What if there was a medical “crystal ball” that let hospitals glimpse three days into the future? Looking ahead would certainly allow physicians to improve their processes when treating the patient in the hospital setting. It also would give case managers a chance to improve their discharge processes overall.

It may not be magic, but researchers from Penn State and Geisinger Health System in Pennsylvania developed a model they say is ahead of its time because it does not focus solely on readmission risks. They named their model REDD: Readmission, Emergency Department, or Death.

To predict a patient’s need for follow-up care, the researchers gathered two years of clinical, administrative, and socioeconomic data from the patient’s previous hospitalizations.

During the six-month pilot, the team analyzed large amounts of these data to identify high-risk patients at the point of discharge, which may tell physicians which interventions may be able to limit adverse events.

“We knew we had a vast amount of data stored in our electronic health record. Geisinger and our department work diligently to discover insights in our data that can improve processes involved in patient care and ultimately improve patient outcomes and satisfaction,” says Eric Reich, manager of healthcare re-engineering at Geisinger.

Geisinger also used the REDD model to allow for targeted intervention, better care, and reduced costs and to avoid CMS 30-day readmission penalties.

Interventions were varied, including scheduling return appointments with primary care providers, educating patients about their prescriptions, having the clinical pharmacist review the discharge medication list, filling the patients’ prescriptions before discharge, and following up the next day with any patient discharged to a skilled nursing facility.

Reich says it made sense for his team to allocate resources for the research, design, development, deployment, and follow-up of a solution, leveraging patient data that could provide information on the risk of patient readmissions.

The results of the analysis showed that readmissions closer to discharge are more likely to be related to factors not identified at the time the patient is discharged, whereas readmissions after 30 days are environmental (e.g., poor shelter, no access to services, little social support, etc.).

“Scientific literature in the topic suggests that below 30 days, readmissions are usually due to hospital processes, while after this period, it is said readmissions are less correlated to the hospital care and more to other factors outside the hospital control,” says Reich.

“This is the main reason why CMS decided to use 30 days as the timeframe to assess preventable hospital readmissions, and after measuring, penalize such hospitals with excessive preventable readmissions,” he says.

If readmissions closer to discharge are related to factors that are not identified at the time the patient is discharged, how can that be fixed? That is not always an easy answer, as readmissions are heavily dependent on a series of patient, hospital, and socioeconomic factors.

“For instance, we know that extreme ages — either young or elderly — are associated with a higher risk of readmissions. Also, the readmission risk is heavily related to the main disease and their individual set of comorbidities,” says Reich.

The timing of critical follow-up care also has been shown to influence readmissions. Patients who visit their doctor or receive a follow-up phone call from a nurse soon after discharge are less likely to be rehospitalized.

Indeed, a new study of about 11,000 heart failure patients in the Kaiser Permanente Northern California health system discharged over a 10-year period found that this critical timing of follow-up matters. The study authors pointed out that it should be done within seven days of hospital discharge to be effective at reducing readmissions within 30 days.

To solve the issue of excessive readmissions within our country, the Geisinger researchers propose the following strategies:

  • further the understanding of individual factors that drive readmissions;
  • communicate and translate this research into systems capable of assisting hospitals in deepening their knowledge about their patients in relation to these outcomes;
  • develop a comprehensive approach to address all these factors to reduce readmissions.

The REDD model, as it was tested during the pilot study, was unable to give reliable estimates on the reduction in readmissions.

There are several barriers that may explain why they could not get concrete answers. Some of them are unique to the healthcare industry, while others are common across companies.

“From an empirical perspective, the risk score of readmissions informs the providers involved in the patient care about the risk for readmissions, and this team discusses options to reduce such risk with the tools they have at their disposal,” he says.

In a perfect world, he says, the risk score would be embedded into the electronic health record and a patient would not be discharged until the care team assessed the condition and additional interventions were considered and/or completed.

“Then, we would be able to track discharged patients with additional interventions allocated to them,” he says. “Finally, we would continuously monitor the performance of the model, update and refine it to ensure it reflected not only the most recent patient data trends, but also used the best algorithms available.”