RESUMO
BACKGROUND: Recidivism is a key outcome measure for injury prevention programs. Firearm injury recidivism rates are difficult to determine because of poor longitudinal follow-up and incomplete, disparate databases. Reported recidivism rates from trauma registries are 2% to 3%. We created a collaborative database merging law enforcement, emergency department, and inpatient trauma registry data to more accurately determine rates of recidivism in patients presenting to our trauma center following firearm injury. METHODS: A collaborative database for Jefferson County, Kentucky, was constructed to include violent firearm injuries encountered by the trauma center or law enforcement from 2008 to 2019. Iterative deterministic data linkage was used to create the database and eliminate redundancies. From patients with at least one hospital encounter, raw recidivism rates were calculated by dividing the number of patients injured at least twice by the total number of patients. Cox proportional hazard models were used to evaluate risk factors for recidivism. The cumulative incidence of recidivism over time was estimated using a Kaplan-Meier survival model. RESULTS: There were 2, 363 assault-type firearm injuries with at least 1 hospital encounter, approximately 9% of which did not survive their initial encounter. The collaborative database demonstrated raw recidivism rates for assault-type firearm injuries of 9.5% compared with 2.5% from the trauma registry alone. Risk factors were young age, male sex, and African American race. The predicted incidence of recidivism was 3.6%, 5.6%, 11.4%, and 15.8% at 1, 2, 5, and 10 years, respectively. CONCLUSION: Both hospital and law enforcement data are critical for determining reinjury rates in patients treated at trauma centers. Recidivism rates following violent firearm injury are four times higher using a collaborative database compared with the inpatient trauma registry alone. Predicted incidence of recidivism at 10 years was at least 16% for all patients, with high-risk subgroups experiencing rates as high as 26%. LEVEL OF EVIDENCE: Epidemiological, level III.
