In-hospital injuries of medical and surgical patients: the predictive effect of a prior injury.

Four hundred three consecutive injury victims admitted via emergency ward over a 3-month period were tracked with 403 contemporaneous controls with medical illness and blindly assessed for in-hospital accidents. There was a high frequency of "incidents" (n = 161 in 107 patients, mainly falls and medication errors), but injury victim admissions resembled medically ill controls when compared by the Kaplan-Meier method for cumulative probability of occurrence of an in-hospital incident. Accelerated failure/time models using the Weibull method to compute average times from admission to incident showed little difference between groups. Admission type (injury victim vs. control) did not predict psychiatric consultation, incident type, or multiple incidents in hospital. Although the injury group had a larger proportion of males and lower mean age, stratification to control for age and sex did not significantly discriminate injury victims from controls in production of incidents: Over all risk of incidents was random. By studying patients during a hospital stay, the effects of differences in individual environment and drug and alcohol intoxication are largely factored out; under these conditions the predictive effect of a prior injury becomes insignificant. In-hospital injury is associated with host factors long known to promote falls: increasing age, debility-cum-mobility, and central nervous system depressant medication.