Dementia recognition, diagnosis, and treatment in the UK, 1997–2017: a change-point analysis

Abstract

BackgroundOver the past decade, numerous national dementia policies and incentive schemes have been introduced in the UK. However, the effects of these initiatives are challenging to assess because of their overlapping nature. We aimed to assess the potential effects of policies on dementia diagnosis, monitoring, and treatment. MethodsWe used UK primary care electronic health record data from the Clinical Practice Research Datalink (CPRD), available through the CALIBER resource, combined with data-driven change-point analyses. We estimated the monthly incidence of dementia diagnoses, dementia-monitoring events, or dementia-related drug prescriptions among patients older than 65 years in CPRD from Jan 1, 1997, to Jan 1, 2017. We compared hypothesis-driven and data-driven approaches to estimate the locations of change-points: an interrupted time-series regression with a step-and-level change impact model versus a continuous model; and a Pruned Exact Linear Time (PELT), a search method based on an optimal partitioning algorithm. FindingsWe identified 356 234 patients that matched the study inclusion criteria and analysed 19 policies or schemes. Across Alzheimer's, vascular dementia, cognitive and memory impairment, and dementia tests, the interrupted time-series approach consistently identified change-points at the introduction of the Quality Outcomes Framework (April, 2006), the Commissioning for Quality and Innovation, the first Prime Minister's Challenge, and the Dementia Enhanced Service (approximately March, 2012). Although the Quality Outcomes Framework change-point marked an expected rise in the incidence of these dementia-related events, the change-point in March, 2012, marked a sharp decrease in incidence. No other change-point locations were concordant among the different models used. InterpretationElectronic health records allow us to explore complex policy interventions, over long periods, at a national scale. Our findings, however, showed that rates of diagnostic, monitoring, or treatment events associated with dementia did not always align to national interventions and the location of change-points varied considerably by method. The temporary potential effects of policies are likely to be minimal compared with the intensifying decrease in dementia incidence. Further research in data-driven methods in the context of overlapping policies is required. FundingNone.