A strategy to identify event specific hospitalizations in large health claims databases

Abstract

Background Health insurance claims data offer a unique opportunity to study disease distribution on a large scale. Challenges arise in the process of accurately analyzing these raw data. One important challenge to overcome is the accurate classification of study outcomes. For example, using claims data, there is no clear way of classifying hospitalizations due to a specific event. This is because of the inherent disjointedness and lack of context that typically come with raw claims data. Methods In this paper, we propose a framework for classifying hospitalizations due to a specific event. We then tested this framework in a private health insurance claims database (Symphony) with approximately 4 million US adults who tested positive with COVID-19 between March and December 2020. Our claims specific COVID-19 related hospitalizations proportion is then compared to nationally reported rates from the Centers for Disease Control by age. Results Across all ages (18 +) the total percentage of Symphony patients who met our definition of hospitalized due to COVID-19 was 7.3% which was similar to the CDC’s estimate of 7.5%. By age group, defined by the CDC, our estimates vs. the CDC’s estimates were 18–49: 2.7% vs. 3%, 50–64: 8.2% vs. 9.2%, and 65 + : 14.6% vs. 28.1%. Conclusions The proposed methodology is a rigorous way to define event specific hospitalizations in claims data. This methodology can be extended to many different types of events and used on a variety of different types of claims databases.

Research note changes in the number of outpatient visits in Japan during the COVID-19 pandemic

“…As our dataset included more than 100 million people, and Japan has universal health insurance coverage, we were able to identify large-scale societal trends in medical consultations associated with the COVID-19 pandemic, such as a decrease in the incidence of infectious diseases other than COVID-19, a temporary limited access to medical facilities, and the avoidance of medical consultations. Our analysis method using actual health insurance claim data can be applied not only in Japan but also in other countries where researchers have access to national information on health insurance claims [6][7][8]. As universal health coverage continues to be implemented in lowand middle-income countries, in keeping with World Health Organization recommendations, consideration should be given to developing health insurance databases also, which can be used for research purposes to draw-up health care policies not only for management of health crises such as the COVID-19 pandemic, but also for strengthening health care systems in general [9,10].…”

Analysis of Oral Anticoagulant Dosing and Adherence to Therapy Among Patients With Nonvalvular Atrial Fibrillation

“…Symphony includes claims submitted to all payer types, including commercial plans, Medicare, and Medicaid. These are adjudicated claims collected from major US clearing houses, as well as large national retail, mail order, and specialty pharmacy chains . This study received approval from the Duke University institutional review board, and informed consent was not needed because the data are anonymous, in accordance with 45 CFR §46.…”

“…These are adjudicated claims collected from major US clearing houses, as well as large national retail, mail order, and specialty pharmacy chains. 6 , 7 This study received approval from the Duke University institutional review board, and informed consent was not needed because the data are anonymous, in accordance with 45 CFR §46. We followed the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guidelines for cohort studies.…”

“…5 chains. 6,7 This study received approval from the Duke University institutional review board, and informed consent was not needed because the data are anonymous, in accordance with 45 CFR §46.…”

Outpatient medications associated with protection from COVID-19 hospitalization

“…Notably, claims data do not include laboratory results, and thus we rely on a treating physician's correct use of the above codes to identify cases. A comprehensive overview of this case definition and the strategy that precedes it can be found elsewhere [47].…”