Stemming from Hospital Acquired Infections

We introduce a Markov chain model to represent a patient's path in terms of the number and type of infections sjhe may have acquired during a hospitalization period. The model allows for categories of patient diagnoses, surgery, the four major types of nosocomial (hospitalacquired) infections and discharqe or death. Data from a national medical records survey including 58,647 patients enable us to estimate transition probabilities and, ultimately, perform statistical tests of fit, including a validation test. Novel parameterizations (functions of the transition matrix) are introduced to answer research questions on timedependent infection rates, time to discharge or death as a function of patient characteristics at admission"and conditional infection rates reflecting intervening variables (e.g., surgery). A STOCHASTIC MODEL TO MEASURE PATIENT EFFECTS STEMMING FROM HOSPITAL ACQUIRED INFECTIONS by George T. Kastner and Richard H. Shachtman Researchers modeling hospital activities often require a quantitative representation of a patient's probabilistic health status changes during a hospitalization stay. Frequently, the stochastic model utilized is a Markov model. See for example Thomas (1968), Bush et a1. (1971) and Kao (1972). These investigators present disease-specific app1ications; Bush et a1. studY adults with primary active tuberculosis while both Thomas and Kao studY the recovery process of coronary care patients. Our model, also a Markov approach, follows a patient's path in terms of type and number of infections s/he may have acquired during a hospitalization period in a particular hospital/patient group, independent of the cause of admission. This model allows us to predict the incidence of nosocomial (hospital acquired) infections (Nls) at the patient-day level. Moreover, our approach suggests novel parameterizations (functions of the transition matrix) to provide decisionmakers with additional criteria for evaluating hospitalization outcomes. We provide answers to several research questions, for example: how can we estimate the time spent in distinct health states? How much longer does a patient with an NI spend in the hospital than a patient with similar characteristics who has not acquired an NI? Does her/his