A Markov model to estimate Salmonella morbidity, mortality, illness duration, and cost

Abstract

Approximately 690000-1790000 Salmonella cases, 20000 hospitalizations, and 400 deaths occur in the USA annually, costing approximately $2.6bn. Existing models estimate morbidity, mortality, and cost solely from incidence. They do not estimate illness duration or use time as an independent cost predictor. Existing models may underestimate physician visits, hospitalizations, deaths, and associated costs. We developed a Markov chain Monte Carlo model to estimate illness duration, physician/emergency room visits, inpatient hospitalizations, mortality, and resultant costs for a given Salmonella incidence. Interested parties include society, third-party payers, health providers, federal, state and local governments, businesses, and individual patients and their families. The marginal approach estimates individual disease behavior for every patient, explicitly estimates disease duration and calculates separate time-dependent costs. The aggregate approach is a Markov equivalent of the existing models; it assumes average disease behavior and cost for a given morbidity/mortality. Transition probabilities were drawn from a meta-analysis of 53 Salmonella studies. Both approaches were tested using the 1993 Salmonella typhimurium outbreak in Gideon, Missouri. This protocol can be applied to estimate morbidity, mortality and cost of specific outbreaks, provide better national Salmonella burden estimates, and estimate the benefits of reducing Salmonella risk.