Medications Prescription at Hospital Discharge in Patients with Validated Diagnosis of Dementia

Background: Health databases may be used to assess cases of dementia (D), among the main causes of morbidity in the elderly, if the data are accurate. Objectives: We present the preliminary results of a cohort study aimed at assessing a) validity of discharge diagnosis of D; b) medication prescribed at discharge; c) prescription redemption after discharge, through record-linkage with outpatient prescription database. Methods: Sources of Data: Friuli Venezia Giulia (FVG), Italy, Regional Health Database of Hospitalizations and hospital electronic medical records (HEMR). Study population: all records of discharge from the Udine University Hospital (FVG) 2012-2014 with International Classification of Diseases, 9th Revision Clinical Modification (ICD-9-CM) code for D (senile, presenile, vascular 290.0-290.43; degenerative 290.9; alcohol persistent 291.2; drug persistent 292.82; D in other diseases 294.1-294.8; frontotemporal 331.0-331.19; Lewy body 331.82; Creutzfeldt Jakob 046.1) in any position were selected from the Hospitalizations Database. The diagnosis was confirmed through HEMR review by trained Medical Doctors. Case confirmation required a written diagnosis of D and/or severe cognitive deficit. Reasons for not confirmation included wrong code, evidence of another disease. Statistical analysis: Positive Predictive Value (PPV), with Wilson 95% Confidence Interval (95%CI), as the ratio of confirmed to potential cases. Results: From 1 July to 31 December 2014, 445 hospitalizations with discharge code for D occurred, for 424 (94.4%) the information in HEMR was complete, in 404 of these the diagnosis was confirmed (PPV 96.2%; 95%CI 94.4-98.0). The most common diagnoses were the senile 290.0, (N=186, 46.0%), vascular 290.4X, (N=158, 39.1%) and presenile 290.1X, (N=26, 6.4%) D subtypes. Proton pump inhibitors (A02BC, N=210, 52.0%), Platelet aggregation inhibitors (B01AC, N=166, 41.1%) and Antipsychotics (N05A, N=141, 34.9%) were the most prescribed medications. Conclusions: Consistently with prior studies, codes for D showed high validity. Review of hospital charts is required when the information in HEMR is lacking