Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Donglin Li, MD, MPH and Long H. Ngo, PhD. Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, RO-102, Boston, MA 02215, 617 667 1340, dli@bidmc.harvard.edu

To estimate the likelihood of late stage cancer diagnosis comparing persons age <65 with and without Social Security Disability Insurance (SSDI) and Medicare (i.e., disability status), we used Surveillance Epidemiology and End Results (SEER) Program linked to Medicare claims for cases diagnosed from 1988 to 1999. We studied 47,704 incident cases of non-small cell lung cancer (NSCLC) diagnosed in persons aged 24-64.

We used multivariable ordinal polychotomous logistic regression to examine the association between late stage diagnosis and disability status after adjusting for sex, age at diagnosis, race/ethnicity, marital status, SEER tumor registry, and year of diagnosis. This model assumes proportional odds of the risk of having late stage diagnosis. However, this test was not informative because the power to detect non-proportionality approached 1 due to the very large sample size. We compared results using both ordinal and nominal models, and discuss the pros and cons of each method in terms of interpretation and computation.

In the ordinal model, we found that the adjusted odds ratio (aOR) for later stage diagnoses was 0.76, 95% CI=(0.72, 0.81), suggesting that persons with disabilities are less likely to have later stage diagnoses than others of the same age. In the multi-nominal model, we found that the estimated aOR of stage IV vs. I was 0.81 (0.77, 0.84), stage IV vs. II 0.89 (0.82, 0.96), stage IV vs. III 0.89 (0.86, 0.92).

Learning Objectives:

• Compare ordinal and nominal logistic regression modeling and discuss pros and cons of each method.
• Obtain some ideas of applying the statistical analysis strategies to epidemiologic study.

Keywords: Statistics, Medicare

Presenting author's disclosure statement:

Not Answered