STATUS: Completed Project
National-level statistics on opioid-related hospitalizations are limited and often incomplete. Electronic health record (EHR) contain much more detailed data such as clinical notes and laboratory results, which allow for a wider perspective on the hospitalization. These combined data will allow health policymakers, clinicians, and researchers to develop effective prevention strategies and improve patient care. The National Center for Health Statistics (NCHS) houses three data sources that, when combined, will offer broad, national-level data on hospital care and death related to opioid-involved drug overdose. The first source, the National Hospital Care Survey (NHCS), attempts to collects inpatient, emergency department, and outpatient claims and EHR data from a sample of approximately 600 hospitals. The second source is the National Death Index (NDI), which includes all deaths occurring within the U.S. along with cause of death. The third source, The National Vital Statistics System restricted-use Drug-Involved Mortality files (DIM) on drug overdose death includes information on specific drugs involved in overdose death. Each of these three sources has limitations, such as the identification of specific opioids and the inclusion of deaths occurring outside of hospital settings. Prior projects in the Office of the Secretary Patient-Centered Outcomes Research Trust Fund (OS-PCORTF) portfolio including Enhancing Data Resources for Researching Patterns of Mortality in Patient-Centered Outcomes in Research: Projects 1-4 and Improving the Mortality Data Infrastructure for Patient-Centered Outcomes Research project have improved specificity within each source. The combination of these three data sources will allow researchers to follow patients with an opioid event from presentation at a hospital to death (if applicable) and retroactively analyze previous hospital encounters for patients that died of an opioid overdose for more information.
PROJECT PURPOSE & GOALS
The overall goal of this project was to improve surveillance and expand researchers’ access to data on hospital care patterns and risk factors associated with opioid overdose deaths.
- Create a merged data set for research by linking data between hospital care and mortality by merging the NHCS, NDI, and DIM and analyze this dataset to explore characteristics of individuals who have opioid-related events, patterns of hospital use in months prior to death, and comparison of patients and services.
- Improve researchers' ability to identify opioid-specific hospital encounters and deaths by improving existing techniques and developing new methods in vocabulary and procedure coding.
- Enrich the opioid-specific hospital care data available in the linked NHCS, NDI, and DIM dataset with enhanced hospital and death certificate opioid identification.
- Invest in the infrastructure to improve the collection and reporting of hospital data as well as disseminate the methodologies, analyses, and knowledge obtained to promote the use of the enhanced data set for patient-centered outcomes research (PCOR).
PROJECT ACHIEVEMENTS AND HIGHLIGHTS
- The project team linked 2014 NCHS and 2014/2015 NDI data using deterministic and probabilistic record linkage methods. The resulting dataset was linked to the 2014/2015 DIM file.
- The team produced a linked 2016 NHCS/NDI/DIM data set, which includes information on hospital care, mortality post-hospital discharge, and specific drugs mentioned in the literal text on the death certificate.
- The project team built the NHCS Annual Hospital Report (AHR) secure web-based portal, which allows participating hospitals that submit 12 months of NHCS data to NCHS to produce customizable reports on hospital encounters. Participating hospitals have access to a patient mortality report for their hospital that presents results from the linked NHCS and NDI data.
PUBLICATIONS, PRESENTATIONS, AND OTHER PUBLICALLY AVAILABLE RESOURCES
- The linked 2014 NHCS data, 2014/2015 NDI data, and 2014/2015 DIM dataset is available through the NHCS Research Data Center. A summary report provides a detailed description of the data sources and methods used to link the data.
- The linked 2016 NHCS data, 2016/2017 NDI data, and 2016/2017 DIM dataset is available through the NHCS Research Data Center. A summary report provides a detailed description of the data sources and methods used to link the data.
- The report “The Linkage of the 2014 National Hospital Care Survey to the 2014/2015 National Death Index: Methodology Overview and Analytic Considerations” describes the linkage process between the 2014 NHCS and the 2014/2015 NDI data, and it includes a brief overview of the data sources, a description of the methods used for linkage, and analytic guidance for researchers. The report is available here: https://www.cdc.gov/nchs/data/datalinkage/NHCS14_NDI14_15_Methodology_Analytic_Consider.pdf.
- A methodology report describes the development of the enhanced opioid-identification algorithm utilizing the 2016 NHCS data. The report is available here: https://www.cdc.gov/nchs/data/series/sr_02/sr2-188.pdf.
- A National Health Statistics report utilized the enhanced opioid algorithm developed by the project by analyzing opioid-involved ED visits in the 2014 NHCS and the 2013-2016 National Hospital Ambulatory Care Survey. The report is available here: https://www.cdc.gov/nchs/data/nhsr/nhsr149-508.pdf.
- A National Health Statistics report utilized the 2016 NHCS data linked to the 2016/2017 NDI and DIM data to report on mortality following nonfatal opioid overdose visits to the emergency department. The report is available here: https://www.cdc.gov/nchs/data/nhsr/nhsr187.pdf.
- Hospitals that submitted 12 months of administrative claims to the 2019 NHCS can view their data in the NCHS AHR portal. A demonstration version of the portal is available here: https://www.cdc.gov/nchs/nhcs/annual_hosp_report_portal.htm.
- “Assessing National Hospital Care Survey and National Hospital Ambulatory Medical Care Survey Data: A Comparison of Opioid and Respiratory Disease Encounters,” at the 2020 Joint Statistical Meetings. Presentation slides are available here: https://www.cdc.gov/nchs/data/nhcs/JSM-508.pdf.
- “Using Linked Data and Natural Language Processing to Support Patient-Centered Outcomes Research on Opioids” at the 2018 AcademyHealth Annual Research Meeting.
- “Utilizing EHR Data in NCHS Data Systems” at the December 2018 NCHS Board of Scientific Counselors meeting.
- “Addressing the Opioid Epidemic: Harnessing the Power of Data for Patient-Centered Research” at the December 2018 ASPE-hosted meeting.
- An overview of the project at the 2019 National Center for Health Statistics and Statistics Canada Interchange.
- “The National Hospital Care Survey: Modernizing the Monitoring of the Nation’s Health Care by Linking Electronic Health Records to Death Record Information and Administrative Data” at the University of Kentucky School of Public Health in October 2019.
- “Using linked hospital care and mortality data to enhance identification of opioid-involved health outcomes” virtually at the 2020 Rx Drug Abuse & Heroin Summit.
- “Identification of Opioid Involved Health Outcomes Using Linked Hospital Care and Mortality Data” at the 2020 International Conference on Health Policy Statistics.
- “Linked Health Data from the National Center for Health Statistics: A Study of Opioid-Involved Emergency Department Visits, Hospitalizations and Mortality” virtually at the 2020 AcademyHealth Annual Research Meeting.
- A webinar for NHCS sampled hospitals at the National Center for Health Statistics to provide an overview on the release of the Annual Hospital Report portal.
- “Using machine learning and natural language processing to enhance the identification of opioid-involved encounters in hospital electronic health records” virtually at the 2021 AcademyHealth Annual Research Meeting.
- “Drug overdose mortality risk for patients presenting to the emergency department for nonfatal opioid overdose” virtually at the 2021 AcademyHealth Annual Research Meeting.
- “Harnessing natural language processing and machine learning to enhance identification of opioid-involved health outcomes in the National Hospital Care Survey” at the 2021 National Center for Health Statistics and Statistics Canada Interchange.
- “Enhancing identification of opioid-involved health outcomes and identifying co-occurring disorders among opioid users using linked hospital care and mortality data” virtually at the 2022 CTN NIDA Steering Committee Meeting.
- “Differences in opioid-involved hospitalizations identified by enhanced opioid-involved algorithm approaches using natural language processing and medical codes,” at the 2022 AcademyHealth Annual Research Meeting.
- “Natural language processing for EHR clinical notes data” at the 2022 AIM-Ahead workshop.
- “Adapting COVID-19 early release data to determine impact of opioid use in hospital emergency departments,” at the 2022 Joint Statistical Meeting.
Below is a list of ASPE-funded PCORTF projects that are related to this project
Enhancing Data Resources for Researching Patterns of Mortality in Patient Centered Outcomes Research – Through collaboration between the CDC, Centers for Medicare & Medicaid Services, and U.S. Food and Drug Administration, the overall goal of this project was to increase the availability of information on the cause of death by linking NDI data to other sources. Enabling linkages will allow researchers to develop national estimates of cause-specific death rates following emergency department (ED) visits and/or hospital stays for specific conditions. The project produced linkages of patient EHRs and national mortality data. Additionally, it linked the NDI’s death and cause of death data with the Master Beneficiary Summary File and the Medicaid Enrollee Supplemental File. The project team also created new methods to optimize data linkages when using large national data files.
Improving the Mortality Data Infrastructure for Patient-Centered Outcomes – Comprised of all U.S. mortality events since 1979, the NDI database allows researchers to match entries in the NDI to those participating in longitudinal clinical and epidemiologic studies to determine both fact and cause of death. A significant challenge with the NDI has been the lag between the date of death and the availability of the record for matching purposes. The CDC’s NCHS worked to improve the infrastructure to support more timely and complete mortality data collection through: 1) more timely delivery of state death records (e.g., cause of death) to the NDI database; and 2) linking NDI records with nationally collected hospital datasets to obtain a more complete picture of patient care.