Understanding hospital readmissions after pancreaticoduodenectomy: can we prevent them?: a 10-year contemporary experience with 1,173 patients at the Massachusetts General Hospital.

INTRODUCTION: The morbidity and mortality of pancreaticoduodenectomy (PD) have significantly decreased over the past decades to the point that they are no longer the sole indicators of quality and safety. In recent times, hospital readmission is increasingly used as a quality metric for surgical performance and has direct implications on health-care costs. We sought to delineate the natural history and predictive factors of readmissions after PD.
METHODS: The clinicopathologic and long-term follow-up data of 1,173 consecutive patients who underwent PD between August 2002 and August 2012 at the Massachusetts General Hospital were reviewed. The NSQIP database was linked with our clinical database to supplement perioperative data. Readmissions unrelated to the index admission were omitted.
RESULTS: We identified 173 (15 %) patients who required readmission after PD within the study period. The readmission rate was higher in the second half of the decade when compared to the first half (18.6 vs 12.3 %, p = 0.003), despite a stable 7-day median length of stay. Readmitted patients were analyzed against those without readmissions after PD. The demographics and tumor pathology of both groups did not differ significantly. In the multivariate logistic regression analysis, pancreatic fistula (18.5 vs 11.3 %, OR 1.86, p = 0.004), multivisceral resection at time of PD (3.5 vs 0.6 %, OR 4.02, p = 0.02), length of initial hospital stay >7 days (59.5 vs 42.5 %, OR 1.57, p = 0.01), and ICU admissions (11.6 vs 3.4 %, OR 2.90, p = 0.0005) were independently associated with readmissions. There were no postoperative biochemical variables that were predictive of readmissions. Fifty percent (n = 87) of the readmissions occurred within 7 days from initial operative discharge. The reasons for immediate (≤7 days) and nonimmediate (>7 days) readmissions differed; ileus, delayed gastric emptying, and pneumonia were more common in early readmissions, whereas wound infection, failure to thrive, and intra-abdominal hemorrhage were associated with late readmissions. The incidences of readmissions due to pancreatic fistulas and intra-abdominal abscesses were equally distributed between both time frames. The frequency of readmission after PD is 15 % and has been on the uptrend over the last decade.
CONCLUSION: The complexity of initial resection and pancreatic fistula were independently associated with hospital readmissions after PD. Further efforts should be centered on preventing early readmissions, which constitute half of all readmissions.