AT THE ASA ANNUAL MEETING
CHICAGO (FRONTLINE MEDICAL NEWS) – Preadmission and postdischarge factors were important predictors of postoperative readmission in a large cohort of surgical patients, but the hospital course had little incremental impact on either readmissions or postdischarge complications in the cohort, according to a retrospective study of Veterans Affairs data.
The findings suggest that efforts to reduce postoperative readmissions should focus on enhanced postdischarge surveillance and early intervention, Dr. Melanie S. Morris of the University of Alabama at Birmingham reported at the annual meeting of the American Surgical Association.
To assess the relative contributions of patient factors, operative characteristics, and postoperative hospital course on readmissions, she and her colleagues evaluated 243,956 general, vascular, and orthopedic surgery patients in 121 VA hospitals. The overall readmission rate among the cohort was 11.1%, and for general, vascular, and orthopedic surgeries, the rates were 12.9%, 15.4%, and 7.6%, respectively; the average postoperative length of stay was 6.9 days, and 6.1% of patients experienced a predischarge complication.
Almost all readmissions occurred within 2 weeks of discharge, and for general surgery patients, most occurred within 1 week. The readmission rate for vascular surgery patients remained high beyond the 2-week mark.
An examination of the reasons for readmission showed that wound complications were the most common reason for readmission, and this was particularly true for vascular surgery patients, in whom 44% of readmissions were for wound complications, Dr. Morris said.
Gastrointestinal complications including ileus and obstruction were also common, accounting for nearly 28% of readmissions among general surgery patients, she said.
Importantly, when including preoperative data (such as demographics, comorbidities, social and behavioral factors, labs and vital signs, and planned procedure type), the variability in readmissions could only be explained 8.6% of the time, she said.
“Adding in operative data, such as procedure complexity and intraoperative blood transfusions, as well as postoperative course, added very little to our predictive ability. Including both of those groups, we could only explain 10% of the variation in readmission,” she said.
Including postdischarge data such as complications and emergency department utilization in the model increased predictive ability to 18%.
R2 and C-statistics comparing the sequentially built model showed that demographics and comorbidities contributed the most to predicting readmission risk, Dr. Morris said.
Modeling based on readmission reason and specialty improved predictive ability. For example, almost 12% of readmissions for wound complications among vascular surgery patients were predictable.
“Our best predictive ability was for orthopedic patients who were readmitted with pneumonia. We were able to predict that 14% of the time,” she said.
The findings were derived by merging VA Surgical Quality Improvement Program data from inpatient operations performed between 2007 and 2014 and involving at least a 2-day postoperative hospital stay, with clinical data including laboratory findings, vitals, prior health care utilization, and postoperative complications.
“We then grouped our variables of interest into the following categories: preoperative, operative, postoperative but predischarge, and postdischarge,” she explained, noting that logistic models predicting 30-day readmission were constructed by sequentially adding groups into the model. Models were compared by way of adjusted R2 and C-statistics.
Assuming postoperative readmissions are preventable suggests that they are linked to the quality of care during the index hospitalization. The current findings demonstrate the challenges in predicting readmissions, and are important given that hospitals with higher-than-expected readmission rates for certain diagnoses and procedures are fined by the Centers for Medicare & Medicaid Services; 54% of hospitals were fined in 2015, she said.
“Readmission is difficult to predict at the time of discharge despite exhaustive statistical modeling with very granular clinical patient-level detail. Preoperative patient factors and postdischarge complications contribute the most to predictive models. Efforts to decrease readmissions should focus on modifiable patient-level factors, transitions of care, and minimizing postoperative complications,” she concluded.
Dr. Morris reported having no disclosures.
The complete manuscript of this presentation is anticipated to be published in Annals of Surgery pending editorial review.