OBJECTIVES: Benzodiazepine (BZD) treatment of prolonged periods and higher dosage is not recommended for the elderly. Because of free choice of physicians and the absence of gatekeepers in Taiwan, this study aimed to identify prescription patterns and expenditure for long-term BZD treatment in elderly outpatients, and examine factors associated with the prescription overlap. METHODS: This study used a sampling database from the Bureau of National Health Insurance. Subjects were included if they were aged 65 or over and had received any BZD prescription in 2002. Received BZDs were divided into prescribed and overlapped quantities in terms of defined daily doses (DDDs) and expenditure. Single long-term prescriber was defined as receiving more than 180 supply days from one department. The logistic model and multivariate regression models were used to evaluate associations of factors with the prescription overlap and quantities of overlap, respectively. RESULTS: In 2002, 1705 out of 7561 elderly individuals had received BZDs with more than 180 prescription-covered days. Prescription overlap represented 10% of both their DDDs and expenditure. There were 107 subjects who had no single long-term prescriber and 362 subjects with more than 30 prescription-overlap days. Men, mental illness(es), comorbidities, multiple prescribers with or without a single long-term prescriber, and higher prescribed dose were associated with an increased likelihood of prescription overlap. Increased age, depression, a single long-term prescriber with or without multiple prescribers, and increased prescribed dose were correlated with increased overlapped DDD and expenditure. CONCLUSIONS: Eighteen percent of the elderly with long-term BZD treatment were not under the authority of a single prescriber, and the prescription overlap was highly concentrated in 21% of them. Multiple prescribers increased the probability of prescription overlap. Individuals with a single long-term prescriber and multiple prescribers had a higher risk of prescription overlap, and higher overlapped BZD DDDs and expenditure.