Comparing Performance of Comorbidity Indices in Predicting Functional Status, Health-Related Quality of Life, and Total Health Care Use in Older Adults With Back Pain

Sean D Rundell; Linda Resnik; Patrick J Heagerty; Amit Kumar; Jeffrey G Jarvik

doi:10.2519/jospt.2019.8764

Comparing Performance of Comorbidity Indices in Predicting Functional Status, Health-Related Quality of Life, and Total Health Care Use in Older Adults With Back Pain

J Orthop Sports Phys Ther. 2019 Jul 23:1-18. doi: 10.2519/jospt.2019.8764. Online ahead of print.

Authors

Sean D Rundell^{1

2}, Linda Resnik^{3

4}, Patrick J Heagerty^{2

5}, Amit Kumar⁶, Jeffrey G Jarvik^{2

7}

Affiliations

¹ Department of Rehabilitation Medicine, University of Washington, Seattle, WA.
² Comparative Effectiveness, Cost, and Outcomes Research Center; University of Washington, Seattle, WA.
³ Department of Health Services, Policy and Practice; Brown University, Providence, RI.
⁴ Research, Providence VA Medical Center, Providence, RI.
⁵ Department of Biostatistics, University of Washington, Seattle, WA.
⁶ Department of Physical Therapy, Northern Arizona University, Flagstaff, AZ.
⁷ Departments of Radiology, Health Services, and Neurological Surgery, University of Washington, Seattle, WA.

PMID: 31337267
DOI: 10.2519/jospt.2019.8764

Abstract

Study design: Secondary analysis of prospective cohort study.

Background: Common comorbidity indices were developed to predict mortality and may not be optimal for functional outcomes.

Objective: Determine how well the Functional Comorbidity Index (FCI) predicts outcomes in older adults with back pain versus the Quan/Charlson and Elixhauser comorbidity indices.

Methods: We included 5155 adults ≥65 years with new primary care visits for back pain. Comorbidity was measured using diagnosis codes 12 months prior to the new visit. Outcomes of functional limitation (Roland Morris Disability Questionnaire [RMDQ]), health-related quality-of-life (EQ5D), and total health care use (sum of Relative Value Units [RVUs]) were measured 12 months the new visit. We compared multivariable models containing preselected prognostic factors.

Results: Spearman's Correlation Coefficients among the indices were ≥0.70. Multivariable models for RMDQ had similar R2 and mean squared error (MSE) of prediction when using the FCI (R²=0.190; MSE= 6.19), Quan/Charlson (R²=0.184; MSE=6.20), or Elixhauser (R²=0.189; MSE=6.19). Multivariable models for EQ5D showed small differences in R² and MSE when using the FCI (R²=0.157; MSE= 0.163), Quan/Charlson (R²=0.148; MSE=0.164), or Elixhauser (R²=0.154; MSE=0.163). Multivariable models for health care use had similar Akaike's information criterion (AICs) when using the FCI (AIC=10.04), Quan/Charlson (AIC=10.04), or Elixhauser (AIC=10.01).

Conclusion: All indices performed similarly in predicting outcomes. There does not seem to be an advantage to using one index over another for older adults with back pain. There is still a need to develop better function-based risk adjustment models that improve prediction of functional outcomes versus standard comorbidity indices.

Level of evidence: Level 2 Prognostic Evidence. J Orthop Sports Phys Ther, Epub 23 Jul 2019. doi:10.2519/jospt.2019.8764.

Keywords: comorbidity measures; risk adjustment; spine.