MatrixCare is Leveraging Microsoft Azure Deep Machine Learning to Continuously Improve the Cost and Quality of Care for America’s Seniors
Over the last fifteen years, the proliferation of Electronic Health Record (EHR) solutions and the digitization of the patient record has been met with mixed reviews across the entire US healthcare landscape. Mixed, because the first couple of generations of EHRs didn’t quite live up to their vast potential and were oftentimes an impediment to caregiver productivity. However, the best of these solutions is now beginning to transition into a third generation of technology, where artificial intelligence and deep machine learning are being used to guide personalized care decisions to maximize both the patient experience and provider/operator/payor outcomes.
Generation 1 EHRs – “Electronic Filing Cabinets”
The harsh reality is that the first generation of EHR solutions, many still in use today, did little more than move the paper from the manila folder in the physical filing cabinet into the virtual manila folder in the electronic filing cabinet. Literally—with the electronic health record comprised of little more than scanned images of paper documents. Even the user interfaces were built to replicate existing paper-based forms. The focus was squarely on data entry and the result was an end-user experience that was difficult at best. Rudimentary workflow management and limited, retrospective, reporting added little value to the atrocious user experience.
These Generation 1 solutions were not architected to improve caregiver productivity and efficiency and, in fact, led to lower levels of productivity and caregiver satisfaction. Various broad-based studies have pegged the caregiver dissatisfaction rate for these solutions at over 90%.
Generation 2 EHRs – Guiding Care Decisions by Population Segment and Condition(s)
In response to the loud dissatisfaction and occasional revolt against Generation 1 solutions, leading EHR vendors have made progress with second-generation technology that represents a significant improvement over earlier systems. Human factors engineering has been applied to user interfaces to improve their intuitiveness and efficiency and lower training requirements. Multi-dimensional analytics tools are used to identify variance in key process and outcome measures. Clinical decision support—driven by episodic discrete data and logical rules/expressions—are used to coach care decisions in line with the validated standard of care for the combination of a given population segment and condition(s)/change of condition(s). Care coordination tools are employed to focus the full care team on the patient and progress against the care plan. Today, this combination of capabilities represents the state-of-the-art in deployed EHR technology.
Generation 3 EHRs – “Intelligent EHRs” Driving Personalized Care Decisions
Taking advantage of the huge increases in the capability of top-tier cloud computing environments to manage big data environments and the maturation of machine-learning toolsets, MatrixCare, and other innovation leaders are beginning to realize the incredible potential of EHRs to vastly improve the efficiency and quality of care. This new, third-generation level of capabilities is built upon full, longitudinal personal health records (PHRs) augmented with Internet-of-Things (IoT)/personal device data and inclusive of social, environmental, financial, behavioral, and genetic factors. On this person-centered foundation, leading innovators are leveraging the targeted application of deep machine learning to both structured and unstructured data in the comprehensive PHR to identify early indicators for changes in condition and opportunities for intervention.CareCommunity— a comprehensive PHR used to identify early indicators for changes in condition and opportunities for intervention. Click To Tweet
In one example of this targeted application of deep machine learning, MatrixCare, leveraging the Microsoft Azure cloud and the Azure Deep Machine Learning Studio, is training a neural network with millions of free-text progress notes analyzed against incidents of falls to identify key phrases/combinations of phrases that predict falls with a very high confidence level (90%+). An Azure web service will allow any progress note to be analyzed in real time using the trained neural network. The progress note entry/update function within MatrixCare can then be enhanced to call the Azure web service and, if indicated, pre-emptively suggest the invoking of the Johns Hopkins falls management protocols to lower fall risk, the gold standard for falls management in MatrixCare’s targeted senior population.
In another example, MatrixCare is leveraging the Microsoft Azure cloud and the Microsoft Azure Deep Machine Learning Studio to train a neural network with millions of anonymized Senior Living resident profiles and negotiated service plan features analyzed against resulting resident satisfaction and operator margin data. This trained network will then be able to suggest the optimal service plan for each specific resident in terms of projected resident satisfaction, length of stay, and operator margin. An Azure web service will allow any resident profile to be analyzed in real time using the trained neural network, and the Negotiated Service Plan entry/update function within MatrixCare will then present the optimal service plan based on the resident profile.
The Widening Performance Gap
Much as the Amazon online shopping experience learns about each shopper and makes shopper-specific recommendations to continuously improve both the shopper’s experience and Amazon’s financial and operational outcomes (e.g. average order size, revenue per shopper, etc.), “Intelligent EHRs” like MatrixCare learn about each person and make person-specific recommendations to improve both the person’s experience and the healthcare provider/operator/payor’s clinical and financial outcomes performance.
As America continues the migration to value-based care and long-term post-acute care (LTPAC) networks continue to narrow, outcomes performance will become the ultimate determinant of success or failure. LTPAC providers and operators will either suffer the same fate as the Main Street retailer who watched Amazon take their business by offering an ever-better, continuously refined experience with each successive engagement or equip themselves with the necessary tools to drive continuously higher outcomes performance.