Do hospital electronic medical records reliably register pre-admission medications in patients with dementia?

Background: Pre-admission medication list was reported to be inaccurate in up to 60% of inpatients. Inaccuracies concur to inappropriate prescribing, medication discontinuation, failure in recognizing adverse events and have been associated with adverse outcomes. Accuracy is challenging in patients with dementia at risk of severe adverse events, frequently in polypharmacy and hospitalized. Objectives: To estimate the percentage of discrepancy and the agreement between hospital electronic medical records (EMRs) pre-admission medication list and dispensing data in patients with dementia. Methods: Source of information: Hospital services and outpatient prescription databases, hospital EMRs; Study design: retrospective cohort study; Study population: all patients hospitalized at the Udine University Hospital, Italy, from 01.01.2012 to 31.12.2014 with primary or secondary ICD-9-CM discharge code for dementia and continuous enrolment for ≥1 year before admission; Data collection: for each hospitalization (a) the EMRs pre-admission medication list and (b) all prescriptions dispensed within 3 months prior to the date of admission through record linkage with the prescription database. An omission was defined as any dispensed medication not registered in EMR; an addition as any not dispensed medication registered in EMR. Statistical analysis: we calculated: (a) percentage of omissions and additions; (b) Kappa coefficient and prevalence and bias-adjusted Kappa (PABAK); The analysis was performed with SAS© software 9.3 (SAS, Cary, NC, USA). The protocol was approved by the Friuli Venezia Giulia regional Ethics Committee. Results: Among 2,777 (89.5%) of 3,104 hospitalizations (exclusions: hospitalizations with EMR unavailable, record linkage unsuccessful, of non-residents), 58.8% had ≥1 medication registered in EMR and 84.7% ≥1 pre-admission dispensing, Kappa 0.10 and PABAK 0.22. In 68.2% (65.9% excluding vitamins, minerals, topicals) ≥ 1 omission occurred and in 44.5% (43.5%) ≥1 addition. Percentage of omissions was 69.1% in respiratory medications (ATC class R), 41.7% in cardiovascular (C) and 42.9% in nervous system (N); of additions 42.9%, 20.1% and 34.6%, respectively. Omissions of anti-dementia agents occurred in 41.5% and additions in 48.2% of hospitalizations. Conclusions: Discrepancies were common, particularly omissions. EMR list has limited utility as a unique source of information on pre-admission medication use.