Linking North Carolina's emergency department and ambulance records for disease surveillance

North Carolina has created computerized repositories of records of all Emergency Departments (ED) visits in the state and records of all patients transported by ambulance. Our objective was to develop and evaluate an algorithm to link patient records from these databases in order to strengthen public health surveillance activities in the state. Although these databases do not contain common patient identifiers such as social security number or name, they do contain patient sex and date of birth, hospital, and date and time (of emergency medical services (EMS) notification, for ambulance data, or of arrival at the ED, for ED data).

All 8600 ED records and all 1337 ambulance records reported for one day were obtained. We postulated that matches of ED and ambulance records on sex, date of birth, hospital, and time of ED arrival occurring within two hours after ambulance notification would be essentially unique. Classification of matches as correct or incorrect was determined by manual review of the chief complaint, which was also included in both datasets.

Roughly 11% of ED records with indication that the patient was transported to hospital by ambulance were matched to an ambulance record. Manual comparison of the ED and ambulance chief complaint indicated 100% matching accuracy. Primary reasons identified for low match percentage were missing or incomplete ambulance records. In conclusion, we demonstrated that highly valid probabilistic linkages can be established between ED and ambulance data; however, a higher level of participation of EMS providers in the state's mandatory reporting of ambulance runs is necessary in order for such linkage activities to be useful for surveillance purposes. Ongoing work is addressing data quality and completeness in both data sources and application of the matching algorithm to data from a single county where both ED and ambulance data are complete.