Combining Claims Data, EHR Best for Identifying Seriously Ill Populations
Key findings
- This study investigated the feasibility of identifying adults with serious medical conditions (SMC) by applying ICD-10 diagnosis codes to electronic health records (EHR) and claims data
- The study subjects were 11,516,072 adults enrolled in a commercial or Medicare Advantage health plan (claim-based sample) and 17,367,524 adults with two or more qualifying encounters in an integrated delivery network (EHR-based sample)
- People with a serious medical condition were identifiable in both data sources (10% of the claims-based sample; 9% of the EHR-based sample), which also confirmed high rates of chronic conditions, hospital admissions and emergency department admissions
- Claims data alone detected 93% of patients with SMC but claims data alone detected only 56%, compared with the combination of EHR and claims data
- When possible, both types of data sources should be used for population-based planning of palliative care services and for benchmarking across health care systems and providers
Conceptually, serious illness is easy to understand: a health condition that carries a high risk of death and either negatively affects a person's daily function or quality of life or excessively strains their caregivers. However, no standard operational definition of serious illness exists because information on function, quality of life and caregiver strain usually are not recorded in patient records.
Subscribe to the latest updates from Geriatrics & Palliative Care Advances in Motion
Amy S. Kelley, MD, MSHS, of the Icahn School of Medicine at Mount Sinai, Christine Seel Ritchie, MD, MSPH, director of research in the Division of Palliative Care and Geriatric Medicine and director of the Mongan Institute Center for Aging and Serious Illness at Massachusetts General Hospital, and colleagues have demonstrated the feasibility of applying diagnosis codes to identify adults with serious illness in a commercially insured population that included Medicare Advantage enrollees. The details appear in the Journal of Pain and Symptom Management.
Study Methods
The team queried the Optum Labs Data Warehouse, which includes administrative claims data for commercially insured and Medicare Advantage enrollees and standardized electronic health record (EHR) data from a nationwide network of provider groups. They initially compared two cohorts:
- Claims-based population: 11,516,072 adults enrolled continuously in a commercial or Medicare Advantage health plan in 2016 (23% were ≥65 years old)
- EHR-based population: 17,367,524 adults with two or more qualifying encounters in an integrated delivery network in 2016 (24% were ≥65 years old)
The team defined a serious medical condition (SMC) as any of hundreds of ICD-10 diagnosis codes entered in the data sources in 2016. The full list is available as a spreadsheet appended to the article; in brief, the codes of interest were for:
- Advanced cancer
- End-stage or stage 5 kidney disease
- Dementia
- Advanced lung disease and advanced congestive heart failure (only if using home oxygen or hospitalized for the condition)
- Advanced liver disease
- Diabetes with severe complications
- Advanced Parkinson's disease with an indicator of dementia or selected durable medical equipment (DME)
- Other neurodegenerative diseases
- Hip fracture at age ≥70
- Stroke requiring hospital admission
- HIV infection with AIDS complications
Identification of SMCs
People with an SMC were identifiable in both data sources, although the populations differed:
Claims-based sample: 10%
- Ages 18–64: 5.4%
- Ages ≥65: 27%
EHR-based sample: 9%
- Ages 18–64: 5.6%
- Ages ≥65: 21%
Health Characteristics and Health Care Utilization
Substantial proportions of the four populations with an SMC (21%–24% of younger adults; 41%–44% of adults ages ≥65) also had two or more chronic conditions.
Measures of utilization across SMC populations were similar in many respects. High rates of hospital admissions (14%–26%) and ED admissions (22%–43%) confirmed a need to improve care for seriously ill individuals.
The claims-based sample revealed a 1.2% prevalence of DME use, which was rarely observed in the EHR-based sample (0.1%). Thus, conditions requiring the use of DME as an indicator of disease severity will be more readily identified using claims data.
Subgroup Analysis
Of 860,000 individuals who had both claims and EHR data available:
- 172,540 people with an SMC were identified by at least one data source
- Only 86,460 people with an SMC (50%) were identified by both data sources
- 48% of the SMC population identified in claims was not identified by EHR
- 13% of the SMC population identified by EHR was not identified in claims
Implications for Health Care Systems
Claims data alone detected 93% of patients with SMC but EHR data alone detected only 56%, compared with the combination of claims and EHR data. When possible, the use of both types of data sources is recommended for population-based planning of palliative care services and for benchmarking across health care systems and providers.
Novel techniques, such as natural language processing of narrative notes in EHR, may someday facilitate the use of these data sources.
view original journal article Subscription may be required
Visit the Division of Palliative Care and Geriatric Medicine
Refer a patient to the Division of Palliative Care and Geriatric Medicine