Can clinical practice indicator relating to long-acting benzodiazepine use in the elderly be easily generated in a hospital setting?

Background: To improve the appropriate use of long-acting benzodiazepine (la.bzd) prescriptions in the elderly, the Haute Autorité de Santé (HAS) has developed clinical practice indicators (CPI). The alert indicator (AI) evaluates the prevalence of la.bzd prescription among older people. The mastering indicator (MI) corresponds to the prevalence of elderly with a justified, i.e., appropriate, la.bzd prescription among all the elderly with la.bzd prescriptions.

Objective: The objective of this study was to test the feasibility of routine generation of CPI regarding la.bzd prescriptions among the elderly in the hospital setting.

Design: This was a retrospective study.

Setting: The study was conducted in two university hospitals located in Paris.

Subject: Eligible cases were patients aged 65 years and older who were hospitalized in acute care units from January to June 2014.

Method: The AI calculation was based on information extracted from medical databases from these hospitals. The appropriateness of la.bzd prescription was assessed by a physician and a pharmacist and was based on review of computerized patient records and prescriptions, using an ad hoc algorithm. The MI was then calculated. Variation in the level of indicators was explored according to the characteristics of patients and of their hospitalization using chi² test. Factors associated with a potentially inappropriate prescription (PIP) of la.bzd were studied using univariate and multivariate logistic regression.

Result: Among the 4167 patients included in the study, 362 had la.bzd prescriptions, i.e., the AI was 9%. Prescriptions were found to be appropriate for 83 patients, i.e., the MI was 23% and PIP was 77%. The MI varied between 13 and 31% according to characteristics of patients and of hospitalization. In multivariate analysis, factors associated with PIP were age, number of comorbidities, type of care unit, and concurrent prescription of a neuroleptic or hypnotic.

Conclusion: Generation of the AI was routinely possible but only for acute care units with computerized prescriptions, corresponding to 78% of patients. Production of the MI has required medical record review for all patients with a la.bzd prescription and cannot be automated. However, difficulties in generating the MI have identified areas for significant improvement. Moreover, strategies to improve the care of older people with a la.bzd prescription could be targeted using characteristics of patients and of hospitalization associated with PIP. The future deployment of a single electronic medical record in all care departments would make it easier to mine the data and make possible automated production of CPI.