Researchers find that functional status, rather than comorbidity, is a better predictor of whether someone will be readmitted to the hospital.
The way the Centers for Medicare & Medicaid Services predicts readmissions is likely neither the most accurate nor the fairest, researchers at Harvard Medical School claim.
A study published in the May issue of the Journal of General Internal Medicine found that functional status, rather than comorbidities, was a better predictor of whether someone would be readmitted to the hospital.
“This raises a question of whether Medicare is really using the best predictors to really understand readmission,” as well as questions about how fairly hospitals are being financially penalized, says principal investigator Jeffrey Schneider, MD, medical director of the Trauma, Burn and Orthopedic Program at Spaulding Rehabilitation Hospital in Boston and assistant professor of physical medicine and rehabilitationÂ at Harvard Medical School.
Jeffrey Schneider, MD
Schneider points out that CMS fined more than 2,200 hospitals a total of $280 million in 2013 for excess 30-day hospital readmissions, so having accurate readmission models is critical.
But the ones CMS uses “are not very good predictive models, and they have relied heavily on simple demographic data like age and gender and comorbidities,” he says.
Moreover, “there’s mounting evidence that function is a good predictor of all sorts of hospital outcomes.”
The researchers conducted a retrospective study of 120,957 patients in the Uniform Data System for Medical Rehabilitation database who were admitted to inpatient rehabilitation facilities under the medically complex impairment group code between 2002 and 2011.
Schneider says they chose to study this “medical complex” population “because it is heterogeneous and we think well-represents a wide swath of patients who are in a hospital for medical reasons.”
“Rehabilitation hospitals routinely collect functional measures and that data is available in a large administrative database,” he says. The researchers measured functional status using the Functional Independence Measure (FIM), which looks at 18 tasks such as eating, dressing, bathing, toileting, grooming, and climbing stairs. Each of the 18 items is rated on a seven-point scale from completely dependent on someone else for help to totally independent.
FIM data is collected on a patient’s admission to a rehab facilityâwhich is usually on the same day as their discharge from an acute care facility. “In that way it’s also a surrogate marker of their functional status when they left acute care,” he says.
Function or Comorbidity?
Researchers built models based on functional status and gender to predict readmission at three, seven, and 30 days, and compared them to three different models based on comorbidities and gender.
“We really just wanted to answer this question: If function was a better measure of readmission than comorbidity,” Schneider says. “We didn’t seek to build the best model.”
The researchers then determined the c-statisticâthe measure of a model’s overall ability to predict an outcome, which ranges from 0.5 (chance) to 1 (perfect predictor)âof the models.
They found that the model with gender and function was significantly better at predicting readmissions, Schneider says.
Models based on function and gender for three-, seven-, and 30-day readmissions (c-statistics 0.691, 0.637, and 0.649, respectively) performed significantly better than even the best-performing model based on comorbidities and gender (c-statistics 0.572, 0.570, and 0.573, respectively).
Even adding comorbidities to the function-based models didn’t help much, creating c-statistic differences of only 0.013, 0.017, and 0.015 for 3-, 7-, and 30-day readmissions, respectively, for the best-performing model.
‘It’s So Intuitive’
Why is function a good predictor? Schneider says it may represent something else, such as the severity of a patient’s illness. Cancer patients, for instance, have a wide degree of functional statuses depending on how sick they are. In this way, “it’s so intuitive” that function would be a good predictor of readmissions, he says. If you can’t care for yourself, you’ll likely end up back in the hospital.
In addition, “comorbidity is a fixed variable,” Schneider says, but function is not. And since function is a better predictor of readmission, even at shorter time intervals, assessing a patient’s functional status and doing things to improve it could be a way of reducing preventable readmissions, especially the three- and seven-day readmissions.
“Acute care hospitals are not routinely collecting a functional measure of their patients,” Schneider says. He also points out that recent research on functional interventionsâsuch as early mobilization in the ICUâin acute care hospitals is showing to improve patient outcomes.
“I think the next wave for hospitalsâ¦ is [thinking about] how to make use of this information,” Schneider says, by piloting functional interventions and determining functional measures at discharge to help with risk-stratifying for readmissions.
On a larger scale, there’s also the policy perspective that CMS’s readmissions models aren’t as good as they could be. Schneider says he and his colleagues are conducting another, even larger study, using the same framework, but looking at but all patients in a rehab hospital, not only at medically complex ones. He says it hasn’t been published yet, but the findings will be pretty similar.
“I think it’s really worthwhile,” he says.
To read the original article on HealthLeaders Media, click here.