Making Better Use of Health Care Data
At Sanford Health, a $4.5 billion rural integrated health care system, we deliver care to over 2.5 million people in 300 communities across 250,000 square miles. In the process, we collect and store vast quantities of patient data – everything from admission, diagnostic, treatment and discharge data to online interactions between patients and providers, as well as data on providers themselves. All this data clearly represents a rich resource with the potential to improve care, but until recently was underutilized. The question was, how best to leverage it.
While we have a mature data infrastructure including a centralized data and analytics team, a standalone virtual data warehouse linking all data silos, and strict enterprise-wide data governance, we reasoned that the best way forward would be to collaborate with other institutions that had additional and complementary data capabilities and expertise.
We reached out to potential academic partners who were leading the way in data science, from university departments of math, science, and computer informatics to business and medical schools and invited them to collaborate with us on projects that could improve health care quality and lower costs. In exchange, Sanford created contracts that gave these partners access to data whose use had previously been constrained by concerns about data privacy and competitive-use agreements. With this access, academic partners are advancing their own research while providing real-world insights into care delivery.
The resulting Sanford Data Collaborative, now in its second year, has attracted regional and national partners and is already beginning to deliver data-driven innovations that are improving care delivery, patient engagement, and care access. Here we describe three that hold particular promise.
Developing Prescriptive Algorithms
With any chronic condition, inattentive management and inconsistent follow-up care increase the risk of urgent or emergency care visits as well as unplanned inpatient admissions. Identifying patients most at risk for these types of visits, and identifying the clinical and behavioral characteristics associated with them, can help frontline clinicians provide targeted management. In partnership with the University of North Dakota School of Medicine’s Population Health Department, we developed an algorithm that can predict diabetic patients’ risk of unplanned medical visits. This algorithm, leveraging advanced machine learning analytics, can predict with nearly 80% certainty the likelihood that a given diabetic patient will incur a costly and unwanted unplanned visit. To make this prediction, the algorithm analyzes smoking status, BMI, and current number of diagnoses on the patient’s “problem list,” all of which are amenable to intervention. The algorithm is currently being validated in pilot clinics with the goal of then scaling it up for enterprise use.
Augmenting Patient Engagement
The degree of patients’ engagement in their own health care is a significant predictor of their health care behaviors and, ultimately, health outcomes. Unfortunately, measuring engagement is difficult because it’s time consuming, generally has low participation rates and patients who are more engaged tend to be more willing to take a survey, potentially skewing the data. Given these barriers, health care systems are left with little information about how patient engagement drives care utilization and behavior. To tackle this issue, Sanford partnered with investigators from the Data Science and Health Sciences schools at South Dakota State University. SDSU’s team developed a patient-engagement score algorithm for people with multiple chronic conditions using pre-existing patient behavior data. The resulting engagement scores predict the likelihood of patient emergency department visits and hospitalization more accurately than previous methods. Key data informing the score include prior online health portal usage and likelihood of showing up at appointments, both of which can be improved through targeted intervention. Next steps include evaluation of short- and long-term impacts of these interventions on ED use and hospitalizations and the resulting impact on outcomes and costs.