THE NOVEL GENERATION AND VALIDATION OF SURVIVAL CURVES IN ONCOLOGY UTILIZING ELECTRONIC MEDICAL RECORDS LINKED TO POINT OF SERVICE CLAIMS DATA
Author(s)
Wade RL1, Korytowsky B2, Singh P2, Dev P1, Bobiak S2, Cariola P1
1QuintilesIMS, Plymouth Meeting, PA, USA, 2Bristol-Myers Squibb, Princeton, NJ, USA
OBJECTIVES: Survival is an important endpoint in outcomes research. Identifying mortality utilizing EMR and claims data is challenging, as they lack reliable mortality records. We describe a method of generating survival curves by linking EMR to claims data, and initial validation of a mortality proxy. METHODS: Using the QuintilesIMS Oncology EMR, we identified patients with stage III/IV gastric, glioblastoma (GBM), head and neck, melanoma or lung cancers (NSCLC and SCLC). The first systemic treatment between 1/1/2014 and 6/30/2015 was the index date. Using a HIPAA-compliant encrypted ID, patients were linked to QuintilesIMS longitudinal prescription data (LRx), a consistent HEM/ONC panel of outpatient medical claims (Dx), and a consumer panel death index incorporating the SSDMF. The last observed claim in LRx/Dx followed by ≥6 months of no activity defined the mortality proxy. Validation of the proxy was determined through agreement the month and year of death within the death index. Survival time from index was estimated using Kaplan Meier analysis. RESULTS: A total of 11,010 patients were evaluated, 4,895 linked to LRx/Dx data, and 2,912 (60%) met the death proxy; of these; 1,076 (37%) linked to the consumer panel, with 376 having month/year of death. When compared to the LRx/Dx claims proxy, 100% agreed on death, with 295 (78%) agreeing on exact month/year of death, 62 (16%) agreeing +/- 1 month, 12 (3%) agreeing +/- 2 months, and 7 (2%) had agreement of >2 months. The 12 month survival from index treatment in the study population was 56% for gastric cancer, 46% for GBM, 61% for head and neck cancer, 62% for melanoma, 43% for NSCLC, and 33% for SCLC. CONCLUSIONS: A death proxy using linked LRx/Dx claims in stage III/IV cancer patients correlated accurately with a consumer death index, and with further validation may serve as a useful tool in outcomes research studies.
Conference/Value in Health Info
2017-11, ISPOR Europe 2017, Glasgow, Scotland
Value in Health, Vol. 20, No. 9 (October 2017)
Code
PCN56
Topic
Clinical Outcomes
Topic Subcategory
Relating Intermediate to Long-term Outcomes
Disease
Oncology