The Value-based Care (VBC) healthcare delivery model was codified almost 15 years ago. A paradigm shift from the ‘fee-for-service’ model, VBC, meant that providers are compensated based on the health outcomes of their patients. The traditional ‘fee-for-service’ model often emphasized volume and individual procedures instead of taking an integrated approach to health and wellness. Consequently, this led to fragmented care quality, driving up costs. In contrast, the VBC model emphasizes evidence-based outcomes that capture how provider interventions have helped patients improve their health cost-effectively.
Since the VBC model prioritized patient outcomes and cost-effectiveness, providers are encouraged to focus on preventive care and adopt a more proactive approach instead of simply providing care for the sick. This shift will require healthcare organizations to leverage technology and data analytics to pursue multi-dimensional outcomes, such as identifying high-risk patients, improving care coordination, and more. For the VBC model to work better, providers need to be able to act with consideration for their patients’ entire history, identify opportunities for preventive care and monitor their health status on an ongoing basis.
However, providers are hamstrung by formidable challenges. For one, is the software that providers use built to function in line with the fundamentals of VBC? Unfortunately, most software was built in pre-VBC times, leaving much to be desired. So, to identify and document real risks and initiate preventive actions, one needs to manually decipher insights from the current EHR (Electronic Health Records) platforms. The exercise is time-consuming and prevents timely interventions, adding to the physicians’ burdens.
AI-based predictive models: Force-multipliers in Value-based Care
Artificial Intelligence (AI) can be a force multiplier in this endeavor. AI-based predictive models can offer actionable insights and empower healthcare organizations to arrive at data-driven decisions, make efficient resource allocations, and deliver care based on patient personalization. AI in Value-Based Care can leverage past data and analytics to predict future outcomes by uncovering meaningful relationships between discrete data points. Let us consider three real-world implementations of AI-based predictive models:
1. Improving Early Disease Detection & Intervention
A key driver in the success of the VBC models is the early identification of high-risk patients. These patients and their history may contain several clues that could foretell the development of chronic health conditions in the future. AI-based predictive models play a vital role in this early-stage identification by processing vast amounts of patient data – EHRs (Electronic Health Records), Patient-Generated Health Data (PGHD), lab test results, etc. – and identifying patterns and risk factors.
Early identification can help providers launch targeted interventions like screenings, and medication, leading to complete prevention or, at least, mitigation. Such efforts not only improve patient outcomes but also reduce long-term healthcare costs.
Example – Early Identification of Heart Failure with an AI model
An AI-enabled electrocardiogram (ECG) system has been used to identify patients at a high risk of heart failure correctly. The system correctly classified long-term cardiovascular outcomes in patients with normal Ejection Fraction (a measure of the percentage of blood leaving the heart each time it squeezes).
The AI model was developed from ECG data from 61,525 patients. Further, the model was internally and externally validated with data from 3,810 and 5,760 patients. These results were published in Frontiers – a leading research publisher and open science platform.
2. Optimizing Resource Allocation to Boost Operational Efficiency
For VBC models to succeed, resource allocation should be efficient, both from a perspective of optimal outcomes and cost efficiency. In this regard, AI in Value-Based Care can analyze historical data and patterns by blending those in with current factors and forecast demand to optimize operations. To forecast patient flow, these predictive models can analyze factors such as patient demographics, seasonal variations, appointment history, etc. These insights can help providers allocate staff, equipment, and facilities effectively.
Example – Predictive model used to forecast hospital bed demand
An AI tool has been used by a leading hospital in the UK to successfully predict how many patients from Accidents & Emergencies would require to be admitted to the hospital. The study showed that the tool was more precise than previous benchmarks. The AI tool relies on historical data and seasonal trends while accounting for patients yet to arrive at the hospital.
While the tool does not churn out a single-figure prediction for the day, its predictions are in the form of a probability distribution for the number of beds required in the next four and eight hours. The forecasts from the AI tool become available four times a day when they are emailed to hospital administration for the next steps. The study results have been published in Nature – a digital medicine journal.
3. Optimizing Risk Adjustment with Machine Learning (A HealthAsyst success story)
Traditional actuarial methods depend on population-level data. These models work with broadly defined risk categories to estimate the cost of care for different individuals and groups. However, this leaves out a lot of individualized nuances. AI-based models can process vast patient-specific data to create detailed risk profiles. Consequently, there is more scope for identifying subtle patterns and factors contributing to individual-level health risks. Therefore, integrating AI into actuarial practices can help design tailor-made interventions, improve financial forecasting, and improve patient outcomes and effective risk management.
Example –Improvement of Personalization and Precision of Health Risk Scores using Machine Learning
HealthAsyst collaborated with a US healthcare analytics solutions provider to develop a risk score model using Machine Learning (ML) techniques. The result was a dynamic risk score for every individual. This risk score was more accurate than traditional actuarial practices and allowed for greater personalization and insights at a member level.
Supercharge your Value-Based Care Practice with the HealthAsyst Advantage
At HealthAsyst, we follow developments in the med-tech space with keen interest. We have witnessed interesting use cases of AI in Value-Based Care that have had a transformative effect with regard to better outcomes and superior care quality. Therefore, we understand the immense potential of this synergy. Our past work experience in this regard, deep knowledge of the healthcare space, combined with 24+ years of software engineering excellence, puts us in an advantageous position to serve clients. If you want to bring greater efficiency to your Value-Based Care Practice, reach out to us at itservices@healthasyst.com
