Over the past several years, nearly all of the leading health systems in the United States have adopted some form of AI, such as machine learning (ML) or natural language processing (NLP), to assist in managing healthcare operations. The unprecedented growth and development of AI has been instrumental in transforming health system administration, healthcare data analytics and patient diagnosis and treatment.
Despite this growth, health systems have yet to maximize the potential of AI to improve upon load balancing and optimization of patient throughput. With staffing shortages and lack of resources at an all-time high due to the COVID-19 pandemic, it is vital for health systems to use their resources in the most effective way possible. By applying ML and predictive modeling, health systems are able to create the optimal match between patient and provider in any scenario, from low-acuity urgent care to specialty care. This not only optimizes their own resources but also ensures outcomes that lead to high patient satisfaction and retention.
Using Predictive Modeling to Match Patient and Provider
All consumers of healthcare are familiar with the challenge of deciding on a health system and choosing a provider to call their own. These choices are often made based on random selection or arbitrary criteria, and there is no guarantee that the provider selected will be the best fit. We’ve all experienced the frustration of not knowing which provider to see or, even more maddening, waiting an inordinately long time for an appointment.
Fortunately, there is another method of matching patients and providers that can take the guesswork out of these decisions. Health systems can apply machine learning algorithms that consider patient need, patient convenience and provider capacity, among other factors, that ultimately create the optimal match to balance the load for health systems and ensure exceptional patient care.
Information entered into the predictive model can include (but is not limited to) the reason for a patient’s visit, symptoms, geographic location and general demographic information. The model can then identify how the patient should best be treated (telehealth or in-person) and which providers can treat their needs. Additional factors are also taken into account including patient satisfaction scores, provider scores for treating certain symptoms and patient convenience, such as distance and wait times. Using these data sets, the model outputs a match that optimizes these factors, ultimately directing the patient to a booking that results in a satisfactory experience for both the patient and the health system.
In actual usage within health systems, this method for matching patient to the provider is still in its early stages – the predictive model currently factors in patient symptoms, basic provider data and convenience based on time and location. However, even in its limited usage so far, health systems have already seen improvements in NPS scores and other key metrics. Soon, predictive modeling will be able to deliver matches that are even more finely tuned to the patient’s individual preferences. Data-driven intelligence companies plan to expand the capabilities of the model to account for patient morbidities, medical history and past experiences with different health systems.
Customizing These Tools for the Health System’s Needs
Patient satisfaction is the single most important factor for health systems, but health systems cannot ensure overall patient satisfaction and retention without optimizing load balancing among practitioners. Load balancing is especially critical in today’s market where we continuously see staffing shortages and patient overload. If health systems are overloaded, patients will have worse outcomes and decreased satisfaction with their care, and the health system’s NPS scores will ultimately suffer.
Fortunately, data-driven AI and predictive modeling can be used to keep the health systems running smoothly while optimizing outcomes for patients. The AI can not only help the health system load balance their providers across departments and facilities, both in-person and digital, but also better route patients to the appropriate level of care.
Not all health systems have the same needs or preferences when it comes to seeing patients. Some may want to minimize turnaround time, getting patients in and out of the door as quickly as possible. Others may want to achieve the highest patient satisfaction scores by ensuring that a patient is cared for by the best possible provider, regardless of wait time. While the predictive model is ultimately driven by the data, the algorithms can be fine-tuned so that the health systems achieve their intended results. This ability to tailor the model for specific provider and patient needs offers flexibility to health systems and ensures successful outcomes, even as needs fluctuate over time.
Three Use Cases for Provider-Agnostic Care
Patients currently seek most of their care needs through a quasi “hunting and pecking” mechanism where they research a specific health system, choose a physician and book the earliest available appointment. However, this approach can be time-consuming and stressful. Doing one’s due diligence to compare health systems and providers is not an easy task, but understanding the alternatives that one has as a patient is non-trivial. With an inconsistent set of resources, it can be an overwhelming and confusing process, especially for someone who does not often seek medical care. It also takes up valuable time that a patient could spend receiving treatment – especially critical when a patient needs care urgently.
While this provider-specific model of finding care may suffice, it does not guarantee that satisfaction or efficiency are being optimized. Meanwhile, the increasing usage of AI and ML in healthcare is likely to further sway the shift towards provider-agnostic care, where patients spend less time searching for the right provider and instead are promptly matched to the best provider through predictive modeling. Taking a provider-agnostic approach helps minimize the laborious research and even guesswork required to select a provider. This model of healthcare delivery better aligns with the needs of both the patient and the health system to ensure better outcomes.
The majority of healthcare can be classified into three scenarios, and provider-agnostic care can be applied in all three. For low-acuity care situations, such as a sore throat, a provider-agnostic approach works best. There is no need for the patient to choose ahead of time what provider they will see as most providers are able to treat such low-acuity conditions. In this case, a provider-agnostic approach allows the patient to receive care as quickly as possible.
For specialty care, such as cardiology, a provider-agnostic approach is most effective for patients to receive the best care for their individual needs. By using predictive modeling that accounts for patient history, preferences and provider data, the confidence level of ensuring an ideal outcome is greatly improved in contrast to a patient selecting a provider individually.
The final scenario for provider-agnostic care is in primary care. In this case, though the patient will continue to see the same provider on a routine basis, a provider-agnostic approach is beneficial to get the patient in the door. Like in the case of specialty care, predictive modeling ensures the best match between a patient and a primary care physician for their first appointment, factoring in a range of aforementioned variables. Subsequently, the patient will continue to see that provider due to the quality of care and alignment between patient need and provider delivery. This ultimately contributes to higher patient retention for providers.
The Future of Healthcare Delivery
If it seems like provider-agnostic care would be a major shift from the way we currently find healthcare, it’s not. A provider-agnostic model is already the standard in most urgent care and retail clinics, and these venues of care have been successful in treating patients effectively and efficiently. In fact, many patients prefer provider-agnostic care because it is more convenient than having to choose a provider and book an appointment ahead of time. In a time when the convenience of care is extremely important to many consumers, taking a provider-agnostic approach can help health systems keep patients healthy and happy.
While most health systems have already adopted AI and ML to a certain extent, it’s time for health systems to maximize the potential of these tools when it comes to load balancing and optimization. Not only will it improve the schedules and job satisfaction of providers, but it will ensure that patients receive the best care possible for their individualized needs. As data-driven intelligence companies work to further develop this technology, the benefits that can be achieved by employing predictive modeling will only continue to expand.
About Chad Waldman
Chad Waldman is the Vice President of Engineering at DexCare, setting the technology strategy and building respected teams to implement DexCare’s vision. With more than 20 years under his belt, Chad has been leading engineering teams with a focus on greenfield technology in both the education and healthcare spaces. Prior to DexCare’s spin-out, Chad worked at Providence St. Joseph, driving much of the solution that makes up DexCare today. He has a passion around fostering talented individuals, working hands-on with coworkers and end users, and jumping at challenges.