Background/Purpose: Maintaining physical function is crucially important in aging. Lower body function is often measured with the chair stand test (CST), also known as the 5 times sit-to-stand test, which is the time it takes to rise from a seated to a standing position 5 times without using one’s arms. Such performance tests may face analytic challenges due to the skewness of completion times and potentially informative missing values from participants who are either unable or refuse to complete the test or decline to perform the test entirely. Our purpose was to compare different analytic strategies using CST as an outcome to assess which approach provided the strongest cross-sectional association with the self-reported WOMAC function scale.

Methods: We included participants were from the Multicenter Osteoarthritis Study (MOST), a NIH-funded longitudinal cohort of individuals with or at risk of knee osteoarthritis (OA). We compared participants with and without missing CST by t-test, chi-square test, and logistic regression.  We conducted the following 4 analytic approaches to address missing CST data: 1) Linear regression of CST times, with and without multiple imputation to include participants missing CST; 2) Linear regression of the natural logarithm of CST times to address skewness, with and without multiple imputation; 3) Non-missing CST times were split into quartiles 1 (Slowest) to 4 (Fastest) and participants with missing values placed into group 0; groups 0-4 were analyzed using proportional odds logistic regression; 4) CST values were ranked in ascending order with missing values given ranks above those for observed time; the ranks were then analyzed using m-estimation with rank-based regression. All regression models included WOMAC function, age, sex, BMI, WOMAC pain, depressive symptoms, and presence of radiographic knee OA (ROA). The strength of association with WOMAC function was assessed by the regression beta slope estimate divided by its standard error (Z statistic).

Results: 1931 participants were included (mean age 61.5, 59% female, mean BMI 30.5, 60% with ROA). 4% of participants had missing values for CST, and among those who performed the CST, there was a pronounced skewness (Figure 1) with mean of 11.8 and a range from 4 to 41 seconds. Higher (worse) values of CST were significantly associated with higher age, depressive symptoms, BMI, WOMAC pain, WOMAC function and presence of ROA. Age and WOMAC function retained statistical significance for association with worse CST in multivariable analysis. The association between WOMAC function and CST was statistically significant in each of the 4 approaches’ regression models (all p< 0.0001), with the strongest result using rank-based regression of CST (approach #4).  

Conclusion: Missing values for the CST were rare in MOST participants (~4%) and all regression models showed strong positive and significant associations between WOMAC function and CST. However, the analytic approach that treated individuals with missing CST values as the worst ranking performers offered the strongest association with WOMAC function. This approach may be useful in other scenarios of missing or incomplete performance of functional tests.

Abbreviations: MI – multiple imputation; LN(CST) – natural logarithm of chair stand time

All values adjusted for age, sex, BMI, WOMAC pain, depressive symptoms, and presence of radiographic osteoarthritis.

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/optimizing-analyses-of-chair-stand-test-outcome-data-in-the-multicenter-osteoarthritis-study-exploration-of-approaches-to-handling-missing-and-or-skewed-data-with-performance-based-function-measures/