Background: This study aimed to develop a multidimensional comorbidity index (MCI) that summarizes information from health insurance claims of patients with advanced ovarian cancer. We aimed to assess whether this index could mitigate confounding by baseline health status in a comparative effectiveness study of neoadjuvant chemotherapy (NACT) versus primary cytoreductive surgery (PDS). Methods: We utilized Medicare claims linked to Surveillance, Epidemiology, and End Result (SEER) registry data to identify patients diagnosed with stage IIIC and IV ovarian cancer between 2010 and 2015. The patient cohort was divided into development (2010-2014) and validation (2014) cohorts. To develop the MCI, we employed partial least squares (PLS) regression, a supervised machine learning algorithm, to extract latent factors that optimally captured the variation in 1-year mortality and health insurance claims in the 12 months before ovarian cancer diagnosis. We assessed the discrimination (c-index) and calibration of the MCI for 1-year mortality and compared its performance to the Charlson comorbidity index (CCI). Finally, we evaluated the MCI's ability to reduce confounding in a comparative effectiveness study comparing long-term all-cause mortality after NACT compared with PCS, accepting as the ground truth that NACT results in similar long-term survival as PDS. Results: The study included 4,760 patients in the development cohort and 941 in the validation cohort. The optimal MCI was composed of two latent variables that explained 17.1% of the variation in one-year mortality and 6.2% of the variation in claims history. In the validation cohort, the MCI demonstrated good discrimination for 1-year mortality (c-index: 0.72, 95% CI: 0.67-0.76), while the CCI had poor discrimination (c-index: 0.60, 95% CI: 0.58-0.61). Calibration plots showed agreement between predicted and observed 1-year mortality risk for the MCI. When comparing NACT with PCS, controlling for tumor characteristics, demographic factors, and the CCI, NACT was associated with a higher long-term hazard of death (HR: 1.14, 95% CI: 1.05-1.25). However, when using the MCI instead of the CCI, there was no longer evidence of a difference (HR: 1.03, 95% CI: 0.95-1.12). Conclusions: The MCI outperformed the conventional CCI in predicting 1-year mortality, and reducing confounding due to differences in baseline health status among patients with advanced ovarian cancer.