Probabilistic linkage of computerized ambulance and inpatient hospital discharge records: A potential tool for evaluation of emergency medical services

Study objectives: Emergency medical services (EMS) is an important part of the health care system. The effect of EMS on morbidity, mortality, and cost s of illness is difficult to evaluate because hospital information is not a vailable in out-of-hospital databases. We used probabilistic linkage to cre ate such a database from ambulance and inpatient data and demonstrate the p otential for linkage to facilitate evaluation of EMS responses resulting in hospital admission. Methods: Statewide ambulance and inpatient hospital discharge records were available for 1994 through 1946. Ambulance records indicating admission to the emergency department or hospital (165,649 records) were linked to inpat ient hospital records indicating emergency admission (146,292 records) by u sing probabilistic linkage. Out-of-hospital data (dispatch code, treatments rendered, and ages), linkage rates, and inpatient data (discharge status, charges, length of stay, and payer category) were analyzed. Results: We linked 24,299 (14.7%) ambulance events to inpatient hospital di scharges. If we had used exact linkage methods, we would have only linked 1 4,621 record pairs, a loss of nearly 40%. linkage rates were relatively con stant between years (approximately 15%) but differed by ambulance dispatch codes. Out-of-hospital dispatch codes with high linkage rates included brea thing problems (22.6%), chest pain (21.5%), diabetic problems (16.9%), drow ning incidents (14.9%), falls (19.2%), strokes (32.8%), and unconsciousness or fainting episodes (16.1%). Linkage to the hospital record provided acce ss to hospital outcome data. Inpatient mortality was 6.8%. Survivors were d ischarged home (60.7%), transferred to other acute-care facilities (3.6%) o r intermediate-care facilities (23.3%), or discharged with home health care provision (4.9%). The median length of stay was 3 days, and median charges were $6,620; total inpatient charges were $286,737,067. Conclusion: Probabilistic linkage enables ambulance and hospital discharge records to be linked together and potentially increases our ability to crit ically evaluate EMS by providing access to hospital-based outcomes. Such ev aluation will be fur ther improved by linking to ED, other outpatient, and other public health data sources.