SBIR Phase I: Predictive Analytics and Machine Learning Modeling for New Patient Cancer Referrals

Award ID: 2304498

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to decrease patient referral wait times. Referral wait times are often long since offices need to retrieve a large amount of medical information on a patient before they are seen by a doctor. Unfortunately, medical records are often not stored in one place, making it difficult to gather the needed medical histories. Quick and complete medical record retrieval is especially important for cancer patients, whose conditions can quickly change. Critical patients need to be seen by doctors in a timely manner to begin treatment. The company is creating a technology that could help quickly retrieve medical information to decrease the time from referral to appointment. The company expects these algorithms to expedite document reconciliation by 7 days, thereby reducing the time from referral for the new patient appointment by 1 week. By facilitating quicker and more meaningful record retrieval, the algorithms are expected to improve treatment initiation by 7-14 days. The company plans to commercialize its technology for use in large academic healthcare systems, first focusing on those with high-volume cancer centers. This Small Business Innovation Research (SBIR) Phase I project will advance a new patient referral predictive analytics software platform for cancer centers. This platform will streamline referrals, increase resource utilization, and optimize care pathways. The companys deep learning algorithms will be developed to streamline record retrieval for new patient appointments and recognize critical medical conditions, resource capacity, local referral patterns, and at-risk socioeconomic factors. This intervention may reduce the mortality risk by 3.2-6.4% per week per patient. To achieve these objectives, the software will contain two major components a cloud-based platform for medical information exchange and an machine learning (ML)-based analytics platform. Once fully developed and launched, it is anticipated that real-world deidentified and aggregated clinical data from the exchange platform will be used to further train and refine the ML model. Prior to this stage, data from large publicly available and multi-institutional databases will be used to provide training data points for the model. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Alastair Monk