Data analysis

ATP and culture plate data were highly skewed and analyzed with the median, range and non-parametric statistics. Differences in ATP RLU’s from baseline to post cleaning and ultraviolet disinfection were compared using the median and Wilcoxon signed rank tests. Differences in microbial cultures between baseline, post cleaning, and post ultraviolet disinfection were described using frequencies and percentages. The Cochran’s Q non-parametric test for related samples was used to compare differences in culture results between the three time points.

Our analysis of infection rates used ratios to account for the lack of independence between rates of nursing-home acquired and hospital-acquired infections. Hospital-acquired infections have become more common as medical treatments and patient complexity have increased and are the sixth leading cause of death in the United States [18, 19]. Our data showed substantial increases each year in hospital acquired infections between 2012 and 2015. Nursing homes are federally mandated to consider an infection nursing-home acquired if symptoms emerge 48 h or more after hospital discharge [15]. This siloing of infection attribution is problematic because the incubation periods for many common viral and bacterial illnesses are longer than 48 h [20, 21]. The reciprocal relationship between hospital and nursing home acquired infection has been demonstrated [22]. Using a model based on actual patient and agency data to simulate the movement of infection between hospitals and nursing homes, the influence of hospitalization on nursing home MRSA prevalence and nursing homes influence on hospital MRSA prevalence levels was demonstrated [22]. Hence, the hospital acquired infection rates were analyzed relative to nosocomial infection rates.

To analyze differences in infection rates from pre to posttest, the independent variables were the 36 months of pre disinfection time (2012, 2013, and 2014) and 12 months of post disinfection time (2015). The dependent variables are hospital acquired and nursing home acquired infection rates for the urinary tract, respiratory tract, and skin. Enteric infections occurred at such a low frequency inferential analyses could not be performed. Univariate analyses of variance (ANOVAs) were initially examined to determine if there were significant differences in rates of urinary, respiratory, and skin and enteric infections between the three pre and one post ultraviolet irradiation study periods. The effect of the ultraviolet disinfection was computed using ANOVA with an a priori contrast comparing the ratio of hospital acquired to nursing home acquired infections from pre to posttest. If the hospital acquired rate increased and the nosocomial rate decreased or stayed the same the ratio increased. If the hospital acquired rate decreased and the nosocomial rate stayed the same or decreased the ratio decreased.

Rates of infection-related hospitalization were compared from pre to posttest using Chi Square. Hospitalization rates per month were examined relative to the monthly census during that time period.