Forecasting models of infections due to carbapenem-resistant Gram-negative bacteria in an intensive care unit in an endemic area

Abstract Objectives To forecast the monthly incidence rates of infections (infections/1000 bed-days, IBD) due to carbapenem-resistant Klebsiella pneumoniae (CRKP), Pseudomonas aeruginosa (CRPA), Acinetobacter baumannii (CRAB) and total Gram-negative bacteria (CRGNB) in an endemic intensive care unit (ICU) during the subsequent year (December 2016-December 2017). Methods An observational 52-month period (August 2012-November 2016) was used. Two forecasting models, simple seasonal model for CRGNB, CRKP and CRPA, and Winters' additive model for CRAB infections, were applied. Results They predicted highest infection rates for CRKP, CRAB and CRGNB in January and September 2017 (23.8/23.4, 24.6/28.5 and 46.8/46.7 IBD, respectively) and for CRPA in February (8.3) and March 2017 (7.9). The highest observed rates for CRKP, CRAB, and CRGNB were indeed in January and September 2017 (25.6/19.04, 34.2/23.8 and 59.8/42.8 IBD, respectively); and for CRPA in February (15.2) and March of the same year (12.7). The increased rates may be associated with personnel's annual work program and behavioral factors. Conclusions Forecasting models in endemic ICU's may assist in the implementation strategies of infection control measures.