Sharps injury surveillance in home healthcare

The main purpose of this dissertation is to improve sharp injury surveillance through improved statistical methods and analytical approaches. The data come primarily from Project SHARRP, a study of sharps injury and body/blood exposure in home healthcare based at the University of Massachusetts Lowell. An additional investigation used Monte Carlo simulations designed to represent typical injury surveillance data from a hospital injury reporting system. This dissertation is comprised of three manuscripts.; The first paper evaluated risk factors associated with sharps injuries (SI) and other blood/body fluid exposures (BBFE) among home healthcare (HHC) nurses and aides in eastern Massachusetts. Information on SI and BBFE experience, a variety of work environment and organizational factors, as well as whether an SI was reported to the employer was gathered using a self-administered survey questionnaire between October 2006 and March 2007. The rates of SI and BBFE, correcting for hours worked, were 5.1/100 FTE and 6.3/100 FTE among nurses and 1.0/100 FTE and 6.5/100 FTE among 282 aides. The study calculated that a part-time HHC nurse had nearly a 50% chance of having a sharps injury every ten years. The risk was higher for per-diem nurses, and lower for full-timers. Several other factors were associated with increased risk of injury or exposure, and these may be useful for preventive interventions to reduce risk.; The second paper assessed the utility of studying near-misses in the prevention of sharps injuries and other blood/body fluid exposures in the home healthcare. A conceptual framework for understanding the relevance of near-misses was developed, as an alternative to the standard safety pyramid. In order for near-misses to be useful in identifying risk factors for injury, these two occurrences must lie along a common causal pathway (the "common cause hypothesis"). Analyses of the Project SHARRP injury and blood/body fluid exposure data supported the common cause hypothesis, suggesting that one could combine events and near-misses to increase statistical power to detect risk factors. The study proposed a new theoretical model (the Snowman) as an alternative to the conventional safety pyramid, because the new model is better able to represent the subjectivity of the reporting of near-misses and injuries, and the "gray area" for events which maybe reported by some as and injury and by others as a near-miss.; The third paper investigated sources of overdispersion in Poisson regression with public health surveillance data, and compared alternative regression models to identify methods which were suitable for analyzing overdispersed surveillance data. Hospital based sharps injury datasets with varying degrees and sources of overdispersion were developed by Monte Carlo simulation and then the performance of different regression methods was compared with these simulated data. The study demonstrated clear evidence of overdispersion caused by unobserved covariates and by using a miss-specified offset. The study concluded that reducing chances of missing important predictors at the design level and specifying proper offsets are most important to minimize the problem of overdispersion. If overdispersion is present, negative binomial regression models may be preferred to conventional Poisson regression.; The dissertation concludes that important methodologic improvements in the analysis and interpretation of occupational health surveillance data can strengthen the public health impact of this branch of epidemiology.