PREDICTING MORTALITY IN CLAIMS DATA AMONG PATIENTS WITH ADVANCED COLORECTAL CANCER

Author(s)

Wang WJ1, Meyer CS2, Wu N3, Sheinson D4
1The Comparative Health Outcomes, Policy, and Economics (CHOICE) Institute, University of Washington, Seattle, WA, USA, 2Genentech, Inc., Austin, TX, USA, 3Regeneron Pharmaceuticals, Inc., Tarrytown, NY, USA, 4Genentech, Inc., South San Francisco, CA, USA

OBJECTIVES

Commercial claims data lack information on vital status, potentially biasing outcomes in observational studies. An algorithm to predict mortality in claims data developed for patients with type 2 diabetes (Joyce, 2004) has been applied to other diseases without validation. The present study aimed to modify and validate the existing algorithm in a target population of patients with advanced colorectal cancer (CRC).

METHODS

The SEER-Medicare linked database (2015 linkage) which includes mortality information was used to modify and validate the existing algorithm. Patients newly diagnosed in 2013 with advanced CRC were followed to their last claim before end of enrollment or December 31, 2014. The distributions of diagnosis and procedure codes in their last month of claims were compared between those known to have died at end of enrollment versus survived. Based on this comparison as well as clinician input, a total of 84 diagnosis and procedure codes (e.g. liver failure, hospice) were selected for the modified algorithm. Patients were classified as having died if they had any of these codes during the month prior to their last claim. Performance of the algorithm was evaluated by estimating sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).

RESULTS

Among 1754 advanced CRC patients who met selection criteria, a total of 964 (54.96%) died by the end of 2014. The modified mortality algorithm’s sensitivity, specificity, PPV, and NPV were 92.01%, 91.39%, 92.88%, and 90.36% respectively. This improved upon the performance metrics of the original diabetes algorithm applied to advanced CRC patients, which were 66.29%, 86.58%, 85.77%, and 67.79%.

CONCLUSIONS

Modification of the existing algorithm for type 2 diabetes greatly improved its ability to predict mortality among advanced CRC patients following their last observed claim. Further development and validation are planned in advanced CRC as well as other cancer populations.

Conference/Value in Health Info

2020-05, ISPOR 2020, Orlando, FL, USA

Value in Health, Volume 23, Issue 5, S1 (May 2020)

Code

PCN276

Topic

Methodological & Statistical Research

Topic Subcategory

Artificial Intelligence, Machine Learning, Predictive Analytics

Disease

Oncology

Explore Related HEOR by Topic


Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×