Accuracy of Electronic Medical Record Follow-Up Data for Estimating the Survival Time of Patients With Cancer

Integrated database-based Screening Cohort for Asian Nomadic descendants in China (Scan-China): Insights on prospective ethnicity-focused cancer screening

Established in 2017, the Screening Cohort for Asian Nomadic descendants in China (Scan-China) has benefited over 180,000 members of a multi-ethnic population, particularly individuals of Mongolian descent compared with the general population (Han ethnicity), in the Inner Mongolia Autonomous Region, China. This cohort study aims to evaluate the effectiveness of cancer screening and serve as a real-world data platform for cancer studies. The 6 most prevalent cancers in China are considered-namely, breast, lung, colorectal, gastric, liver and esophageal cancer. After baseline cancer risk assessments and screening tests, both active and passive follow-up (based on the healthcare insurance database, cancer registry, the front page of hospital medical records, and death certificates) will be conducted to trace participants' onset and progression of cancers and other prevalent chronic diseases. Scan-China has preliminarily found a disproportionately lower screening participation rate and higher incidence/mortality rates of esophageal and breast cancer among the Mongolian population than among their Han counterparts. Further research will explore the cancer burden, natural history, treatment patterns, and risk factors of the target cancers.

Performance drift in a mortality prediction algorithm among patients with cancer during the SARS-CoV-2 pandemic

Sudden changes in health care utilization during the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic may have impacted the performance of clinical predictive models that were trained prior to the pandemic. In this study, we evaluated the performance over time of a machine learning, electronic health record-based mortality prediction algorithm currently used in clinical practice to identify patients with cancer who may benefit from early advance care planning conversations. We show that during the pandemic period, algorithm identification of high-risk patients had a substantial and sustained decline. Decreases in laboratory utilization during the peak of the pandemic may have contributed to drift. Calibration and overall discrimination did not markedly decline during the pandemic. This argues for careful attention to the performance and retraining of predictive algorithms that use inputs from the pandemic period.