Bringing functional status into a big data world: Validation of national Veterans Affairs functional status data

Background: The ability to perform basic daily activities (“functional status”) is key to older adults’ quality of life and strongly predicts health outcomes. However, data on functional status are seldom collected during routine clinical care in a way that makes them available for clinical use and research. Objectives: To validate functional status data that Veterans Affairs (VA) medical centers recently started collecting during routine clinical care, compared to the same data collected in a structured research setting. Design: Prospective validation study. Setting: Seven VA medical centers that collected complete data on 5 activities of daily living (ADLs) and 8 instrumental activities of daily living (IADLs) from older patients attending primary care appointments. Participants: Randomly selected patients aged 75 and older who had new ADL and IADL data collected during a primary care appointment (N = 252). We oversampled patients with ADL dependence and applied these sampling weights to our analyses. Measurements: Telephone-based interviews using a validated measure to assess the same 5 ADLs and 8 IADLs. Results: Mean age was 83 years, 96% were male, and 75% were white. Of 85 participants whom VA data identified as dependent in 1 or more ADLs, 74 (87%) reported being dependent by interview; of 167 whom VA data identified as independent in ADLs, 149 (89%) reported being independent. The sample-weighted sensitivity of the VA data for identifying ADL dependence was 45% (95% CI, 29%, 62%) compared to the reference standard, the specificity was 99% (95% CI, 99%, >99%), and the positive predictive value was 87% (95% CI, 79%, 93%). The weighted kappa statistic was 0.55 (95% CI, 0.41, 0.68) for the agreement between VA data and research-collected data in identifying ADL dependence. Conclusion: Overall agreement of VA functional status data with a reference standard was moderate, with fair sensitivity but high specificity and positive predictive value.