Clare Medical Develops AI Based Platform To Predict 30 Day Hospital Admissions
19 Nov, 2019, 11:03 ET
MOUNT HOLLY, N.J., Nov. 19, 2019 /PRNewswire/ -- Clare Medical of New Jersey www.getclare.com, a provider of in-home comprehensive medical care to seniors, announced today the validation of a novel, proprietary Artificial Intelligence (AI)-based diagnostic tool to accurately predict, within an error range of only 3%, the likelihood of patients requiring admission to a hospital within 30 days. The data it generates can be deployed to avoid hospital admissions and reduce hospital and emergency room utilization, which are major costs to the healthcare system. This is the first of a suite of tools currently in development at Clare which are expected to provide accurate forecasts for a range of clinically meaningful events, whose onset can be halted or delayed by early intervention and providing appropriate corrective measures.
The 30-day risk of hospital admission and readmission are crucial metrics in determining how healthcare organizations and providers are graded. Because of the high costs of hospital admissions - organizations that can manage risk appropriately and succeed in reducing hospitalizations are likely to be well-positioned and have significant value in the current healthcare system. According to Elie Donath , MD,MPH,MBA , one of Clare's primary care medical directors and project lead, commented, "Although a variety of risk tools have been constructed as general measures of patient health – we don't know of any that are focused on this specific critical metric, drawing from such a vast data repository, while also relying on a deep learning-based framework that manages to deliver on this level of well-validated performance measures. We believe the development of this diagnostic tool has significant value and deployment of this model and the data it generates can be used to deliver better preventative care, contain costs and ultimately improve the lives of patients."
The platform was built on internal data derived from the company's EHR, as well as publicly sourced datasets. It uses an advanced form of artificial intelligence that gathers vital pieces of information from a variety of sources contained within a patient's chart – including provider notes, vital signs, medications, ICD10 coding and laboratory data – to output a "30-day hospital admission" probability score. The algorithm has been tested and validated in a variety of settings. Most recently, as part of an internal pilot trial, Clare gathered a large cohort of high-risk patients and computed a probability score for each of them and then prospectively monitored them over time to assess the likelihood of actually their requiring a hospital admission, in the 30-day period following the time the risk score was ascertained. In this pilot, across the entire prospective cohort, Clare's algorithm predicted, on average the 30-day risk score within 3% of what it was. The model's viability and accuracy were further confirmed when assessed with more sophisticated techniques (including AUC-ROC and F1 score) as it performed exceptionally on those measures, as well.
Clare is also exploring an expansion of the AI platform beyond predicting "30-day admissions" to predicting 30-day risk scores associated with a variety of additional clinical outcomes (including pneumonia, urinary tract infection, CHF exacerbation, COPD exacerbation among others).
Clare Medical of New Jersey provides comprehensive in-home medical care to seniors and individuals with multiple chronic conditions throughout New Jersey. Our team of mobile clinicians provides timely comprehensive proactive care to patients residing in Assisted Living and Independent Living communities and those residing in their own home. Clare partners with over 60 organizations throughout the state which refer Clare patients in need of in-home medical care and partners with the local provider community physicians and clinicians to service their home -bound patients.
Contact: Ron Lipstein, [email protected]
SOURCE Clare Medical
Share this article