SBIR-STTR Award

War-related syndromes have existed throughout modern wars; however, studies to-date have not provided evidence to point to a specific group of symptoms and their causes. Nevertheless, these studies have confirmed increased rates of reported symptoms and general deterioration of perceived health by war veterans. Military health care systems today offer unique opportunities in monitoring and predicting health-related events of military personnel. A closer investigation of war-related symptoms is warranted to evaluate the adequacy of the information collected and the analytical approaches employed. A data mining infrastructure can facilitate (i) collection of specific fields not previously gathered, and (ii) avoid forced entry of data fields from field personnel that are not relevant. Existing biosurveillance systems are well suited for counts-based analyses and efficient data collection. They do not offer analytical techniques to visualize dependencies between known and unexplained symptoms. New approaches in assessing information content will allow examination of syndromes and their associations with medical and non-medical inputs. Further efforts can then refine the architecture and apply a suite of analyses to provide decision management tools. The availability of such a system can not only leverage health care data for disease modeling but also enable timely response management.

The field of medically unexplained symptoms (MUS) in post-deployment situations presents a unique opportunity for enhancing further understanding and prospective monitoring. Silico Insights structured its Phase I efforts in two areas: literature review, and assessment of critical data elements for reliable clinical analysis. In general, the availability of electronic information and infrastructure today is heartening for a proposed Phase II development of a MUS-based data mining infrastructure. Specifically, existing clinical cohort-type analyses of post-deployment responses provide insight on what data may provide most information, what analyses may extract the most features and how a proactive approach may be implemented for MUS. In implementing steps towards a productive completion of a Phase II effort, the following key questions will be addressed: 1. What data hold the most information relevant to MUS? 2. What data elements among these data are not currently collected reliably? 3. What are the minimum "new" data elements most critical for future MUS analyses? 4. How can these critical elements be reliably collected and combined with other required existing data fields to provide the basis for MUS data mining? 5. What analytics would be needed for extracting the most value from such a database.

Keywords:
DATA MINING, MEDICALLY UNEXPLAINED SYMPTOMS, UNSUPERVISED ALGORITHMS