A little more than a decade ago, the Institute for Healthcare Improvement established healthcare’s Triple Aim, providing health systems with three ultimate goals:
- Improve the patient experience.
- Reduce the costs of healthcare.
- Improve the health of populations.
As the healthcare industry continues to adopt value-based models, managing population health has become more important. Health systems are placing more emphasis on effective care management—a team-based, patient centric approach is designed to improve care while reducing the need for medical services—in their efforts to achieve the Triple Aim, with substantial financial incentives to do so.
Traditional care management efforts have seen some early success, but they aren’t taking advantage of all the rich patient data available in medical charts. Providers and payers are investing in machine learning technology, which has the potential to transform unstructured healthcare data like doctor’s notes into actionable insights to guide more effective care management activities.
What is machine learning?
Machine learning is a computational technique that uses algorithms to learn from historical datasets and make accurate predictions about the future. These algorithms are trained to look at specific variables, then develop models that predict events or find patterns in new data.
In healthcare, machine learning models can be applied to a wide variety of datasets. For example, machine learning could be used to predict which hospital inpatients are most likely to be readmitted, flag patients who haven’t received appropriate preventive screenings, or identify patterns in prescribing behaviors within a specific medical system.
Machine learning empowers care managers to make more intelligent suggestions.
Care managers play a vital role in ensuring that patients receive appropriate and high-quality care. Focused on individual patients, care managers help identify specific care needs and route patients to the right services. The more information care managers can obtain on their patients, the better equipped they’ll be to guide people appropriately throughout the healthcare system.
As more patient data becomes available than ever before, it’s imperative that care managers are able to leverage machine learning capabilities to surface relevant trends and insights.
Machine learning is able to sort through the massive volumes of data quickly so care managers are able to efficiently extract useful intelligence from the data. Equipped with this information, care managers can provide suggestions tailored to the unique needs of each patient. As a result, patients receive more personalized care and positive outcomes, and care managers are better positioned to help more patients.
Machine learning better identifies potential chronic issues or comorbidities.
Each year, the U.S. healthcare system spends approximately $1.65 trillion on treating patients with one or more chronic diseases, with expected expenditures potentially reaching $6 trillion by 2050. With chronic disease affecting 133 million Americans (40% of the nation), the need for disease management and prevention is immense.
Based on a study from the U.S. Agency for Healthcare Research and Quality (AHRQ), an estimated 4.4 million hospital admissions totaling $30.8 billion could have been prevented. Of that $30.8 billion, nearly half of the costs were attributed to heart disease and diabetes complications – both chronic diseases.
Through machine learning, chronic care delivery can shift from reactive to proactive, focusing less on a one-size-fits-all approach and more on individualized care. By leveraging patient data, machine learning algorithms can quickly determine personalized interventions for each patient, including everything from increased monitoring to specified treatment plans. With these steps in place, interventions can occur before chronic issues appear or become acute that result in costly hospital stays, preventable emergency care, and an overall lower quality of life.
For example, Boston University’s Center for Information and Systems Engineering conducted a study that used patients’ electronic health records (EHRs) and machine learning to predict hospitalizations due to diabetes and heart diseases. In the study, hospitals provided patients’ anonymized EHR data, which included demographics, diagnoses, admissions, procedures, medications prescribed, and lab results. Machine learning algorithms were then used to process the large set of data and predict who might need to be hospitalized. Boston University was able to predict hospitalizations due to heart disease and diabetes about a year in advance with an 82% accuracy rate. These predictions allowed them to intervene sooner by treating the disease in an outpatient setting and avoiding costly hospitalizations or emergency room visits.
Machine learning is a worthwhile investment.
As value-based care increasingly becomes the primary focus of today’s healthcare industry, technologies like machine learning are emerging as invaluable resources for healthcare organizations to provide more personalized care management efforts.
Healthcare organizations have been reticent to adopt new and complex technologies like machine learning, but the unique opportunities to leverage the influx of big data through machine learning further cements its value to providers. Whether through internal analytic teams or external technology vendors, machine learning is a necessary and valuable investment. Late adopters will miss out on opportunities to leverage their patient data to drive more intelligent, effective patient care.
Machine learning applications have the potential to save the healthcare system billions of dollars and simultaneously help provider systems enhance care coordination, eliminate duplicative care, and help patients stay healthier. These financial and clinical benefits stretch across the entire healthcare spectrum and can benefit providers, health plans and patients alike.