Catastrophe model for the exposure to blood-borne pathogens and other accidents in health care settings.

Catastrophe models, which describe and predict discontinuous changes in system state variables, were used to model the exposure to blood and bodily fluids and more conventional occupational accidents among 1708 health care workers. Workers at three hospitals completed a survey measuring HIV-relevant exposures (needlesticks, cuts, splashes, contact with open wounds), the accident rate for broadly-defined injuries, and several occupationally relevant themes: safety climate, shift work, depression symptoms, work pace, verbal abuse, and professional group membership. A cusp (cubic polynomial) model predicting HIV-relevant exposures specifically was more accurate (R2 = 0.56) than a comparable linear model containing the same variables (R2 = 0.07). Some of the foregoing variables predisposed workers to greater differences in HIV-relevant and general accident exposures: shiftwork, climate, depressive symptoms, and work pace. Other variables governed how close an individual was to a critical threshold where a harmful incident would take place: verbal abuse, professional group membership. Similarly, a cusp model for accident incidents predicted from HIV-relevant exposures and occupational variables was also more accurate (R2 = 0.75) than comparison models. Two variables predisposed the worker to a greater accident risk: depression symptoms and shift work. Four other variables predisposed the worker to lesser accident risk: job satisfaction, safety climate, environmental stressors, and work pace. Compliance with the universal precautions and HIV-related training were not relevant to either of the models.