Descriptive Analysis of 30-Day Readmission After Inpatient Surgery Discharge in the Veterans Health Administration

Importance  For the first time to our knowledge, this study analyzes and reports the 30-day all-cause readmission rates for surgical procedures performed in the Veterans Health Administration (VHA).

Objective  To analyze and report 30-day all-cause readmission rates following discharge from 9 surgical specialties in the VHA for a 10-year period.

Design, Setting, and Participants  In a retrospective observational study, Veterans Affairs Surgical Quality Improvement Program data for surgery records and VHA administrative discharge data were linked to evaluate all-cause readmission within 30 days of discharge from the surgical inpatient stay. The study population represents 9 surgical specialty groups: general, urology, neurosurgery, orthopedic, otolaryngology, plastic, thoracic, peripheral vascular, and cardiac. Trends of postoperative hospital admission rates for each surgery were investigated using 10 years (fiscal years 2001-2010; N = 894 943) of linked data.

Main Outcomes and Measures  All-cause 30-day readmission after surgery for each specialty.

Results  During the 10-year period, the overall 30-day all-cause readmission rate following inpatient surgery discharge significantly decreased from 12.9% to 12.2% (P < .001). Unadjusted readmission rates varied by surgical specialty: general, 12.5%; urology, 9.0%; neurosurgery, 10.5%; orthopedic, 9.6%; otolaryngology, 9.5%; plastic, 12.2%; thoracic, 14.4%; peripheral vascular, 16.0%; and cardiac, 16.6%. The following specialties were found to have a significant decline in readmission rates: orthopedic (P = .004), otolaryngology (P = .005), plastic (P = .02), thoracic (P = .04), peripheral vascular (P < .001), and cardiac (P = .003). Postoperative hospital length of stay in individual specialties decreased during this period (each P < .05) except for thoracic and cardiac surgery, which remained unchanged. Readmission diagnoses varied by specialty; postoperative infection was the most common readmission diagnosis in 7 specialties and the second most common in the other 2 specialties (urology and thoracic). Urinary tract infection and digestive system complications were also common readmission diagnoses.

Conclusions and Relevance  This retrospective observational study showed decreasing 30-day readmission rates associated with a decline in postoperative hospital length of stay for 9 surgical specialties in the VHA during a 10-year period. Further study will be required to capture data from patients who had surgery at a VHA facility but were readmitted in the private sector.

Patients, caregivers, clinicians, and payers are concerned about rates of all-cause 30-day readmission following an inpatient surgical procedure.¹ The 2 primary concerns are the quality of medical care and the associated cost.^2-4 The Centers for Medicare and Medicaid Services (CMS) report a national average all-cause 30-day readmission rate of approximately 18%, which has remained unchanged for several years despite the significant efforts of many hospitals and health care systems to lower this rate.^5-7 The Medicare Payment Advisory Commission reported that readmission within 30 days of discharge cost Medicare $15 billion in 2005, with $12 billion related to potentially preventable readmissions.^1,8 In 2009, the CMS began to publicly report 30-day risk-standardized all-cause readmission measures for patients with acute myocardial infarction, heart failure, and pneumonia.^1,9 As a result, hospital readmission became a metric for health care system performance. The CMS publicly announced in the fiscal year 2014 Inpatient Prospective Payment Systems that their Readmission Reduction Program will include readmissions following total knee and total hip arthroplasty in fiscal year 2015.¹⁰

There are numerous published results about hospital readmission, but many of the studies have focused on the specific conditions or populations to replicate CMS methods. To our knowledge, 30-day all-cause hospital readmission rates following surgical procedures in the Veterans Health Administration (VHA) have not been published to date.

Quiz Ref IDSeveral readmission studies^11-16 have investigated whether decreases in the initial hospital length of stay (LOS) correlate with increased rates of subsequent readmission in VHA and non-VHA populations, and the results have been mixed. Kaboli et al¹¹ studied VHA acute medical admissions, noting that both the length of admission and 30-day readmission rate decreased from 1997 to 2010. In the Medicare population with heart failure, however, a decrease in LOS from 1993 to 2006 was associated with a contemporaneous increase in 30-day readmission.¹² A study of colorectal resection readmission in Ireland found that longer hospital LOS for colorectal resection was associated with increased risk of emergency readmission.¹³ Two studies of surgical patient populations at the Cleveland Clinic did not detect a correlation between hospital LOS and early readmission.^14,15 An analysis of readmission following coronary artery bypass graft surgery in the Medicare population from 2007 to 2008 found an association between reduced postoperative LOS and increased risk of 30-day readmission.¹⁶

This 10-year retrospective review of VHA surgical outcomes data describes and summarizes trends in 30-day all-cause readmission rates and postoperative hospital LOS (POHLOS) for inpatient surgical procedures. In addition, we describe variation between surgical specialties and evaluate readmission diagnoses as a basis for future strategies to potentially reduce preventable hospital readmission.

This is a retrospective cohort study using 2 large maintained data sources, Veterans Affairs Surgical Quality Improvement Program (VASQIP) data and VHA discharge data. The VHA has monitored patients who had cardiac surgery beginning in 1987, and VASQIP monitors patients who had surgery in 8 noncardiac specialties since 1991: general, urology, neurosurgery, orthopedic, otolaryngology, plastic, thoracic (noncardiac), and peripheral vascular.^17-19 Eligibility for VASQIP is determined by Current Procedural Terminology codes with measurable risk of 30-day mortality and morbidity attributable to the surgical procedure. The VASQIP data include clinical variables for the preoperative to 30-day postoperative period, including 30-day mortality and POHLOS.

Quiz Ref IDPatients who underwent VASQIP-assessed inpatient surgery performed by general, urologic, neurosurgery, orthopedic, otolaryngology, plastic, thoracic, peripheral vascular, or cardiac surgeons at VHA facilities from October 1, 2000, through September 30, 2010 (fiscal years 2001-2010) were identified. The specialty code of the surgeon as captured in VASQIP defined the specialties. A surgical case is included in this study as an inpatient surgery if the patient was hospitalized prior to or admitted within 1 day of surgery. Outpatient surgical procedures assessed by VASQIP, patients with blinded administrative identifiers, and patients with a known 30-day mortality outcome following surgery were excluded.

The VHA discharge data were obtained from the VA Austin Information Technology Center. These data include all inpatient admissions of veteran patients to VHA facilities. These VHA administrative data provide both surgery and nonsurgery admission dates and identifiers, but clinical information is limited to diagnosis and procedure codes. Per VHA policy, the information presented in this article represents a VHA operations activity, does not constitute research, and therefore does not require informed consent or institutional review board review.

The primary outcome of interest was all-cause 30-day readmission from the date of hospital discharge for patients undergoing a VASQIP-assessed inpatient surgical procedure. Readmission was evaluated for each surgical case rather than at the patient level. Therefore, multiple readmissions for any given patient were counted. This study investigated all-cause readmission owing to the diversity of surgical procedures included. The POHLOS was defined as the number of days from the surgery date to the hospital discharge date as recorded in VASQIP.

Trends of postoperative hospital admission rates for each surgical specialty and mean POHLOS for each surgery were investigated using 10 years of linked data. Cochran-Armitage 2-tailed trend tests were performed to detect changes in 30-day readmission rates for each specialty and Mann-Kendall trend tests were performed to detect change in mean POHLOS for each specialty during the 10-year period. P < .05 was considered statistically significant.

Common readmission diagnoses were summarized using International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes in 3 ways: (1) rankings of readmission diagnoses for each specialty; (2) overall total proportion for all 9 specialties combined; and (3) grouping of similar ICD-9 codes using the Clinical Classification Software (CCS) for ICD-9,Clinical Modification. The CCS is a tool for clustering patient diagnoses and procedures into a manageable number of clinically meaningful categories developed at the Agency for Healthcare Research and Quality.²⁰

Readmission diagnoses were investigated using the most recent 3 years of data (fiscal years 2008-2010) rather than the full 10-year data set to focus on a patient population with characteristics more likely to be similar to the current patient population.

All statistical analyses were performed using SAS version 9.3 statistical software (SAS Institute, Inc).

There were 749 362 VASQIP-assessed inpatient surgical case records for the 10-year period (fiscal years 2001-2010). The number of VHA administrative admission records (surgical and nonsurgical) for the same 10-year period was 6 420 457. Joining these 2 data sets resulted in 894 943 linked readmission records. After restriction to the 9 major surgical specialties, exclusion of 30-day mortalities, and exclusion of duplicate records, there were 666 579 major surgery cases, of which 80 798 had a readmission in the 30 days following discharge from inpatient surgery. There were 1787 unique ICD-9 diagnosis codes among the 24 684 readmission cases between fiscal years 2008 and 2010. Within this period, 1974 administrative records did not have a principal admission diagnosis code assigned.

Quiz Ref IDCochran-Armitage 2-tailed trend tests were performed to detect trends in the overall 30-day all-cause readmission rate during the 10-year study period. The overall readmission rate showed a statistically significant decreasing trend from 12.9% to 12.2% (P < .001; slope = −0.00883). Unadjusted readmission rates varied by specialty (Figure 1): general, 12.5%; urology, 9.0%; neurosurgery, 10.5%; orthopedic, 9.6%; otolaryngology, 9.5%; plastic, 12.2%; thoracic, 14.4%; peripheral vascular, 16.0%; and cardiac, 16.6%. Significant but slight decreasing trends during the 10-year period were found for orthopedic (P = .004), otolaryngology (P = .005), plastic (P = .02), thoracic (P = .04), peripheral vascular (P < .001), and cardiac (P = .003) surgery. Readmission rates were unchanged for general (P = .94), neurosurgery (P = .82), and urology (P = .26) specialties. Quiz Ref IDThe overall mean POHLOS for the initial surgical inpatient stay during the 10-year period significantly decreased from 10.6 to 9.2 days (P < .001). The mean POHLOS varied by surgical specialty (Figure 2): general, 10.9 days; urology, 5.9 days; neurosurgery, 8.9 days; orthopedic, 8.1 days; otolaryngology, 9.2 days; plastic, 17.4 days; thoracic, 10.8 days; peripheral vascular, 9.5 days; and cardiac, 11.3 days. The mean POHLOS for individual specialties significantly decreased during this period (each P < .05) except thoracic and cardiac, which remained unchanged. In cardiac surgery, the trend of mean POHLOS had a positive slope (0.047) but was not significant (P = .23). Plastic surgery showed the steepest declining slope (−1.06), from 20.7 days to 15.2 days, but still showed the longest mean POHLOS (17.4 days) during the 10-year period.

Table 1 displays the top 5 readmission diagnosis codes for each specialty. The top 10 were also evaluated (data not shown). The ICD-9 code 998.59, other postoperative infection, was the most common readmission diagnosis for all specialties except urology and thoracic, ranging in prevalence among the specialties from 4.5% to 12.3%. Urinary tract infection (UTI) (ICD-9 code 599.0) was the most common readmission diagnosis for urology (7.4%), and this diagnosis was also common for other specialties. Pneumonia (ICD-9 code 486) ranked within the top 10 diagnoses for thoracic, cardiac, orthopedic, general, and neurosurgery specialties (range, 1.5%-5.7%). Digestive system complications (DSCs) (ICD-9 code 997.4) and dehydration (ICD-9 code 276.51) were the second and fourth most common readmission diagnoses, respectively, following general specialty inpatient admissions but were not common for other specialties. Other nervous system complications (ICD-9 code 997.09) and other acute postoperative pain (ICD-9 code 338.18) were the second and third most common readmission diagnoses, respectively, for neurosurgery but were not common diagnoses for other specialties. Many common readmission diagnoses were related to infections across the 9 specialties: other postoperative infection (ICD-9 code 998.59), infection and inflammatory reaction due to internal joint prosthesis (ICD-9 code 996.66), UTI (ICD-9 code 599.0), infected postoperative seroma (ICD-9 code 998.51), infection (chronic) of amputation stump (ICD-9 code 997.62), and infection and inflammatory reaction due to other vascular device, implant, and graft (ICD-9 code 996.62).

Table 2 displays the top 10 readmission ICD-9 diagnosis codes in the study cohort. Overall, postoperative infection (ICD-9 code 998.59) was the most common readmission diagnosis (9.8%) and UTI (ICD-9 code 599.0) was the second most common readmission diagnosis (2.5%). Other reasons in the top 10 included pneumonia (2.0%), DSCs (2.0%), congestive heart failure (CHF) (1.9%), acute renal failure (1.8%), disruption of external operation wound (1.7%), hematoma complicating a procedure (1.6%), infection (chronic) of amputation stump (1.3%), and dehydration (1.2%).

Because the readmission diagnoses are highly varied, an attempt was made to group the codes using the CCS for ICD-9, Clinical Modification. We performed single-level CCS groupings to investigate readmission patterns. Table 3 summarizes the CCS groupings for the top 50% of the ICD-9 readmission diagnosis codes. The single-level CCS clustered the 1787 ICD-9 codes into 209 groups. The CCS category of complications of surgical procedures or medical care (category 238) included 24.0% of the readmission reasons, 5459 readmission cases, and 60 ICD-9 codes. The second most common CCS category was complication of device, implant, or graft (category 237), which included 5.9% of the readmission reasons, 1337 cases, and 52 ICD-9 codes. The third most common CCS category was UTI (category 159), which included 2.8% of the readmission reasons, 645 cases, and 9 ICD-9 codes.

The CCS grouping result is partially consistent with Table 1 and Table 2 because other postoperative infection was included in the top CCS category, complications of surgical procedures or medical care, and UTI and CHF were also within the top 5 common readmission diagnosis groups. However, the CCS category of complication of device, implant, or graft (5.9%), the second most common group, was not apparent when individual ICD-9 codes were used.

It should be noted that the single-level CCS combined widely varying diagnoses. For example, the category of complications of surgical procedures or medical care (category 238) combined other postoperative infection (ICD-9 code 998.59), DSCs (ICD-9 code 997.4), disruption of external operation wound (ICD-9 code 998.32), hematoma complicating a procedure (ICD-9 code 998.12), and hemorrhage complicating a procedure (ICD-9 code 998.11) into a single category. After performing the CCS categorization, the readmission diagnoses were still highly varied and did not uncover additional patterns of readmission for this study.

Thoracic surgery had 2 cancer-related readmission diagnosis codes, malignant neoplasm of bronchus and lung (ICD-9 code 162.3; 6.4%) and unspecified malignant neoplasm of bronchus and lung (ICD-9 code 162.9; 3.2%). Other malignant neoplasm readmission diagnoses were commonly found for otolaryngology and urology. Otolaryngology had common diagnoses of malignant neoplasm of thyroid gland (ICD-9 code 193; 5.0%) and unspecified malignant neoplasm of larynx (ICD-9 code 161.9; 2.6%), while urology had a common diagnosis of malignant neoplasm of bladder (ICD-9 code 188.9; 3.7%).

Diagnoses of DSCs (ICD-9 code 997.4), CHF (ICD-9 code 428.0), and dehydration (ICD-9 code 276.51) were the fourth, fifth, and tenth common readmission diagnoses (2.0%, 1.9%, and 1.2%, respectively) by overall percentage but were not common across all 9 specialties. Among these diagnoses, 77.5% of DSCs and 60.7% of dehydration occurred following general surgery and 62.0% of CHF occurred following cardiac and peripheral vascular surgery inpatient admissions. Among all readmissions, 26.9% occurred following general surgery, 11.1% following cardiac surgery, and 17.1% following peripheral vascular surgery. The high percentages of these codes in the study cohort were driven by the readmission volume for specific specialties.

Using 2 large national databases, we examined trends of all-cause 30-day readmission rates and mean POHLOS following major surgery in 9 specialties at VHA facilities between October 1, 2000, and September 30, 2010. We found that the overall readmission rate significantly declined from 12.9% to 12.2% during the 10-year period concurrent with a significant decrease in mean POHLOS from 10.6 to 9.2 days.

When postdischarge readmissions did occur, our results identified postoperative infection, UTI, and pneumonia as the most common diagnoses. This finding is consistent with other studies that also indicated postoperative infection as the most common reason for readmission.^9,21,22

Kassin et al⁹ studied risk factors for 30-day readmission following general surgery in the VA and found that surgical infection and gastrointestinal complications were the most common reasons for readmission. Schairer et al²¹ reported that infection was the most common readmission reason after total knee arthroplasty, and Hannan et al²² reported similar results that postoperative infections and heart failure were common readmission reasons after coronary artery bypass graft surgery. Vorhies et al,²³ however, reported that the most common readmission reasons following total hip arthroplasty in the Medicare population were cardiac complications. Our study suggests that a future strategy to minimize postoperative infection will positively affect postoperative readmission rates across all specialties.

The mean POHLOS significantly declined during the 10-year period for 7 surgical specialties (all specialties except cardiac and thoracic, which remained unchanged). During the same period, the readmission rates significantly improved in 6 specialties (all specialties except general, urology, and neurosurgery, which remained unchanged). The declining trends of POHLOS with a contemporaneous decline in readmission in the VHA are consistent with the medical acute inpatient population within the VHA for a similar period. Kaboli et al¹¹ observed LOS and readmission from 1997 to 2010 in 129 acute care VHA hospitals and found that LOS improved and the 30-day readmission rate did not increase during the 14 years. There are similar published results from surgical populations outside the VHA. Vorhies et al²³ observed association between LOS and readmission rates among a sample of Medicare patients undergoing total hip arthroplasty from 2002 to 2007 and found that a reduction in LOS was not associated with an increased readmission rate.

By contrast, some studies in the population with heart failure have found hospital LOS decreases and increasing readmission rates.^12,16,24 Bueno et al¹² investigated trends of LOS and readmission rate among Medicare patients hospitalized for heart failure from 1993 to 2006 and found that the mean LOS decreased from 8.81 to 6.33 days but the 30-day readmission rate increased from 17.2% to 20.1% during the 13-year period. Li et al¹⁶ also published similar results, with instrumental variable analysis showing a reduction in POHLOS associated with an increased risk for 30-day readmission among Medicare patients undergoing bypass surgery. Carey and Lin²⁴ investigated acute care hospitals in California in 2008 and estimated probability models for patients with heart attack and heart failure. Their models predicted that for every 1-day increase in LOS, there was a corresponding reduction in the risk of readmission (7%-18% reduction for patients with heart attack and 1%-8% reduction for patients with heart failure).

Evaluation of some patterns of readmission noted in this analysis would require additional detailed data on the patient’s care plan and evaluation of procedure-specific outcomes. The cancer-related readmission diagnosis codes observed in thoracic, otolaryngology, and urology surgery may be representative of planned readmission. The data available to us could not discern this level of detail. In addition, the reason for prolonged POHLOS in plastic surgery was not identified.

This study has several limitations. Patients with a 30-day mortality outcome following surgery were excluded (n = 20 937), so no relationship of readmission with mortality can be inferred. This study did not attempt to quantify private-sector admissions for patients who had surgery in VHA facilities. These exclusions may also bias the readmission rates experienced by VHA patients downward. The veteran population has characteristics that differ from the general population, as they are more likely to be male and have higher comorbidities; thus, this research may not be generalizable to the nonveteran population.^25,26 This study was designed to be a retrospective analysis of readmission rates and POHLOS by specialty grouping. We were unable to determine cause of readmission from available data. We have not investigated differences in surgical case complexity or patient characteristics such as age or comorbid conditions that contribute to the risk of readmission within the scope of this study. The data used for this analysis also predate the VHA’s National Utilization Management Integration review processes, which became a VHA-wide requirement in the third quarter of fiscal year 2010 with the release of VHA directive 1117.²⁷ National Utilization Management Integration is designed to evaluate the appropriateness of acute admissions and continued stay days and to optimize the LOS in acute care. We were not able to link our records to the results of these reviews or infer any changes in readmission rates or POHLOS following the national rollout of National Utilization Management Integration.

This study summarized 30-day all-cause readmission rates, mean POHLOS, and principal readmission diagnoses for each of 9 surgical specialties in the VHA between fiscal years 2001 and 2010. We found that the overall readmission rate significantly declined from 12.9% to 12.2% during the 10-year period associated with a significant decrease in mean POHLOS from 10.6 to 9.2 days. Quiz Ref IDThe readmission rate ranged from 9.0% (urology) to 16.6% (cardiac) and the mean POHLOS ranged from 5.9 days (urology) to 17.4 days (plastic). The readmission diagnoses were diverse, but postoperative infections, UTI, and pneumonia were common for all specialties. Further study is required to identify whether patients who had surgery in VHA facilities return to a VHA facility or private-sector facility, and validated readmission risk prediction models should be developed for each surgical specialty.

Corresponding Author: Soonhee Han, MS, Veterans Affairs National Surgery Office, 4100 E Mississippi Ave, Ste 310, Denver, CO 80246 (soonhee.han@va.gov).

Accepted for Publication: April 8, 2014.

Published Online: September 17, 2014. doi:10.1001/jamasurg.2014.1706.

Author Contributions: Ms Han had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: Han, Smith.

Drafting of the manuscript: Han, Gunnar.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Han, Smith.

Administrative, technical, or material support: Gunnar.

Study supervision: Smith, Gunnar.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the Veterans Affairs National Surgery Office.

Role of the Funder/Sponsor: The Veterans Affairs National Surgery Office had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: This article represents the opinions of the authors and not necessarily those of the Department of Veterans Affairs or the US government.

Previous Presentation: This paper was presented at the 2014 Annual Meeting of the Association of VA Surgeons; April 7, 2014; New Haven, Connecticut.