Only 1/3 of hospitals use predictive analytics, even though integrating patient information, data mining, stats and machine learning are the industry’s prized tools for rolling out population health management. Patient information today stems from a variety of records and resources, and while population health management is critical to improving clinical/financial outcomes for patients and for taking the guesswork out of healthcare, implementation of predictive analytics begins with pushing providers to performance-based care and standardization into workflows.

Informing redesigned and standardized processes with better, timelier insights provides tremendous opportunities to
improve the efficacy of care, the effectiveness of healthcare resources and, ultimately, the value achieved. Predictive
technologies are capable of providing more integrated, action-oriented care management that incorporates feedback
and metrics to drive measurable improvements in patient outcome. A case in point is the application of predictive insights by hospital pharmacists to help standardize and optimize the delivery of medications for high-risk patients prior to hospital discharge.

Poor medication adherence costs the healthcare system at least $100 billion a year in additional doctor visits, emergency department visits and hospitalizations, and further claims the lives of 125,000 Americans annually. Improving medication adherence can be difficult because there are so many potential points of failure, including medication fill errors, patients taking the wrong medications, duplicative therapy and patients not following medication instructions. Leading hospitals have realized that a primary opportunity to impact medication-related, avoidable readmissions is squarely within their control: delivering medications to high-risk patients at bedside prior to discharge, sometimes called “meds to beds.”

Identifying at-risk patients is the first challenge. Almost all hospitals employ some kind of risk score to identify patients who may be at risk for readmission. However, for an effective data-driven “meds to beds” program, hospitals must identify patients who are vulnerable to readmission because of their medication risk. Most risk prediction models in use by hospitals do not include risk factors that are specific to medication adherence, such as gaps in medication fill patterns prior to admission, the numbers of concurrent medications, newly prescribed medications for chronic conditions, medications that are difficult for patients to manage and social determinants.

With hospitals under constant pressure to contain staffing costs, improved patient targeting ensures that the staff devoted to a “meds to beds” program reach the patients who most need it.

A “meds to beds” program that leverages a data-driven approach targeting high-risk patients can be staffed to engage only 30% of the inpatient population but impact more than 60% of the hospital’s total readmission risk. The typical all-cause readmission rate for a hospital across all inpatient stays may be around 9%, whereas the top 30% of patients at highest risk for medication adherence failure can have readmission rates of more than 20%, if they are not getting their medications prior to discharge.

Recently, a case study by University of Tennessee Medical Center demonstrated that by using a data-driven approach to
deliver medications to high-risk patients prior to discharge, it was able to reduce 30-day readmissions by more than 20%
compared to patients of similar risk who did not have their medications delivered.

Predictive models can improve the efficiency of the process for the hospital pharmacy. By excluding patients likely to be discharged to locations other than their homes, staff outreach efforts are focused on those patients who will be responsible for their own prescription fulfillment post-discharge. Incorporating predictions of patient discharge dates can be used to sequence outreach efforts and coordinate the bedside delivery of medication. Analytics can also identify opt-out patterns, enabling refinement of outreach and enrollment workflow, quantification of missed opportunities and evaluation of pharmacy staffing decisions.

“Meds to beds” is a simple, yet powerful, example of the impact of predictive analytics on a redesigned care process. At-risk patients start their post-discharge recovery with the right medication, hospitals avoid preventable readmissions and onsite pharmacies are able to improve gross margins and increase staff effectiveness.

Smiley is the CEO of Loopback Analytics Dallas. Founded in 2009, Loopback Analytics is a pioneer in empowering health systems and post-acute care organizations to more effectively manage care transitions and reimbursement challenges in a “pay-for-outcome” environment.

The company’s comprehensive management platform helps identify at-risk patient populations, match interventions and measure efficacy to improve patient, clinical and economic outcomes. Visit