Aitken Zoe, Walmsley Sarah, M Bishop Glenda, Badji Samia, Fortune Nicola
Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, VIC, 3010, Australia.
Centre for Health Economics, Monash Business School, Monash University, Melbourne, VIC, 3145, Australia.
Popul Health Metr. 2025 Jun 6;23(1):22. doi: 10.1186/s12963-025-00386-w.
In this scoping review, we aimed to examine evidence on methods used to construct disability indicators in linked administrative datasets and describe the approaches used to assess the validity of the indicators.
Medline (Ovid) and Embase (Ovid) were searched for studies published between January 2010 and June 2023. Original, peer-reviewed studies that aimed to construct a disability indicator using linked administrative data sources were included. Studies identifying any types of disability were included, but not those which defined the target population in terms of specific health conditions. We produced a narrative synthesis of findings related to disability indicator construction methods and validation approaches.
Thirty-six relevant studies were included, with 30 of those identifying a cohort of people with intellectual and/or developmental disability. Health data sources were most commonly used for indicator construction, with 33 of the studies using at least one health data source. Disability and education sector data sources were also commonly used. Diagnostic codes were used for disability identification in 34 of the 36 studies; 16 used diagnostic codes alone and 18 used diagnostic codes along with other information. A subgroup of 19 studies had a primary aim to create a disability cohort or estimate disability prevalence. Thirteen of these 19 studies compared their estimated prevalence rates with previously published estimates. Only five studies conducted testing to investigate the extent to which their derived disability indicator captured the intended target population.
We found a paucity of evidence on methods for identifying a target population of people with diverse disabilities. In the existing literature, diagnostic information is relied upon heavily for disability identification, likely due to a lack of other types of disability-relevant information in administrative data sources. Use of derived disability indicators within linked data holds potential to advance research regarding people with disability. It is crucial, however, to conduct and report validation testing to understand the strengths and limitations of the indicators and inform their use for specific purposes.
在本范围综述中，我们旨在研究在关联行政数据集中构建残疾指标所使用方法的证据，并描述用于评估这些指标有效性的方法。
检索了Medline（Ovid）和Embase（Ovid）中2010年1月至2023年6月发表的研究。纳入旨在使用关联行政数据源构建残疾指标的原创性、同行评审研究。纳入识别任何类型残疾的研究，但不包括那些根据特定健康状况定义目标人群的研究。我们对与残疾指标构建方法和验证方法相关的研究结果进行了叙述性综合。
纳入了36项相关研究，其中30项确定了一组智力和/或发育残疾人群。健康数据源最常用于指标构建，33项研究使用了至少一个健康数据源。残疾和教育部门数据源也常用。36项研究中有34项使用诊断代码进行残疾识别；16项仅使用诊断代码，18项将诊断代码与其他信息一起使用。19项研究的一个子组主要目的是创建残疾队列或估计残疾患病率。这19项研究中有13项将其估计患病率与先前发表的估计值进行了比较。只有5项研究进行了测试，以调查其得出的残疾指标捕获预期目标人群的程度。
我们发现关于识别不同残疾目标人群方法的证据不足。在现有文献中，残疾识别严重依赖诊断信息可能是由于行政数据源中缺乏其他类型与残疾相关的信息。在关联数据中使用派生的残疾指标有可能推进关于残疾人研究。然而，进行并报告验证测试以了解指标的优势和局限性，并为其特定用途提供信息至关重要
