Developing a Model to Predict and Prevent Readmission in Patients With Cirrhosis

Patient-reported outcome measures may have a small impact on predicting which patients with cirrhosis will be readmitted to the hospital.

Patient-reported outcome measures (PROMs) may help to improve prediction models for which patients with cirrhosis with be readmitted to the hospital, according to a study published in Clinical Gastroenterology and Hepatology.

Patients with cirrhosis have high rates of early, often preventable, readmission following hospital discharge. In addition to being expensive, these readmissions are associated with poor outcomes. However, interventions to reduce these readmissions have shown limited evidence for efficacy, underlining the need for risk stratification tools to refine and target interventions to the patients who will benefit the most.

“Multiple models have been developed to predict readmission, with disappointing results,” the authors wrote.

The authors hypothesized that socioeconomic status, functional status, and quality of life would enhance a prediction tool and sought to develop a model using these factors to predict readmissions in hospitalized patients with cirrhosis. They performed a prospective cohort study of 654 adults with cirrhosis who were admitted to Indiana University Hospital between June 2014 and March 2020.

They collected functional status using the functional status questionnaire and disease-specific quality of life using the chronic liver disease questionnaire (CLDQ). The patients were asked to reflect on the time prior to their admission when answering the questionnaires.

The researchers built 5 models:

  • Model 1: only used electronically available variables
  • Model 2: added clinically available variables to model 1
  • Model 3: added socioeconomic variables to model 2
  • Model 4: added functional status to model 3
  • Model 5: added quality of life to model 4

More than one-third (35.3%) of patients were admitted in the prior 30 days, 79.8% had a history of ascites, and 70.2% had hepatic encephalopathy (HE). The median Charlson Comorbidity Index score was 6; 41.9% had diabetes, and 23.5% had chronic kidney disease.

HE was the most common reason for admission (21.9%), followed by ascites/volume overload (20.5%) and abdominal pain (17.9%). Half (50.6%) of the patients were unemployed, 69.9% had a high school education or less, and 27.2% had saved less than 6 months of expenses.

More than one-third (37.8%) had a readmission: 22% for HE, 16% for acute kidney injury/electrolyte disturbance, 13% for gastrointestinal bleeding, and 10% for ascites/volume overload. Thirty-nine patients (6%) died within 30 days of discharge.

The factors associated with readmission in univariate analysis were cerebrovascular disease, prior admissions and emergency department visits, ascites, Model for End Stage Liver Disease (MELD) score, albumin, hemoglobin, use of an antibiotic for HE, Child-Pugh score, discharge transportation, and discharge to a facility. No socioeconomic variables were associated with readmission, and of the PROMs, only basic activities of daily living (ADL) impairment was associated. The presence of any cancer was associated with a reduced risk of readmission.

In a multivariable analysis, predictors of readmission included cerebrovascular disease, ascites, admission in the prior 30 days, admission via emergency department, lower admission albumin, higher discharge MELD score, and discharge via public transportation. Basic ADL impairment and impaired CLDQ activity domain was associated with readmission. However, having children in the home, one of the socioeconomic variables, was associated with a reduced readmission risk, as were cancer and admission for infection.

The authors evaluated the prediction of the models using the continuous net reclassification index (NRI) and integrated discrimination improvement (IDI). Using NRI, there were significant improvements when including PROMs, but there were no significant differences between models 3, 4, and 5. Using IDI, none of the models did better than any of the others.

Overall, adding socioeconomic information and PROMs had “marginal effects on prediction,” according to the authors. They did wonder if other PROMs, such as health literacy, self-management, and caregiver burden, could add value and improve prediction.

The findings did suggest that interventions related to poor social support could help with readmissions, the authors noted.

“In addition to targeting the population of high utilizers with more severe liver disease, prevention efforts must also address the negative effects of poor social support and disability,” they concluded. “These goals may be achieved through complex, multicomponent interventions, focused on enhancing self-care and linking patients to community resources.”

Reference

Orman ES, Ghabril MS, Desai AP, et al. Patient-reported outcome measures modestly enhance prediction of readmission in patients with cirrhosis. Clin Gastroenterol Hepatol. Published online July 23, 2021. doi:10.1016/j.cgh.2021.07.032