[Utilising statutory health insurance data to evaluate pharmaceutical interventions in secondary care - a pilot study].

Background: Intensive pharmaceutical and medical care can lead to fewer drug-related problems (DRPs) in hospitals. Currently available methods to track drug changes after transition from inpatient to outpatient care are susceptible to systemic bias. Therefore we analysed the feasibility of a data linkage between prescription data extracted from hospital medical records and claims data from a health insurance company.
Methods: At six Saxonian hospitals, patients with a written informed consent were consecutively assigned to a control (CG) or intervention group (IG) depending on the time of admission. Clinical pharmacists documented predetermined DRPs and prescribed medication on the day of hospital admission and in the discharge letter. In case of DRP (IG) or potentially life-threatening DRPs (CG), drug changes were recommended to the hospital physician. These data were patient-individually linked to claims data from a health insurance company comprising a period of six months before and six months after hospitalisation (data linkage). We analysed data consistency within the data linkage and the post-hospital prevalence of DRPs identified in the hospital setting.
Results: We enrolled 532 patients (CG/IG 280/252). The data linkage was feasible for 97.0% (CG) and 96.6% (IG) of the patients, respectively. A total of 318 DRPs (CG/IG 176/142) was detected. Because of restrictions in the reimbursement of drugs in the outpatient setting, 22 (12.5%, CG) and 13 (9.2%, IG) DRPs were not analysable. Insurance claims data during a 6-month follow-up showed no statistically significant difference between the CG (without intervention) and in the IG (with intervention) with respect to DRPs (43.4% vs 38.1%; p = 0.472).
Conclusions: The linkage of inpatient and outpatient data was feasible for the majority of enrolled patients. Compared to similar studies, the risk for systemic bias decreased because fewer patients were lost to follow-up. Within this feasibility study the expected difference between IG and CG could not be demonstrated statistically.
(Copyright © 2017. Published by Elsevier GmbH.)