Why so Fast? A Focus on Reasons for an Increase in Arthritis-Attributable Activity Limitation Trends, 2002-2017

Abstract

Background/Purpose: “Arthritis-attributable activity limitation” (AAAL) is linked to many potentially modifiable characteristics (e.g., work disability, physical inactivity, obesity). By 2015, prevalence of AAAL among adults =18 with arthritis had increased by almost 20% compared with that reported in 2002, and the observed annualized prevalence of AAAL in 2010 had outpaced projections for 2020 (22.7 vs. 22.1 million). Our objective was to investigate the extent to which trends in AAAL prevalence can be explained by changes in characteristics associated with AAAL among U.S. adults with arthritis.

Methods: Cross-Sectional data were obtained for participants =18 years from years 2002-2017 of the National Health Interview Survey (NHIS, average unweighted sample n=30,076), an ongoing, multistage probability survey designed to be nationally representative of the civilian, non-institutionalized U.S. population. Analyses were restricted to those with self-reported doctor-diagnosed arthritis. AAAL was defined as “yes” to “Are you now limited in any way in any of your usual activities because of arthritis or joint symptoms?” Demographic, health, function, and health care access measures were assessed across 30 variables.

We used unconditional logistic regression to estimate prevalence ratios (PR) with 95% confidence intervals (CI) to examine the associations between a priori selected variables and AAAL, modeled the prevalence of AAAL (dependent variable), and estimated a prevalence ratio (PR) for each independent variable in a model containing only that variable and (continuous) calendar year. Next, for 7 variables significantly associated with AAAL (identified with the overall Wald test of model coefficients) we estimated 2 PRs for AAAL trend: 1) average annual increase in AAAL, and 2) overall AAAL increase, (2017 prevalence less 2002 prevalence). Finally, we calculated a series of multivariable-adjusted (MV) models to test changes in the estimated effect of calendar year on AAAL post-adjustment, i.e., addition of a covariate which reduced the significance of the existing trend identified a variable with a significant role in the observed trend relationship.

Results: Between 2002 and 2017, prevalence of AAAL among adults with arthritis significantly increased (37.6% (95% CI=36.1-39.1) to 42.9% (41.4-44.4), p-value < 0.0001 test for trend). Initial model PR=1.11 (1.08-1.15). Only 4 variables reduced the PR, driven by: comorbidities (increased number), employment status (increased retired and not in labor force), body mass index (increased overweight), and race/ethnicity (increased non-Hispanic White). The final MV PR was 1.05 (1.02-1.08). The temporal trend in AAAL is largely explained by the temporal trends of the characteristics accounted for in this analysis (Figure).

Conclusion: In addition to identifying specific variables which explain some of the trend in increased AAAL between 2002-2017, we rejected many others as contributors. Still, unknown/unmeasured variables (e.g., zip code, urbanization, neighborhood) may account for some of the trend. Meanwhile, interventions can be directed at the subgroups identified here to reduce AAAL impact.