Prevalence of missing data in the National Cancer Database and association with overall survival

Abstract

Background: There has been increased enthusiasm regarding the use of real-world data (RWD) for clinical evidence generation. Cancer registries are important RWD sources that rely on data abstraction from the medical record, however, patients with unknown or missing data elements are under-represented in studies that use such data sources. Therefore, we sought to determine the prevalence of missing data and its associated survival outcomes among cancer patients. Methods: All data elements within the National Cancer Database were reviewed for missing or unknown values for the three most common cancers in the United States diagnosed from 2006 to 2015. Patient records with incompletely abstracted data elements were identified. Prevalence of patients with missing data and their associated survival outcomes were determined. Results: A total of 1,198,749 non-small cell lung cancer (NSCLC), 2,120,775 breast cancer, and 1,158,635 prostate cancer patients were included for analysis. For NSCLC, there were 851,295 (71.0%) patients with missing data in data elements identified as variables of interest; 2-year survival was 33.2% for patients with missing data and 51.6% for patients with complete data (p<0.001). For breast cancer, there were 1,161,096 (54.8%) patients with missing data; 2-year survival was 93.2% for patients with missing data and 93.9% for patients with complete data (p<0.001). For prostate cancer, there were 460,167 (39.7%) patients with missing data; 2-year survival was 91.0% for patients with missing data and 95.6% for patients with complete data (p<0.001). There were significant differences in demographics, tumor characteristics, cancer stage, and treatments received between cancer patients with complete and incomplete records. Conclusions: The majority of patients in a large cancer registry-based RWD source have missing data that was unable to be ascertained from the medical record. Moreover, patients with missing data show worse survival than those with more complete documentation. Increasing documentation quality and adoption of rigorous missing data correction methods are needed to best leverage RWD for clinical advancements.

Expanding the Secondary Use of Prostate Cancer Real World Data: Automated Classifiers for Clinical and Pathological Stage

“…While critical to support evidence-based patient care, patients' medical records and cancer registries are often missing or have inaccurate staging information ( 4 – 6 ). Stage information is missing from 10 to 50% of patient records in cancer registries likely because of absent documentation of explicit stage in patients' medical records ( 7 – 9 ). However, the data used to derive cancer stage is often recorded in unstructured text within the electronic health records (EHR).…”