Reliability of coded data to identify earliest indications of cognitive decline, cognitive evaluation and Alzheimer's disease diagnosis: a pilot study in England
BMJ Open. 2018 Mar 22;8(3):e019684
Evaluate the reliability of using diagnosis codes and prescription data to identify the timing of symptomatic onset, cognitiveassessment and diagnosis of Alzheimer's disease (AD) among patients diagnosed with AD.
This was a retrospective cohort study using the UK Clinical Practice Research Datalink (CPRD). The study cohort consisted of a random sample of 50 patients with first AD diagnosis in 2010-2013. Additionally, patients were required to have a valid text-field code and a hospital episode or a referral in the 3 years before the first AD diagnosis. The earliest indications of cognitive impairment, cognitiveassessment and AD diagnosis were identified using two approaches: (1) using an algorithm based on diagnostic codes and prescription drug information and (2) using information compiled from manual review of both text-based and coded data. The reliability of the code-based algorithm for identifying the earliest dates of the three measures described earlier was evaluated relative to the comprehensive second approach. Additionally, common cognitive assessments (with and without results) were described for both approaches.
The two approaches identified the same first dates of cognitive symptoms in 33 (66%) of the 50 patients, first cognitiveassessment in 29 (58%) patients and first AD diagnosis in 43 (86%) patients. Allowing for the dates from the two approaches to be within 30 days, the code-based algorithm's success rates increased to 74%, 70% and 94%, respectively. Mini-Mental State Examination was the most commonly observed cognitive assessment in both approaches; however, of the 53 tests performed, only 19 results were observed in the coded data.
The code-based algorithm shows promise for identifying the first AD diagnosis. However, the reliability of using coded datato identify earliest indications of cognitive impairment and cognitive assessments is questionable. Additionally, CPRD is not a recommended data source to identify results of cognitive assessments.