Measuring dementia incidence within a cohort of 267,153 older Australians using routinely collected linked administrative data

Abstract

To estimate dementia incidence rates using Australian administrative datasets and compare the characteristics of people identified with dementia across different datasets. This data linkage study used a cohort of 267,153 from the Australian 45 and Up Study. Participants completed a survey in 2006-2009 and subsequent dementia was identified through pharmaceutical claims, hospitalisations, aged care eligibility assessments, care needs at residential aged care entry and death certificates. Agespecific, and age-standardised incidence rates, incidence rate ratios and survival from first dementia diagnosis were estimated. Estimated age-standardised dementia incidence rates using all linked datasets was 16.8 cases per 1000 person years for people aged 65+. Comparing incidence rates to the global published rates suggested 77% of cases were identified but this varied by age with highest coverage among those aged 80-84 years (92%). Incidence rate ratios were inconsistent across datasets for: sex, socioeconomic disadvantage, size of support network, marital status, functional limitations and diabetes. Median survival from first dementia diagnosis ranged from 1.80 years in the care needs dataset to 3.74 years in the pharmaceutical claims dataset. Characteristics of people identified with dementia in different administrative datasets reflect the factors that drive interaction with specific services; this may introduce bias in observational studies using a single data-source to identify dementia. Routinely-collected linked administrative data are increasingly being used to monitor endpoints in observational studies and clinical trials 1. Dementia prevention studies may benefit from this approach due to the long time-frame required to study risk factors in this population 2. Within a research setting, maintaining contact with people in older age groups, particularly as they develop cognitive impairment or dementia is often not viable. Administrative data have the potential to increase power within studies by improving completeness of follow-up, and to reduce bias by avoiding the issue of differential drop-out due to cognitive impairment 3. However, there are also potential limitations to using administrative databases for detecting dementia. Administrative health data are those generated routinely via a person's interaction with the health system. They may include records of hospitalisation, physician visits, entry to long term care or dispensing of pharmaceuticals. They are generally collected with payment rather than research in mind, but there is often a degree of data curation that increases accuracy or adds value such as through coding of diseases and medical conditions. As such, they can be a valuable resource for research 4. When using such data to measure endpoints in trials or cohort studies it is important to consider both whether the cases identified have the disease of interest (usually indicated by high positive predictive value (PPV)) and the proportion of total cases that are detected (sensitivity). A recent syst...

Association between the severity of hearing loss and the risk of dementia within the 2010–2017 national insurance service survey in South Korea

“…In Czech Republic, data of patients hospitalized for dementia are provided by the national register of hospitalized patients in all Czech hospitals 13 . In New South Wales (NSW), Australia, recruitment to the 45 and up study involves patients diagnosed with dementia from 2006 to 2009 through random sampling from the Department of Human Services enrollment database (Australia's national universal insurance provider) 14 . Table 1.…”

“…ND, non-disability; HD, hearing disability. www.nature.com/scientificreports/ Although the ICD-10 codes for dementia and additional details for defining dementia, such as medications for treating dementia, hospital admissions, aged care assessments, and underlying or contributing cause of death, are slightly different for each study, our definition for dementia is almost similar to that of previous studies [12][13][14][15][16] . We also used nationwide data from the National Disability Registry (NDR), a social registry, which provides patients with severe-profound HL or deafness with disability benefits such as financial assistance to acquire hearing aids or cochlear implants.…”

“…In Czech Republic, data of patients hospitalized for dementia are provided by the national register of hospitalized patients in all Czech hospitals 13 . In New South Wales (NSW), Australia, recruitment to the 45 and up study involves patients diagnosed with dementia from 2006 to 2009 through random sampling from the Department of Human Services enrollment database (Australia’s national universal insurance provider) 14 . Although the ICD-10 codes for dementia and additional details for defining dementia, such as medications for treating dementia, hospital admissions, aged care assessments, and underlying or contributing cause of death, are slightly different for each study, our definition for dementia is almost similar to that of previous studies 12 – 16 .…”

Dementia case ascertainment using aged care assessment data

“…Two prior cohort studies have assessed the value of aged care assessments for estimating dementia prevalence and incidence, but the proportion of the whole population of people with dementia that can be captured using these data has not yet been examined. 8,9 As such, this study aimed to establish the extent to which aged care assessment data can capture the total national population of people with dementia.…”

“…Using multiple data sources is important given that each source will produce a unique population, based on differences between those who interact with each service. 9 There are important limitations to using aged care assessment data to capture the full population of people with dementia. This includes under-and biased ascertainment, because only those who access aged care and those with a formal diagnosis of dementia are included.…”