A recent study by researchers at the University of California, San Francisco demonstrates how difficult it is to reduce readmissions—especially given the diverse patient populations served by safety-net hospitals with limited resources. In a telephone interview, Margot Kushel, M.D., professor of general medicine at UCSF and a senior author of the study, suggests some promising approaches that hospitals can consider, and says that data analytics can go a long way in helping help hospitals identify at-risk patients.
As noted by HCI Editor-in-Chief Mark Hagland in his blog post, recently released federal records show that Medicare is fining a record number of hospitals for having too many patients returning within a month for additional treatments.
The UCSF researchers note that as many as 30 percent of hospitalized elderly patients are readmitted within 30 days, and many of those readmissions are preventable. To learn if an in-hospital intervention program can have a beneficial impact, they looked at 700 adults age 55 and older who spoke English, Spanish or Chinese (Mandarin or Cantonese), and were being discharged back into the community. The patients had a high rate of multiple co-morbid medical conditions and low health literacy.
The study—which has been published in the Oct. 7 issue of the Annals of Internal Medicine—was carried out at a Northern California acute-care safety-net hospital in an urban setting. It compared a nurse-led intervention program designed to reduce readmissions among ethnically and linguistically diverse older patients to the hospital’s conventional discharge process.
According to Kushel, the researchers were interested in looking at a poverty population in part because hospitals that serve the urban poor have higher rates of readmissions, and there are potentially different factors that lead to those readmissions than the wider population. “Generally those people were pretty sick, with a lot of cardiac disease, pulmonary disease, and they were on 10 or 12 medications, representing the complexity of their illnesses,” she says. Many of the patients were in the 55 to 65 year-old age range, but in terms of their medical burden and functional status, they had more in common with patients in their 70s and 80s in non-safety-net hospitals, she adds.
Although the intervention program was built on already robust standard discharge procedures, it had no impact on readmission rates on the patient population. Thirty days after discharge, the readmission rate of patients who went through the intervention program were the same as those who went through the usual discharge process—about 15 percent. (That’s lower than the national average readmission rate for Medicare patients. Kushel notes that the baseline readmission rate, which was lower than the researchers anticipated, was in part due to the nature of the study: it intentionally excluded patients who didn’t have a phone to follow up with, so homeless and those with marginal housing were not part of the studied population.)
The nurse intervention used in the study was based on an intervention that had been used successfully at in a Boston hospital, which has been used in 200 to 300 hospitals across the country, Kushel says. UCSF’s version was designed not to be resource-intensive, so it could be useful for safety-net hospitals that often struggle with limited resources, she says.
Kushel says that traditional discharge processes are already thorough at UCSF Medical Center. Bedside nurses review recommendations of the medical team with the patient and review follow-up appointments; a 10-day supply of medications will be supplied to patients if there is a concern about their getting access to prescriptions outside the hospital. Social workers are on hand to see if there are services that will be needed in the home.
The nurse intervention program added to those procedures. Nurses spoke to the patients in their native language; they were trained in motivational coaching, and spent nearly three hours with each patient on average. Soon after admission to the hospital patients were informed about their diagnosis, treatment plan, and medications, and were coached about any problems that were anticipated. Patients were given a to-do document, written at a basic reading level in their language and including simple graphics.
Following their discharge, a nurse practitioner called the patients after three days and again after seven to 10 days as follow-up. The intervention also had a “warm line,” a special phone number they could call with any additional questions for 30 days after their hospitalization. Kushel says that about 70 to 80 percent of the patients talked to the nurse practitioner at least once, although fewer used the 30-day call-back number.
“The intervention added components onto an already robust discharge plan, and didn’t do anything other than, maybe, it might have even slightly increased ED visits,” Kushel says, adding that many patients followed the advice of the caregivers after discharge.
Kushel has some thoughts about alternatives that could be worth exploring. One is the use of health coaches or peer counselors to spend time with patients after they leave the hospitals, to try to overcome non-medical barriers such as lack of transportation or housing issues. “That is worth considering. We’ve all been trying to bolster the medical home,” she says.
She cites another study in which patients were asked open-ended questions what affected their health after they left the hospital. Interestingly, almost none of the patients talked about classic medical issues. Instead, the talked about practical issues, such as difficult living circumstances, isolation, or the lack of anyone to check up on them. “It raises the question of whether, when people are in the hospital and they already feel lousy, it is the best time we should be doing it, or whether we should be bolstering the people once they have left the hospital,” she says.
Kushel believes that data analytics can help decrease readmissions by allowing provider organizations do a better job of predicting which patients are most likely to be readmitted; and then focusing their efforts on the percentage of the population that should be getting more intensive services. “Provider judgment has not been very good at predicting who is going to come back to the hospital,” she says. “That is where data analytics has a role: to do a better job of risk stratifying, to figure who needs the most help, and not waste precious resources on someone who will probably be fine anyway.”
She adds that while safety-net hospitals across the country have a lot in common, with diverse populations, literacy issues and language barriers, each hospital is a little different from each other as well. UCSF researchers adapted popular intervention that was studied in a safety-net setting, “But their system was different enough than ours—their population was younger and we chose older people, because that was whom we were concerned with,” she says.
One of the conclusions of the study is hospitals need to take a refined approach in evaluating programs before implementing them. After all, hospitals differ, both in their internal processes and resources, and in the patient populations they serve. What was the lesson learned from this project? “It told us we need to be careful about adapting other interventions without studying them, without checking to see if it works. It was a great program and it didn’t work, and every time you do that, there’s an opportunity cost. We need to spend resources on something that does meet our needs and does work,” Kushel says.
