hospital pharmacy. The information given by pharmacists was analyzed for 150 pairs of questionnaires (on 150 patients) sent back from pharmacists in both community and hospital pharmacies. The questions in the questionnaire were categorized into [1] those on drugs used by patients (concurrent drugs, daily dose of the drug monitored, and compliance), [2] those on events which the patient had experienced after the prescription of the drug monitored, [3] those on patients (the first date of prescription, reason of prescribing the drug monitored, initial date when the disease developed, underlying diseases or complications and whether and when the patient was lost to follow-up). The questionnaires were examined to determine whether the answer was given to each question. When the answer was given, its quality and quantity were then assessed. The answer to each question given by the pharmacist in the community pharmacy (C) and that by the pharmacist in the hospital pharmacy (H) were compared by the McNemar test after the pairs of answers were classified into the following categories : [1] C is better than H, [2] H is better than C, [3] C and H are similar to each other, and [4] impossible to classify. The difference was considered to be significant where p<0.05.
Results and conclusion : For the initial date when the disease developed and 'underlying diseases or complications', H was significantly better than C. However, for concurrent drugs, compliance and events, C was significantly better than H. Otherwise, the difference was not statistically significant. Being compatible with the superiority of C over H in regard to concurrent drugs and events, the fraction of patients lost to follow-up during the observation period was small not only in H but also in C. This observation may be associated with the fact that almost all prescriptions were issued by a single hospital in more than 60% of community pharmacies in the pilot study, and most patients identified in the study were probably a regular visitor to one of such community pharmacies. The most important information to be provided by the pharmacists in J-PEM is that on events and drugs used by patients. It is thought to be not necessary to complement the information obtained from the community pharmacy by that from the hospital pharmacy.

View full abstract

As the Japanese health insurance system is reimbursed on a fee-for-service basis, the claims data are useful sources for quantitative analysis on practice and prescription patterns as well as the health care resource use. However, the claims database has limitations such as incomplete description of the patients' condition and severity, and the lack of prescription data for low-priced drugs.
For use in pharmacoepidemiologic studies, patient information from medical charts, prescription forms, and/or hospital information systems may need to be included to compensate for the limitations of the claims data.

View full abstract

Objective : To elucidate the availability and applicability of the information contained in health insurance claims to pharmacoepidemiological studies with particular reference to the computerization of insurance claims submitted by dispensing pharmacies.
Data Sources : Outpatient insurance claims of two health insurance societies based in the Kanto region. The survey, funded by the Federation of Health Insurance Societies, was undertaken as part of a demonstration project to develop methodologies for insurers to evaluate the clinical performance of providers.
Study Selection : Disease-specific, provider-specific and clinical procedure specific comparison of the per-claim cost. All claims from pharmacies were matched with the prescribing hospitals or clinics to capture the entire medication.
Data Extraction : All claims were reviewed by trained reviewers and procedure specific costs were classified into five disease categories to obtain the best estimate of disease-specific cost.
Results : Inter-provider variations were assessed based on overlapping of the overall average and the confidence interval of the doubled standard error after adjusting for age. Contrary to popular belief, academic medical centers showed significantly lower medication cost than clinics for the treatment of acute upper respiratory infection. Prescription pattern to favor high cost antibiotics accounted for higher per-claim medication cost of clinics with concentration of high cost claims.
Conclusion : Based on the authors' experience and methodologies developed, computerization of insurance claims will achieve consistency, efficiency and timelines which are of utmost importance for pharmacoepidemiological studies. On the other hand, insurers may preempt the claims review and reimbursement organizations by appealing to the pharmacy claims through electronically matching the prescribing claims. The nature of evidence sought by both pharmacoepidemiological researchers and insurers, whether it be for the purpose of post-marketing survey or cost cutting, would eventually converge.

View full abstract