Death
Epic equivalent: Discharge records (deceased disposition) / State death index / Vital statistics
The death table captures mortality information — one row per deceased person. At Emory, sources include EHR discharge records and state death data supplements.
This table is critical for survival analysis, mortality endpoints, and time-to-event studies.
Epic-to-OMOP Field Mapping
Field reference (click to expand)
| OMOP Field | Epic Equivalent | What It Captures |
|---|---|---|
person_id |
Patient ID / MRN | Identifies the deceased patient |
death_date |
Date of death | As recorded from available sources |
death_datetime |
Time of death | Rarely populated unless from inpatient EHR |
death_type_concept_id |
Data provenance | EHR, claims, registry, state death data |
cause_concept_id |
Primary cause of death | SNOMED concept for cause (if available) |
cause_source_value |
ICD-10 cause code | Original value from death certificate or discharge diagnosis |
What to Watch For
Common pitfalls
- Absence does not mean alive
- If a patient is not in the
deathtable, they may still be deceased — the death simply wasn't captured. This is a right-censoring issue in survival analysis. - Date precision varies
- Deaths from external sources (claims, state indices) may have imprecise dates. Validate for time-to-event analyses.
- Cause of death is often missing
- Structured cause of death requires vital records data. Many EHR-only deaths have no
cause_concept_id.
Research Patterns
| Question | Tables Involved |
|---|---|
| All-cause mortality in the oncology cohort | death + condition_occurrence (cancer) |
| Median survival from Alzheimer's diagnosis | condition_occurrence (Alzheimer's) + death.death_date |
| 30-day post-surgical mortality | procedure_occurrence + death + date interval |
| Cause-of-death distribution in stroke patients | condition_occurrence (stroke) + death.cause_concept_id |
| Mortality and social risk factors | observation (SDoH) + death |