The Relationship Between Timing of Surgical Complications and Hospital Readmission

Importance  Readmissions after surgery are costly and may reflect quality of care in the index hospitalization.

Objectives  To determine the timing of postoperative complications with respect to hospital discharge and the frequency of readmission stratified by predischarge and postdischarge occurrence of complications.

Design, Setting, and Participants  This is a retrospective cohort study of national Veterans Affairs Surgical Quality Improvement Program preoperative risk and outcome data on the Surgical Care Improvement Project cohort for operations performed from January 2005 to August 2009, including colorectal, arthroplasty, vascular, and gynecologic procedures. The association between timing of complication with respect to index hospitalization and 30-day readmission was modeled using generalized estimating equations.

Main Outcome and Measure  All-cause readmission within 30 days of the index surgical hospitalization discharge.

Results  Our study of 59 273 surgical procedures performed at 112 Department of Veterans Affairs (VA) hospitals found an overall complication rate of 22.6% (predischarge complications, 71.9%; postdischarge complications, 28.1%). The proportion of postdischarge complications varied significantly, from 8.7% for respiratory complications to 55.7% for surgical site infection (P < .001). The overall 30-day readmission rate was 11.9%, of which only 56.0% of readmissions were associated with a currently assessed complication. Readmission was predicted by patient comorbid conditions, procedure factors, and the occurrence of postoperative complications. Multivariable generalized estimating equation models of readmission adjusting for patient and procedure characteristics, hospital, and index length of stay found that the occurrence of postdischarge complications had the highest odds of readmission (odds ratio, 7.4-20.8) compared with predischarge complications (odds ratio, 0.9-1.48).

Conclusions and Relevance  More than one-quarter of assessed complications are diagnosed after hospital discharge and strongly predict readmission. Hospital discharge is an insufficient end point for quality assessment. Although readmission is associated with complications, almost half of readmissions are not associated with a complication currently assessed by the Veterans Affairs Surgical Quality Improvement Program.

The 2000 Institute of Medicine report increased emphasis and scrutiny of the safety and quality of hospital care.¹ For surgical care, postoperative complications lead to increased hospital length of stay, morbidity, and mortality.^2-5 Reporting of some postoperative complications is a new mandatory quality initiative for hospitals. However, most hospitals track only events that occur during the hospitalization; thus, postdischarge complications will probably not be included in the hospital reporting. Accurately measuring complications, whether they occur before or after hospital discharge, is necessary to inform efforts to reduce them.

Quiz Ref IDThe Centers for Medicare & Medicaid Services (CMS) has targeted hospital readmission as a quality metric. As a result of the Patient Protection and Affordable Care Act,⁶ beginning in fiscal year 2013, CMS reimbursement to hospitals will be decreased for higher-than-expected risk-adjusted 30-day readmission rates after admissions for myocardial infarction, pneumonia, and congestive heart failure. The diagnoses covered will expand by fiscal year 2015, and CMS requires hospitals to collect and report all-cause 30-day readmissions. It is conceivable that reporting of postoperative in-hospital complications and 30-day all-cause readmissions will together enable a comprehensive assessment of surgical quality of care. However, there is currently scant evidence available to assess the contribution of postoperative complications to hospital readmission.

To make accurate quality assessments and better inform reimbursement decisions, we need more data to understand the timing and consequences of postoperative complications, as well as which readmissions are unplanned and related to surgical complications. With these goals in mind, the objective of this study is to determine when surgical complications occur in relation to hospital discharge. Our secondary aim is to determine the contribution of predischarge and postdischarge complications to readmission within 30 days of hospital discharge.

This is a retrospective cohort study of national Veterans Affairs Surgical Quality Improvement Program (VASQIP) preoperative risk data for the Department of Veterans Affairs (VA) Surgical Care Improvement Project (SCIP) cohort and hospital discharge date records from the Office of Informatics and Analytics External Peer Review Program from January 2005 to August 2009. We chose the SCIP cohort for this study because it includes common major surgical procedures and has been the focus of surgical quality improvement. The VASQIP collects demographic data, preoperative risk and laboratory data, operative data, and 30-day postoperative morbidity and mortality outcomes in most patients undergoing major surgery in the VA system. Clinical nurse reviewers, trained in clinical medicine and quality assurance, complete in-depth training on the data collection procedures and detailed definitions of each of the variables. A recent study of the quality of the data at a sample of VA medical centers showed that the data were complete and highly reliable.⁷

This study protocol was reviewed and approved by the local VA Research and Development Committee and the Institutional Review Board of the Birmingham (Alabama) VA Hospital, as well as by the Surgical Quality Data Use Group and the Office of Informatics and Analytics in the VA Central Office, Washington, DC; a waiver of informed consent was obtained. Our data sources have been published elsewhere.^7-9

The population for this analysis included patients from the SCIP cohort classified by procedure type, defined according to current procedural terminology code as gastrointestinal (GI), gynecologic, orthopedic, vascular, or other. A total of 59 273 SCIP procedures were identified with matched VASQIP records from January 2005 to August 2009.

The main independent variable was timing of a VASQIP-identified postoperative complication grouped by complication category (Table 1). The dependent variable was readmission within 30 days of discharge from the index hospital stay. Information on readmission within 30 days was obtained from the Corporate Data Warehouse.

Patient-level covariates thought to predict readmission were considered in analyses, including demographic variables, lifestyle variables (eg, alcohol and tobacco use), and cardiovascular, pulmonary, renal, hepatobiliary, nutritional, and immune comorbid conditions obtained from VASQIP records. Surgery characteristics included case status, duration of the operation, wound class, surgical specialty, American Society of Anesthesiologists (ASA) status, and surgical complexity, recorded using the resource-based relative value unit (RVU) for each current procedural terminology code.

Unadjusted analysis of association between the mentioned characteristics and readmission within 30 days was performed using χ² and Wilcoxon rank sum tests. Multivariable analyses adjusting for patient and operative characteristics known to be independently associated with readmission were performed, with use of generalized estimating equations to adjust for clustering at the facility level. Full models included adjustment for history of congestive heart failure, sex, smoking, alcohol use, work RVUs, diabetes mellitus, renal failure, weight loss, emergent case status, operative duration, wound class,¹⁰ ASA classification, body mass index, length of stay, and the occurrence of a postoperative complication. Postoperative complications were categorized as none, predischarge, or postdischarge, and patients with no complications were used as the referent group. Only the first complication experienced by each patient was analyzed. Model building was based on forward selection to obtain the most parsimonious model. Significance was defined at α = .05. All analyses were completed using SAS software (version 9.2; SAS Institute Inc).¹¹

Quiz Ref IDOur study of 59 273 major surgical procedures performed at 112 VA hospitals found an overall 30-day complication rate of 22.6% and readmission rate of 11.9%. Index hospitalization length of stay and complication and readmission rates by patient and procedure characteristics are shown in Table 2. Older and male patients had a longer median stay and significantly higher rates of complications and readmission. Complication and readmission rates were increased in patients with congestive heart failure, diabetes, chronic obstructive pulmonary disease, chronic kidney disease, and significant preoperative weight loss (all P < .001). Operative characteristics, including increasing ASA class, emergent surgery, longer operative time, increasing case complexity as measured in RVUs, and increasing wound contamination, were associated with postoperative complications and readmission (all P < .001). Postoperative length of stay and complication and readmission rates differed by procedure type.

The timing of postoperative complications in relationship to the index hospitalization is shown in Table 3. Of the 13 408 complications, 9646 (71.9%) occurred before discharge, but these rates varied significantly by complication type, from 44.4% for surgical site infection to 91.3% for respiratory complications (P < .001). Quiz Ref IDRespiratory complications occurred most frequently and were associated with the longest stays but were mostly diagnosed before patients were discharged. More than half of surgical site infections were diagnosed after discharge, as were approximately one-third of urinary tract and venous thromboembolism complications. The occurrence of a predischarge complication was associated with an 18.3% readmission rate (range, 17.4%-19.4%) compared with 58.9% (range, 54.3%-78.4%) for postdischarge complications (P < .001 for all complication categories).

Patients who had undergone GI surgery had the highest overall complication rate (37.4%), followed by those who had undergone vascular (30.5%), orthopedic (13.1%), or gynecologic surgery (12.7%) (Table 2). Those who had undergone gynecologic surgery had the shortest median stay (3 days) but the highest proportion of postdischarge complications (55.2%). The longest stay was in patients who had undergone GI surgery, only 21.7% of whose complications were diagnosed after discharge. Patients who had undergone orthopedic or vascular surgery had similar median lengths of stay (5 days), with postdischarge complication rates of 31.0% and 34.3%, respectively (Table 3). The readmission rate for postdischarge complications varied widely by procedure type, from 34.2% for gynecologic to 69.7% for vascular procedures (Table 3).

Results of the generalized estimating equations modeling readmission are shown in Table 4. Patient comorbid conditions, including congestive heart failure, diabetes, renal failure, chronic obstructive pulmonary disease, preoperative weight loss, and smoking, were associated with increased readmission rates. Quiz Ref IDProcedure-related factors that increased the risk of readmission were emergent status, longer operative time, increasingly contaminated wound class, and higher ASA class. In the adjusted model, vascular surgery was associated with increased odds of readmission (odds ratio [OR], 1.8; 95% CI, 1.6-2.0) and gynecologic surgery with decreased odds of readmission (OR, 0.65; 95% CI, 0.50-0.85). The occurrence of a predischarge complication is either weakly or not associated with readmission, whereas a postdischarge complication regardless of type was strongly associated with readmission, with ORs ranging from 7.4 to 59.3 (Table 4).

Our study examined the timing of postoperative complications in relation to hospital discharge and the contribution of these complications to readmission. We found that one-third of all complications and more than half of all surgical site infections were diagnosed after hospital discharge. Complication rates and the timing of their diagnosis differed by surgical specialty and length of stay for the index hospitalization. The occurrence of a postdischarge complication was most strongly associated with increased odds of readmission. However, of all readmissions, only slightly more than half were associated with a currently assessed VASQIP complication, regardless of timing.

A significant number of all postoperative complications are diagnosed after hospital discharge. Although the VA surgical cohort includes mostly white men, our results are similar to those of another large study examining the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) data. Their cohort included 329 951 patients at 181 hospitals, with an overall complication rate of 11%. This rate is lower than our overall complication rate of 22.6%, because their study population included a broader range of surgical procedures, such as breast surgery, which had a complication rate of 2%. Complications were diagnosed after discharge in 32.9%, which is consistent with our results. Also consistent with our results, some complications occurred more frequently after discharge, including surgical site infection (66.0%), urinary tract infection (39.4%), pulmonary embolus (42.2%), and deep venous thrombosis (34.5%).¹²

We observed that surgical site infections are most frequently diagnosed after discharge, consistent with findings of other studies. Daneman et al¹³ examined 622 683 elective surgical procedures in Ontario, Canada. They found an overall surgical site infection rate of 13.5%, with more than half of the infections diagnosed after discharge. Postdischarge infection was associated with an increased risk of readmission (OR, 6.16; 95% CI, 5.98-6.35). They concluded that “post discharge infections are difficult to predict and are associated with real consequences”.^13(p192) Educating patients on the signs and symptoms of surgical site infections before hospital discharge is vital. If signs or symptoms of a surgical site infection develop at home, appropriate and timely follow-up may prevent some readmissions.

We found that patients who had undergone GI surgery (primarily colectomy) had the highest rate of complications (37.4%) and a readmission rate of 16.2%, consistent with other reports in the literature. Wick et al¹⁴ examined 10 882 patients who underwent colorectal surgery and found a 30-day readmission rate of 11.4%. Risk factors in their study associated with readmission included surgical site infection, stoma creation, discharge to nursing home, initial stay longer than 7 days, or rectal resection. Readmission after colorectal surgery occurs frequently and costs approximately $9000 per readmission.¹⁴ Targeted interventions to reduce readmissions could lead to improved quality of care and cost savings because patients undergoing colorectal surgery have high complication and readmission rates. Studies are needed to identify processes of care in the index hospitalization that are linked to complications and or readmission. We have reported elsewhere that the use of oral antibiotic bowel preparation for colorectal surgery was associated with lower surgical site infection rates, shorter stays, and fewer readmissions.^15,16

A significant difference was observed in the timing of postoperative complications with respect to the index hospitalization by procedure category. These differences are probably due to the differences in length of stay by procedure category. A shorter stay is associated with an increased rate of complications diagnosed after discharge, as seen in patients undergoing gynecologic surgery who had a median stay of 3 days with 55.2% of their complications diagnosed after discharge. Furthermore, there were significant differences in readmission by procedure category. Thus, any quality assessment of hospital postoperative complication and/or 30-day readmission rates will be highly dependent on the surgical case mix. These will be important considerations when deciding whether a hospital’s excessive surgical readmissions are truly excessive or appropriate for the surgical case mix.

Currently assessed complications in quality tracking programs, such as VASQIP and ACS NSQIP, do not account for the reasons for postoperative readmission, and, in fact, we identified many patient and procedure factors associated with readmission after adjusting for postoperative complications. However, postoperative complications, postdischarge complications in particular, strongly predict readmission. This is consistent with findings of a recent single institution study, which found that the most significant risk factor for readmission was the occurrence of a postoperative complication (OR, 4.2;95% CI, 2.89-6.13).¹⁷ We recently reported reasons for readmission in the colectomy cohort stratified by the use of oral antibiotic bowel preparation, using VASQIP data supplemented with International Classification of Diseases, Ninth Revision codes to identify readmission reasons.¹⁶ From these data, we infer that there are many additional procedure-specific complications, such as ileus or dehydration, that may reflect complications after colectomy but are not included in quality tracking programs designed to assess outcomes across a broad range of procedures. Validating International Classification of Diseases, Ninth Revision codes as metrics of surgical quality and readmissions is an enormous undertaking and will be the focus of future work. Furthermore, many readmissions to the hospital are necessary and represent high-quality care.

More research is needed to determine which postoperative complications lead to preventable readmissions as well as the processes of care associated with preventing them. A recent analysis of the ACS NSQIP data procedures associated with the highest 30-day postoperative readmission rate found that complications, regardless of timing, were associated with the probability and the cost of readmission.^4,18 We had similar observations, but when we assessed timing of complication, we found that postdischarge complications were a much stronger driver of readmission than those occurring in the index hospitalization. In addition, in contrast to ACS NSQIP data, which capture readmission within 30 days of surgery, we captured readmissions occurring up to 30 days after discharge, allowing all patients to have equal time exposure for readmission.

Our study has several limitations. We were unable to identify patients readmitted to a non-VA hospital. Our patient population is mostly white men, and findings may not be applicable to other patient demographics or other surgical procedures not included in the analyses. Similarly, the veteran population is older with more comorbid conditions; thus, our findings may not be generalizable to the private sector. However, our numbers are similar to those other large studies that include more women and racial diversity. Furthermore, in contrast to ACS NSQIP data, which capture readmission within 30 days of the surgical procedure, we were able to assess readmission within 30 days of discharge from the index hospitalization. We were unable to determine whether any of these readmissions were planned, and we did not attempt to assess causes of readmission by International Classification of Diseases, Ninth Revision codes. Our assessment of complications was restricted to the currently assessed VASQIP complications; thus, many readmissions may be related to complications not currently assessed in this program.

Quiz Ref IDPostoperative complications occur frequently after hospital discharge and are strongly associated with hospital readmission; however, only half of readmissions are associated with a currently assessed complication. Hospital discharge is an insufficient end point for quality assessment, and readmission is not a sensitive means for detecting many postdischarge complications. Although readmissions can serve as a trigger to identify potential quality issues related to the index admission, more work is needed to understand which readmissions are potentially preventable and which are unavoidable.

Accepted for Publication: June 21, 2013.

Corresponding Author: Mary T. Hawn, MD, MPH, Section of Gastrointestinal Surgery, Department of Surgery, University of Alabama at Birmingham, 1922 Seventh Ave S, Kracke Bldg 428, Birmingham, AL 35294-0016 (mhawn@uabmc.edu).

Published Online: February 12, 2014. doi:10.1001/jamasurg.2013.4064.

Author Contributions: Ms Deierhoi and Dr Richman had full access to the data used for this study and take responsibility for the integrity of the data, as well as the accuracy and completeness of the analyses.

Study concept and design: Morris, Hawn.

Acquisition of data: Deierhoi, Hawn.

Analysis and interpretation of data: All authors.

Drafting of the manuscript: Morris, Deierhoi, Hawn.

Critical revision of the manuscript for important intellectual content: Morris Richman, Altom, Hawn.

Statistical analysis: Richman.

Obtained funding: Richman, Hawn.

Administrative, technical, or material support: Deierhoi, Altom.

Study supervision: Morris, Hawn.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by the VA Health Services Research & Development Service (grant PPO 10-296) and the Agency for Healthcare Research and Quality (grant T32HS013852).

Role of the Sponsor: The VA Surgical Quality Data Use Group acted as scientific advisors and provided critical review of the data use and analysis presented in this article.

Previous Presentation: Presented in part at the 97th Annual Clinical Congress of the American Surgical Congress; October 1, 2012; Chicago, Illinois.