Probabilistic Disease Surveillance

University Of Pittsburgh At Pittsburgh, Pittsburgh PA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): The proposed research will further develop and evaluate a probabilistic approach to disease surveillance. In this approach, a probabilistic case detection system (CDS) uses Bayesian diagnostic networks to compute the likelihoods of patient findings for each of a set of infectious diseases for every patient in a monitored population. CDS computes these likelihoods from data in electronic medical records, including information derived from free-text reports by natural language processing. CDS makes those estimates available to a probabilistic outbreak detection and characterization component (ODCS). ODCS also utilizes a Bayesian approach to compute the probability that an outbreak is ongoing for each of a set of infectious diseases of interest, given information from CDS. ODCS also computes probability distributions over the current and future size of a detected outbreak and other characteristics such as incubation period used by public health officials when responding to an outbreak. The proposed research will extend the approach, which we have already developed and evaluated for the disease influenza to six additional respiratory infectious diseases. The research will also extend the capabilities of ODCS to utilize non-EMR data, detect an unknown disease, and detect and characterize concurrent outbreaks. The planned evaluations will measure the accuracy of both CDS and ODCS using historical surveillance data from two regions and simulated outbreak data, which we will create by adding outbreak cases generated by an agent-based epidemic simulator to real baseline surveillance data from non-outbreak periods. The innovation being advanced by this research is a novel, integrated, Bayesian approach for the early and accurate detection of cases of diseases that threaten health and for the detection and characterization of outbreaks of diseases that threaten public health. The proposed approach has significant potential to improve the information available to public health officials and physicians, which can be expected to improve clinical and public health decision making, and ultimately to improve population health.