(P-005) Multimorbidity Measures from Health Administrative Data: A Systematic Review

Background: None of the previous systematic reviews have examined administrative data-specific issues for multimorbidity measures and compare their quality assessment.

Objectives: We aimed to identify and characterize adult population-based measures of multimorbidity relying on health administrative data.

Methods: We performed a narrative systematic review of studies describing development and/or validation of population-based multimorbidity measures relying on health administrative databases. We characterized multimorbidity measures based on diseases identified from in-hospital or outpatient data using the International Classification of Disease (ICD) codes. We compared number of diseases considered, level of categorisation of diseases (list of ICD codes considered), process of data extraction (case definition) and validation. We assessed their quality using eight robustness criteria.

Results: We identified 22 multimorbidity measures. The number of diseases in multimorbidity measures varied from 5 to 83 (median=20), and all measures but three included both physical and mental conditions. The majority of multimorbidity measures (14/22) were validated, mostly by assessing association with the risk of mortality, but no study validated against a gold standard. In most studies (18/22), diseases were identified from diagnosis codes extracted from both inpatient and outpatient data. However, heterogeneity in level of categorisation of diseases was noted. Six multimorbidity measures met at least six of the eight robustness criteria that we assessed.

Conclusions: Less than a third of multimorbidity measures presented high robust quality. The heterogeneity between measures limited appropriate comparisons between prevalence obtained from them. Agreement within the scientific community on a common multimorbidity measure validated against a gold standard may improve accuracy of multimorbidity assessment and facilitate comparison over time and regions.