Benzodiazepine use in the real world of psychiatric practice: low-dose, long-term drug taking and low rates of treatment discontinuation.
Veronese A., Garatti M., Cipriani A., Barbui C.
OBJECTIVE: The present study was designed to (1) estimate the frequency of benzodiazepine use in psychiatric practice, (2) investigate factors associated with use, (3) establish whether a relationship exists between benzodiazepine dose and length of use, and (4) investigate factors associated with time to discontinuation. METHODS: This study was conducted in South Verona, Italy. All individuals who were exposed to benzodiazepines during 2005 were extracted from the local Psychiatric Case Register, and the longitudinal history of benzodiazepine exposure was retrospectively described. RESULTS: In 2005, a total of 1,771 individuals were in contact with at least one of the psychiatric facilities of the South Verona catchment area. Of these, 535 were benzodiazepine users, yielding a frequency of use of 30.2% [95% confidence intervals (CI) 28.0, 32.4]. In multivariate logistic regression analysis, lower level of education, diagnosis of affective illness, longer length of illness and higher service use were significantly associated with benzodiazepine exposure. An increase in dosages over time to maintain the drug's effectiveness was not evident from the analysis of the relationship between daily dose and length of therapy. A total of 17.3% (93/535) of patients exposed to benzodiazepines discontinued treatment. Cox regression analysis revealed that age and length of illness were negatively associated with the probability of discontinuing therapy, while the concomitant use of antipsychotics and mood stabilisers was positively associated with discontinuing therapy. CONCLUSION: The finding that in the great majority of psychiatric patients, low doses of benzodiazepines are routinely prescribed on a long-term basis suggests that, in this specific setting of care, treatment recommendations stating that use should be short term may not be applicable.