Home Analysis Think you can predict readmissions? Think again

Think you can predict readmissions? Think again

June 2011

Published in the June 2011 issue of Today’s Hospitalist

HOW ACCURATELY CAN YOU PREDICT which patients will likely be readmitted to the hospital? If you’re like most physicians, you’re probably pretty confident that you can spot patients who will run into trouble and be back to see you. But you’re also probably wrong about just how well you can identify readmissions.

According to a study posted online in March by the Journal of General Internal Medicine, hospital-based clinicians are terrible at predicting patients who are likely to bounce back within 30 days. That group includes physicians, nurses and case managers.

“When it came to predicting readmissions, there was no statistical difference between doctors’ judgment and flipping a coin,” says Nazima Allaudeen, MD, lead author of the article and a hospitalist at the VA Palo Alto Healthcare System.

Physicians’ inability to predict which patients would likely be readmitted was particularly important, the study found, because of the volume of readmissions it uncovered. While previous studies have found that about 17% of hospitalizations lead to a readmission, Dr. Allaudeen’s research learned that nearly one-third of the 160 patients in the study returned within 30 days.

High readmission rates
When it comes to why so many patients bounced back, Dr. Allaudeen explains that the urban population she studied faced not only clinical problems, but social issues as well. “These were incredibly sick patients,” she says, “and they were socially and economically disadvantaged.”

The study methodology may have also had something to do with the high readmission rate. Instead of relying only on billing information to find readmissions, the team checked the records of other local hospitals to find patients who went to another facility for follow-up care, as is sometimes the case with homeless individuals. Researchers also called patients and their family members to determine if patients had been readmitted.

While differences in methodology may have led to higher-than-expected readmissions, they don’t explain the lackluster performance of physicians in predicting patients who would bounce back.

“Physicians came the closest in guessing the overall percentage of patients who’d be readmitted,” says Dr. Allaudeen, “but not which patients.”

Who was missed
Why were clinicians so clueless at predicting readmissions? Dr. Allaudeen gives examples to illustrate the re- admissions that physicians in the study “missed.” There was the patient discharged with a Foley catheter who came back with a urinary tract infection, for instance, and the patient discharged on antibiotics who was re- admitted with C. difficile colitis.

While it doesn’t take a medical degree to predict that a Foley catheter might lead to a UTI, Dr. Allaudeen says that predicting this complication in a patient with multiple medical problems is like spotting a needle in the proverbial haystack. “We know that readmission is a possibility in these cases,” she says, “but there are so many of these maybes, and we are guessing how they will add up.”

What she found particularly interesting was that most readmissions were related to patients’ admitting diagnoses or a related complication.

“You know that C. diff is a complication of being on antibiotics,” Dr. Allaudeen points out, “but the chance of getting this infection is small, and it’s one of 20 different possible complications. There are so many things that can fall apart. The question is which one is it going to be and who is it going to be.”

The way Dr. Allaudeen sees it, the study offers mixed news. “On one hand, it’s kind of freeing to know that readmission is fairly common, so I take it a little less personally,” she explains. “But on the other hand, you realize how vulnerable these patients are.”

No reliable tools
When Dr. Allaudeen presented the study’s results to physicians, most were surprised.

“I would ask physicians if they thought that they were good at predicting readmissions,” she says. “A lot of people would say, ‘This patient is totally coming back,’ or ‘This patient is set and not going to come back.’ We think we can make an educated guess.”

She notes that physicians like hospitalists, who treat patients who present with a broad range of problems, have an especially tough time identifying readmissions. “It’s particularly difficult to predict readmissions with medicine patients because they have so many comorbidities,” Dr. Allaudeen says.

The study raises an interesting question: Is it possible for physicians to accurately identify patients who are at risk of readmission? “I don’t think we have the tools right now to do that,” says Dr. Allaudeen. “Fairly sophisticated analyses are coming up with the same conclusion, which is that we are not reliably able to predict who’s coming back.”

While several studies have looked at a variety of strategies to predict readmissions, she explains, none has been particularly successful. Her research team focused on clinicians’ predictive abilities, but algorithms and tools that have used a more mechanical approach to identify risk factors for readmission have fared almost as poorly.

Everyone is high risk
So what does this mean for hospitalists’ efforts to build a better discharge? Dr. Allaudeen says that she now tries to see every patient as a potential readmission and then act accordingly. Instead of trying to identify those patients at high risk for being readmitted, she now assumes that everyone is high risk.

“I look at patients from the angle of what could go wrong for this patient and try to anticipate that,” she says. “I am very particular about discharge planning and follow-up appointments, follow-up labs and medication reconciliation. I am more likely to order labs if patients have medication changes.”

Because she runs a discharge clinic, Dr. Allaudeen sees first-hand how the best-laid follow-up plans don’t always come to fruition in the real world. “We underestimate how difficult some medication regimens are,” she says, “so I do an in-person medication reconciliation with my patients.”

While conducting such thorough discharges definitely takes more time, Dr. Allaudeen says that a little organization “and a good support staff “help minimize the impact on her schedule. Her hospital has an excellent discharge coordinator who makes sure that every patient has a follow-up appointment within no more than 10 days of discharge.

Besides, she adds, her team’s data show that hospitalists can’t afford to take shortcuts at discharge. “This study tells me that we can’t predict who’s going to come back, so we need to treat all patients as if they may be readmitted. We can’t decide that this patient requires follow-up but another patient doesn’t. ”

“We need to apply interventions across the board, not try to guess who needs them and who doesn’t,” Dr. Allaudeen points out. The problem with that approach is that “it has implications for how we use resources.”

Edward Doyle is Editor of Today’s Hospitalist.