STTR Phase I: An integrated platform for the analysis of patient health record data to enable predictive clinical decision support

The broader impact/commercial potential of this Small Business Technology Transfer (STTR) Phase I project is to reduce preventable patient readmissions, streamline triage, and detect multi-organ diseases early. Currently, failures of care delivery and care coordination, overtreatment, and administrative complexity cost the American healthcare system an estimated 300 billion dollars per year, and are among the largest contributors to mortalities within clinical settings. The recent transition of the healthcare industry to electronic health records offers new opportunities to reduce these costs and mortalities through use of clinical decision support systems. However, existing clinical decision support systems have had limited impact, due in part to their failure to detect trends in patient status and neglect of risk factor interdependence. Further, these systems must be updated manually on a regular basis to combat declining accuracy over time. Thus, there exists a pressing need to improve the technology underlying clinical decision support systems. The proposed technology directly addresses the current limitations of clinical decision support technology while placing no additional burden on clinicians, thus easing its adoption into clinics. In addition, the large value proposition and life-saving potential lend broad commercial appeal to the clinical decision support system being developed in this study. The proposed project advances the analysis of trends in patient health information and the identification of correlations among physiological data that are useful in predicting patient outcomes. The vast amounts of patient health information that are collected in electronic medical records present opportunities for improving the quality of health care, as well as practical challenges that are associated with interpreting such data. These challenges include the processing of measurements taken unreliably and at irregular intervals, the quantification of the interdependence of health risk factors, and the development of infrastructure for effectively interfacing medical records with suites of tools for clinical decision support. This project entails the implementation of sophisticated data imputation procedures for repairing imperfect time series measurements and building trend features for use in disease prediction and patient transfer recommendation tools. Trend information will be combined with correlations between sets of vital signs and lab tests, and then optimized using a statistical scheme for reliably predicting patient outcomes. The integration of this analytic technology with existing clinical information technology infrastructure will empower clinicians to more effectively use the data available to them, reduce the costs associated with overtreatment and extended stay, and improve patient health care outcomes.