Personalized risk assessment of new onset depression in long-term cancer survivors.

e13798 Background: Advances in cancer detection and treatment have improved long-term survival. It is critical to understand the impact that experiencing cancer has on psychological health, given that the prevalence of depression in long-term survivors is higher than in the general population. Our aim was to develop a risk prediction model to identify long-term (5-year) survivors at high risk of developing depression. Methods: We conducted a retrospective study of the SEER-Medicare database beneficiaries with localized breast, prostate, colon, or rectal cancer diagnosed between 2003-2011 who had survived at least 5 years since diagnosis. We required Part D coverage in the breast cohort to account for adjuvant endocrine therapy. Our primary outcome was the cumulative incidence of depression between 5-10 years after diagnosis, which was defined using a claims-based algorithm as a single or recurrent episode of major depressive disorder, recent episode of bipolar I disorder, or dysthymic disorder. For each cancer site, we constructed restricted mean survival time models, a non-parametric method that yields time ratios; a time ratio (TR) < 1 indicates shorter mean time to event (TTE). Results: Our cohort consisted of 80,579 survivors. The cumulative incidence of depression occurring between 5-10 years after diagnosis was 16.9% for breast cancer, 7.8% for prostate cancer, 10.6% for colon cancer, and 8.6% for rectal cancer. The cumulative incidence of new onset depression increased over time and was 7.2% vs. 13.9% among survivors diagnosed before and after 2008, respectively. Compared with patients age < 70 at diagnosis, patients aged 75 and older exhibited 5-10% shorter mean TTE across breast, prostate, and colon cancer. There was no association between Non-Hispanic (NH) Black race or Hispanic ethnicity and depression compared with NH White survivors. Asian survivors were at lower risk of developing depression. Disease specific factors, including stage, grade, hormone receptor status (breast), or PSA (prostate), did not predict depression in long-term survivors. Receipt of treatment in the 3-5 years after initial diagnosis was associated with moderately greater risk for all survivors, with the shortest mean TTE among breast cancer survivors (TR=0.91, 95% CI=0.85-0.97). Across all cohorts, the greatest predictor of depression was a prior history of depression. The incidence of depression in long-term survivors was 42% in patients with a prior history of depression vs. 8% among survivors without a prior history of depression. Conclusions: Diagnoses of depression increased in cancer patients since 2008 and were associated with prior history of depression, advanced age, census tract poverty, and continued treatment 3 years after diagnosis. Treatment type and disease specific factors were not strong predictors of depression.