Postoperative 30-Day Readmission: Time to Focus on What Happens Outside the Hospital

Melanie S. Morris, MD; Laura A. Graham, MPH; Joshua S. Richman, MD, PhD; Robert H. Hollis, MD; Caroline E. Jones, MD; Tyler Wahl, MD; Kamal M. F. Itani, MD; Hillary J. Mull, PhD; Amy K. Rosen, PhD; Laurel Copeland, PhD; Edith Burns, MD; Gordon Telford, MD; Jeffery Whittle, MD, MPH; Mark Wilson, MD; Sara J. Knight, PhD; Mary T. Hawn, MD, MPH


Annals of Surgery. 2016;264(4):621-631. 

In This Article


Our study's innovativeness includes the use of patient-level clinical variables (eg, vital signs, laboratory values, and prior healthcare utilization) in addition to operative characteristics and postoperative course. Despite the granularity of the patient-level and hospital course variables, our best models only explained 10.4% of variation in readmission including information known at the time of discharge. We were able to improve prediction (19.6%) of the variation in readmission with the addition of postdischarge complications. However, clinicians would not have this information at the time of discharge, and postdischarge complications are, themselves, challenging to predict.[7] These results suggest that readmission is difficult to predict, making effective prevention strategies difficult. We propose that efforts to reduce readmissions should include optimizing patients' condition and modifiable risk factors before surgery, enriching postdischarge transitional care, and targeting the most common readmission reasons based on specialty, such as surgical site infection rates and GI-related complications for general surgery patients.

Our results are concordant with 2 recent studies examining readmission that did not have detailed clinical information. Merkow et al[7] examined a National Surgical Quality Improvement Program cohort of 498,875 patients undergoing bariatric surgery, colectomy, hysterectomy, total joint replacement, ventral hernia repair, or lower extremity bypass. Their overall readmission rate was lower than ours at 5.7%; however, they were studying 30-day postprocedure, not postdischarge readmission rates. They noted that only 2.3% of readmissions were for a complication they experienced during their postoperative hospital stay. Their overall model of readmission explained 27% of the variation in readmission risk, which is greater than our predictive ability when we included postdischarge complications. As we found in our study, their most important contributors to readmission risk were patient-level factors, accounting for 24% of the variation. They concluded that readmission after surgery was associated with new postdischarge complications, not exacerbation of predischarge complications. They noted that readmissions are a measure of postdischarge complications and that sometimes readmissions may be appropriate and timely hospital care.[7] Gani et al[8] reported their results of 22,559 patients from 8 different surgical subspecialties performed at a single center from 2009 to 2013 and noted similar importance of patient-level factors. In general, good predictive models have not been demonstrated, and readmission remains difficult to predict. Given that much of the risk in being readmitted is attributable to unidentified or potentially unmodifiable patient factors, the use of readmission as a surrogate for hospital quality of care may not be justified in surgical patients.

Our study and others[7,9] have shown that surgical site infections (SSIs) and wound complications are the most frequent reasons why surgical patients are readmitted. As we have previously shown, one third of all complications and more than half of all SSIs are diagnosed after hospital discharge.[10] Quality improvement efforts should focus on programs to reduce SSI rates through examining and implementing best practices within the hospital, providing patient education on SSI rates, and ensuring appropriate postoperative follow-up care. One example of this is that for patients undergoing elective colectomy, implementing a full mechanical and oral antibiotic bowel preparation was shown to decrease SSI rates by 50%.[11] Implementing SSI reduction bundles has also been shown to effectively decrease SSI rates, although which components are most important is debated.[12] Novel ideas using technology, such as harnessing the electronic medical record patient portal features or emailing pictures of wounds, should be explored. Most readmissions for wound complications occurred within the first week after hospital discharge. Several small studies suggest that early clinic visits (within 5 d of discharge) for certain patient populations would allow for earlier intervention and avoid hospital readmission, although this needs to be further studied.[13–17]

With early recognition and intervention, we may observe increases in readmission. Despite exhaustive work on reducing perioperative complications, it is clear that readmissions will continue to occur as long as we continue to perform surgery. Using readmission as a quality metric may result in unintended consequences. Financial penalties already exist for selected readmissions. The hospital margin on the readmission may be substantially lower than using those resources on a second index admission, as is the case for total knee arthroplasty.[18] Similar economic disincentives exist for surgeons in the current fee-for-service model. The CMS measure does not effectively discriminate among hospitals, given that 54% of all US hospitals were penalized in 2015.[19] Considering that our data demonstrate that sicker and more disadvantaged patients are more likely to be readmitted, hospitals are being penalized for caring for these vulnerable populations. Care must be taken to avoid increasing disparities of care when attempting to improve quality of care with financial penalty.

Using granular clinical data, we found that having a postoperative pain score above 8 was associated with increasing rates of readmission. It is unclear whether higher pain scores reflect underlying evolving complications, medication management issues, or patient variability in adjustment to the post-operative course. These findings highlight the need for alternative pain control methods and pathways such as enhanced recovery after surgery (ERAS). Programs like ERAS focus on non-narcotic pain control and regional anesthetic blocks, improved patient education, and improved care transitions. The relationship between ERAS and hospital readmission needs further investigation. Although ERAS improve many postoperative outcomes, it has had little impact on postoperative ileus rates. GI reasons for readmission, including postoperative ileus and obstruction, accounted for 16.1% of readmissions in our cohort and are poorly understood; it is unclear whether these are preventable readmission reasons.

Preoperative patient factors contributed the most to the variation in readmission risk (R2 7.0%) predictable at the time of discharge. We found that patients with renal dysfunction and cardiac disease had the highest rates of readmission. Patients who were partially or totally dependent in their functional status and admitted to the hospital from a nursing home were more likely to be readmitted. Functional status and frailty are important areas of research to improve postoperative outcomes in our increasingly elderly patient population, as more than 4 million major surgical procedures are performed in the United States annually on patients age 65 and older.[20] A recent review examined surgical mortality risk factors for older patients including cognitive impairment, functional dependence, malnutrition, frailty, and preoperative institutionalization.[21] Of the 28 studies examined, risk ratios for functional dependence relating to mortality ranged from an adjusted hazard ratio of null to an adjusted odds ratio of 18.7. Frailty was associated with a 3- to 13-fold increased risk of discharge to a facility and an increased risk of mortality. Another recent article from the United Kingdom examined frailty in 413 vascular surgery patients and found a greater than 50% readmission rate in frail patients.[22] Prehabilitation programs help modify patient-level factors before surgery in an effort to improve postoperative outcomes. Most published research reports on prehabilitation are small, single-center studies with mixed results, suggesting that this approach needs further refinement. Further research is warranted to understand if a tailored prehabilitation program could decrease readmission rates.

Our study has several limitations. Our cohort represents a national Veterans Affairs patient population with a large sample comprised predominantly of white men; thus, our results may not be generalizable to women or nonfederally insured populations. The VA Health System is more fully integrated than other healthcare settings including the electronic medical record, pharmacy, and primary care. This integration may allow the VA to more fully coordinate care in the perioperative period which may influence readmission rates, thus limiting the ability to generalize these findings to other populations. Further prospective research into the mechanisms underlying readmission are underway to explore the structural characteristics of facilities in that may influence readmission following surgery. Reasons for readmissions were determined by physician review of primary diagnosis codes and may be subject to misclassification. However, the frequencies of readmission reasons were similar to those previously published using clinical chart abstraction.[7] Although we have a great breadth of clinical patient-level data demographics, comorbidities, vital signs, laboratory values, and postoperative complications, there remains the possibility that there are important predictors unavailable to us, especially subjective or patient-reported measures. In addition, we had significant missing data on laboratory and to a lesser extent on vital sign data, which limited our ability to fully understand the importance of these measures in readmission, although in this case "missing" lab values represent cases where tests were not ordered which is itself prognostic information.