A, All patients. B, Patients having any postoperative complication. C, Patients having multiple postoperative complications.
Adjusted for age, sex, race/ethnicity, comorbidities, income, cancer vs benign disease, ostomy creation, admission type (elective vs urgent or emergent), teaching vs nonteaching hospital, and urban vs rural location. Limit lines indicate the 95% CI.
Balentine CJ, Naik AD, Robinson CN, Petersen NJ, Chen GJ, Berger DH, Anaya DA. Association of High-Volume Hospitals With Greater Likelihood of Discharge to Home Following Colorectal Surgery. JAMA Surg. 2014;149(3):244-251. doi:10.1001/jamasurg.2013.3838
Discharge disposition is a patient-centered quality metric that reflects differences in quality of life and recovery following surgery. The effect of hospital volume on quality of recovery measured by rates of successful discharge to home remains unclear.
To test the hypothesis that patients having colorectal surgery at high-volume hospitals would more likely be discharged to home rather than discharged to skilled rehabilitation facilities to complete recovery.
Design, Setting, and Participants
Longitudinal analysis of 2008 hospital inpatient data to identify patients undergoing colorectal surgery who survived to discharge. The setting was the Nationwide Inpatient Sample, the largest all-payer inpatient care database, containing data from more than 1000 hospitals. Participants were 280 644 patients (≥18 years) who underwent colorectal resections for benign or malignant disease and survived to discharge.
Main Outcomes and Measures
The primary end point was discharge to home (with or without home health care) vs discharge to skilled facilities (skilled nursing, short-term recovery, or rehabilitation hospitals or other institutions). The secondary end point was discharge to home with home health care rather than to a skilled facility for patients with postdischarge care needs. Multiple logistic regression using robust standard errors was used to compute the odds ratios of each outcome based on hospital volume, while adjusting for other important variables.
The odds of discharge to home vs discharge to skilled facilities were significantly greater in high-volume hospitals compared with low-volume hospitals (odds ratio, 2.09; 95% CI, 1.70-2.56), with an absolute increase of 9%. For patients with postdischarge care needs, high-volume hospitals were less likely than low-volume hospitals to use skilled facilities rather than home health care (odds ratio, 0.35; 95% CI, 0.27-0.45), with an absolute difference of 10%.
Conclusions and Relevance
Patients having colorectal surgery at high-volume hospitals are significantly more likely to recover and return home after surgery than individuals having operations at low-volume hospitals. This study is the first step in a process of identifying which features of high-volume hospitals contribute toward desirable outcomes. Efforts to identify the reasons for improved recovery at high-volume hospitals can help lower-volume hospitals adopt beneficial practices.
The benefits of performing complex operations at high-volume hospitals rather than at low-volume hospitals have been established for a range of procedures involving cancer and benign disease.1- 4 Comparisons based on hospital volume demonstrate decreased 30-day morbidity and mortality at high-volume hospitals, but the absolute magnitude of this effect is variable and depends on the procedure studied.5,6 While short-term mortality for pancreatic resection is reduced by more than 12% at high-volume hospitals compared with low-volume hospitals, the difference is much smaller for colectomy, for which the absolute difference in mortality is only 1.5%.1 An area that has received less attention is what the quality of recovery is following surgery and how well patients are able to regain preoperative levels of function following discharge. The quality of surgical recovery is partially reflected by whether patients are able to return home following surgery or whether time in skilled facilities (rehabilitation centers, short-term recovery hospitals, skilled nursing facilities, or other institutions) is necessary to regain functional independence. Because discharge to skilled facilities is associated with postoperative complications, more frequent readmissions, and increased mortality, discharge disposition is a clinically salient marker for a difficult and protracted recovery and diminished quality of life following surgery.7- 11 In addition, discharge disposition is a true patient-centered quality measure that represents the difference between recovering at home with family vs protracted recovery periods in a nursing home or other facility. From a quality improvement perspective, discharge disposition is also a useful metric because it is easy to measure and potentially modifiable.
Quality of postsurgical recovery is particularly relevant following postoperative complications. Evidence has suggested that high-quality hospitals are primarily distinguished by the ability to limit mortality through early detection and appropriate management of complications.12- 15 Although survival following complications is important, it remains a crude measure of surgical quality because a considerable difference exists between regaining functional independence and returning home vs requiring a prolonged stay in a rehabilitation facility or nursing home to complete recovery. Although one might expect high-volume hospitals to have greater resources and expertise for enhancing recovery and detecting and managing complications, the relationship between hospital volume and discharge disposition has not been investigated to date. Consequently, it is unclear whether high-volume hospitals are more likely than low-volume hospitals to help patients achieve full recovery and return home after surgery. In addition, for patients with ongoing postdischarge care needs, no data exist to date on whether high-volume hospitals are more likely to use home health care as opposed to skilled facilities.
We hypothesized that recovery following colorectal surgery would be improved at high-volume hospitals compared with low-volume hospitals and that this would be reflected by more frequent discharge to home or the use of home health services rather than discharge to skilled nursing and rehabilitation facilities. We also hypothesized that differences in the likelihood of discharge to home between high-volume hospitals and low-volume hospitals would be especially pronounced for patients recovering from postoperative complications.
The Nationwide Inpatient Sample (NIS) contains data from more than 1000 hospitals, representing a 20% stratified sample of all US community hospitals, excluding Veterans Affairs hospitals and federal facilities.16 After approval by the Baylor College of Medicine and Veterans Affairs institutional review boards, all patients in the 2008 NIS who were 18 years or older, underwent colorectal resections for benign or malignant disease, and survived to discharge were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedure codes and were included in the study. Used were open colon and rectal resection codes (45.71-45.74, 45.76, 45.79, 45.8, 48.5, 48.62-48.63, and 48.69) and laparoscopic colon and rectal resection codes (17.31-17.36, 17.39, 48.42, and 48.51). The 2008 NIS was selected because this was the first year with distinct ICD-9-CM procedure codes for laparoscopic colorectal surgery, and it was thought that surgical approach would be an important covariate when assessing the relationship between hospital volume and discharge disposition.
The main variable of interest was hospital volume, defined as the number of weighted colorectal resections performed during 2008 in the NIS. Hospital volume was modeled as a continuous variable after taking the natural logarithm to normalize and was categorized into 3 equal groups (low, medium, and high) for ease of presentation. The high-volume category represents hospitals with the top one-third of volume for 2008, while the low-volume category represents hospitals with the lowest one-third of volume.
The primary outcome was discharge to home (with or without home health care) vs discharge to a skilled facility. The secondary outcome was discharge to home with home health care vs discharge to a skilled facility for patients with continuing-care needs following discharge. Patients were considered discharged to home if the uniform disposition indicator was coded as routine. Discharge was categorized as with home health care if the disposition was coded as home health care. Discharge to skilled facilities included skilled nursing homes, short-term recovery hospitals, rehabilitation hospitals, or other facilities.
Several states in the NIS report age as a range rather than a continuous variable, so age was modeled using indicator variables, with age younger than 60 years as the reference category compared with ages 60 to 69, 70 to 79, and 80 years or older. Comorbidities were modeled as described by Elixhauser et al.17 Indication for surgery was classified as benign or malignant, and procedure approach was defined as laparoscopic vs open based on ICD-9-CM diagnosis and procedure codes, respectively. Creation of ileostomy or colostomy was also determined using ICD-9-CM codes. Hospital location was based on Core-Based Statistical Areas from the 2000 US Census (http://www.census.gov/hhes/www/housing/housing_patterns/cbsa.html), with metropolitan areas classified as urban and micropolitan or noncore areas considered rural. Teaching hospital status was determined by the presence of an American Medical Association–approved residency program, a ratio of full-time equivalent interns and residents to beds of 0.25 or higher, or membership in the Council of Teaching Hospitals. Complications were identified using ICD-9-CM codes from previously published studies18- 21 and were assessed for face validity by 2 fellowship-trained surgeons (D.H.B. and D.A.A.) with experience in colorectal surgery. Variables were included in the final model as covariates if they were plausibly identified as likely cofounders before modeling, if they were individually significant at P < .05 in the model, or if their inclusion or exclusion changed the point estimate for the independent variable by 10% or more.22,23
Multiple logistic regression was used to evaluate the association between volume and discharge status, and robust standard errors were used to account for clustering within hospitals. Race/ethnicity is not reported by several states in the NIS, so this field is not missing at random. To avoid selection bias from exclusion of these cases, 2 modeling strategies were pursued. First, a separate category was created for missing values so these cases could be included in the analysis. Second, the analysis was repeated with race/ethnicity omitted. None of the subsequent findings differed between these 2 approaches, so results from the first model are presented. Model discrimination was assessed using the C statistic. Tests for main effects and interactions were considered significant at α < .05, although no tests of interaction were significant, so those data are not presented herein. All analyses were conducted using commercially available software (Complex Samples module of SPSS, version 18; SPSS Inc).
A total of 280 644 adult patients in the 2008 NIS underwent colorectal resections and survived to discharge, with 78.3% treated at high-volume hospitals (Table). Patients at low-volume hospitals were older, with 42.6% 70 years or older compared with 35.5% at high-volume hospitals. By contrast, comorbidity burden was greater at high-volume hospitals: 55.8% at high-volume hospitals had 2 or more comorbidities vs 52.9% at low-volume centers. In addition, patients at high-volume hospitals tended to represent higher-income groups and were more likely to have private insurance. High-volume hospitals were also more likely than low-volume centers to perform surgery for benign disease and to perform rectal surgery. The overall complication rates were similar (39.3% at high-volume hospitals and 39.1% at low-volume hospitals), but the proportion of patients with multiple complications was greater at high-volume centers (16.5%) than at low-volume centers (14.4%).
As summarized in the Table and in Figure 1A, the proportion of patients discharged to home from high-volume hospitals was significantly greater than that from low-volume hospitals (85.6% vs 76.1%, P < .001). A similar difference was seen among patients having any postoperative complication (73.2% vs 64.1%, P < .001) (Figure 1B) or for individuals having multiple complications (58.7% vs 49.5%, P < .001) (Figure 1C). Further examination revealed that these differences were primarily due to variations in the use of home health care vs discharge to skilled facilities: the proportion of patients discharged to home without any further assistance was 65.7% at high-volume hospitals and 65.5% at low-volume hospitals (Figure 2A). For the entire cohort, high-volume hospitals used home health care in 19.9% of cases, while 14.4% of patients were discharged to skilled facilities. By contrast, low-volume centers discharged 10.6% of patients with home health services and 23.8% to skilled facilities. Similarly, among patients having a postoperative complication, high-volume hospitals discharged 45.8% to home without assistance, 27.4% to home health care, and 26.8% to skilled facilities. In the same population, low-volume hospitals discharged 48.8% of patients to home, 15.3% to home with home health care, and 35.9% to skilled facilities (Figure 2B). Similar results were seen for individuals with multiple complications, with high-volume hospitals discharging 28.6% of patients to home without assistance, 30.1% to home with home health care, and 41.3% to skilled facilities compared with 33.1%, 16.4%, and 50.5%, respectively, at low-volume hospitals (Figure 2C).
As shown in Figure 3, discharge to home (with or without home health care) rather than to skilled facilities is more than twice as likely at high-volume hospitals compared with low-volume hospitals (odds ratio [OR], 2.09; 95% CI, 1.70-2.56), even after risk adjustment. Similarly, high-volume hospitals are more likely to discharge patients home after they have a complication (OR, 1.81; 95% CI, 1.42-2.31) or multiple complications (OR, 1.63; 95% CI, 1.12-2.39).
Figure 4 shows risk-adjusted ORs for discharge to skilled facilities rather than the use of home health care for individuals requiring postdischarge care. Compared with low-volume centers, high-volume hospitals are 65% less likely to discharge patients to skilled facilities instead of to home health care (OR, 0.35; 95% CI, 0.27-0.45). For patients having postoperative complications, high-volume centers are 55% less likely to discharge to a skilled facility (OR, 0.45; 95% CI, 0.33-0.62), with similar results following multiple complications (OR, 0.48; 95% CI, 0.30-0.77).
Discharge to home health care or to skilled facilities may be influenced by prior use of those resources so that patients who came from nursing homes, for example, would likely return to nursing homes after surgery. To control for the possibility that more patients at low-volume hospitals might have come from skilled facilities, the analysis was repeated after the exclusion of any patient who was transferred from an outside facility rather than being a primary admission. In this population, the odds of discharge to home or home health care rather than to a skilled facility remained significantly increased at high-volume hospitals relative to low-volume hospitals (OR, 2.18; 95% CI, 1.78-2.68). Similarly, the odds of discharge to a skilled facility rather than discharge to home health care was still decreased at high-volume hospitals vs low-volume hospitals (OR, 0.33; 95% CI, 0.25-0.43).
Because insurance status could potentially affect access to home health care providers and skilled facilities, we repeated the analysis for only those patients listing Medicare as the primary payer. Medicare provides coverage for postacute care, so we thought that this could control for differences resulting from heterogeneity of insurance status. The odds of discharge to home rather than to skilled facilities among Medicare patients treated at high-volume hospitals (OR, 1.95; 95% CI, 1.56-2.43) were similar to the odds for the entire cohort, as were odds of discharge to skilled facilities vs to home health care (OR, 0.37; 95% CI, 0.27-0.49).
We also wanted to control for the possibility that access to home health care providers was more readily available at higher-volume centers and that this could affect rates of discharge to home. To address this problem, the Medicare Home Health Compare database24 was linked to the NIS using hospital zip codes, and the number of providers in the same zip code as the hospital was used to assess provider availability. We found a median of 8 providers in the same zip code for low-volume hospitals and a median of 10 providers in the same zip code for high-volume hospitals. When this was normalized for case volume, medians were 0.27 provider per case for low-volume centers and 0.02 provider per case for high-volume centers. Addition of this variable to the model did not change the relationship between volume and odds of discharge to home.
Comparisons of high-volume hospitals vs low-volume hospitals have focused on short-term morbidity and mortality, with less attention paid to whether patients regain independence following surgery.1,5,6,25 To the best of our knowledge, this study is the first to evaluate the quality of recovery from colorectal surgery by exploring the relationship between hospital volume and discharge disposition. While previous studies1,5 showed minimal morbidity and mortality benefits of having colorectal surgery at high-volume hospitals, we found that patients having colorectal surgery at high-volume centers were more than twice as likely to return home following surgery as patients having colorectal surgery at low-volume hospitals. Equally striking was the observation that, when patients had ongoing care needs after discharge, high-volume hospitals were considerably more likely to use home health services than discharge to skilled facilities. Most important, the benefits of having colorectal surgery at high-volume hospitals were maintained even in patients having postoperative complications.
Our findings have several important policy implications. First, considerable cost is associated with discharge to skilled nursing and rehabilitation facilities: Medicare alone spends more than $42 billion each year on postdischarge care.26 Recent work has also suggested that recovery and survival are improved when patients return home following surgery. Legner et al11 examined 89 405 patients who underwent major abdominal surgery and found that 1-year survival was 4 times worse in patients discharged to skilled facilities rather than to their homes. Although patients in this study who used home health services did not fare as well as those who were discharged to home without assistance, the home health group still had improved survival relative to patients discharged to skilled facilities. A second study9 found similar results for trauma patients discharged to skilled nursing care compared with those discharged to home with or without home health services. Another important finding of the latter study was that overall mortality among trauma patients remained steady despite fewer inpatient deaths and shortened length of stay. In essence, the location of patient death shifted from hospitals to skilled facilities, so overall mortality remained static. This finding suggests that strategies focused on rapid discharge rather than on recovery may artificially decrease in-hospital mortality without actually saving lives. Consequently, discharge disposition shows considerable promise as a surgical and hospital quality metric. It is easily measured, potentially modifiable, and associated with patient-centered outcomes, including recovery and quality of life, in addition to standard outcomes such as survival.
This study has several potential limitations. First, risk adjustment using administrative data are subject to misclassification bias. However, little reason exists to assume that hospital volume and misclassification of comorbidity or complications would correlate in a consistent manner, so this problem is unlikely to exaggerate volume effects. It is also important to note that the NIS data set does not include subsequent readmission rates. It could reasonably be argued that our analysis is incomplete because patients discharged to home might be readmitted shortly thereafter. We considered it unlikely that the frequency of readmissions would be high enough to outweigh the magnitude of the effect observed herein, but future work will need to explore this possibility using data sets that are not limited to single inpatient visits. In addition, social support can have an important role in recovery and decision making regarding discharge placement, but the only proxy available in the NIS was income. It is possible those individuals with more resources and support will seek care at high-volume hospitals and that this will influence discharge disposition. Finally, we attempted to analyze the effect of access to home health providers and skilled facilities by separately analyzing patients insured by Medicare, which covers postdischarge care. It is possible that other factors, such as patient and provider knowledge, influence access to care, and this information is not readily available for analysis. Another limitation is the inability to quantify the length of stay at skilled facilities or the duration of home health care. It will be important in future research to further study the effect we observed by evaluating how long patients require additional assistance after discharge and whether this also varies according to hospital volume. While the findings of this study were presented in terms of high-, medium-, and low-volume hospitals, an important point exists regarding the volume outcome relationship that must be emphasized. For ease of data presentation, volume was presented as 3 discrete categories, but the main analysis was performed by treating volume as a continuous variable. Because the findings from this analysis indicate that higher volume is consistently associated with greater odds of discharge to home, no concrete cutoff exists at which higher volume is not associated with discharge to home. For patients seeking to identify a hospital at which to have their surgery, the best strategy if all other factors are equal is to choose the hospital that performs colorectal surgery most frequently. While this is not as simple as categorizing a center as high volume or low volume, it should still be a useful tool in making informed decisions.
This study offers a unique perspective on the relationship between hospital volume and surgical outcomes that extends beyond standard metrics of morbidity and mortality. However, our findings should not be construed as an argument that all colorectal resections should be performed at high-volume hospitals. Instead, this study is the first step in a process of identifying what features of high-volume hospitals contribute to desirable outcomes. It is unclear which characteristics of high-volume hospitals (eg, increased experience, greater resources for recognizing and treating complications, or different patterns and practices of care) are related to discharge disposition. Given that discharge to home with home health care rather than to a skilled nursing facility accounts for much of the observed difference between high-volume and low-volume hospitals, it is possible that the financial arrangements or infrastructure necessary to support home health services are more viable for high-volume centers. Efforts to identify the reasons (clinical and health services) for improved recovery at high-volume hospitals can assist lower-volume hospitals in adopting beneficial practices. These practice changes can then help the health care system approach a key goal of medicine in providing the best care to all patients, regardless of where they are treated.
Accepted for Publication: May 14, 2013.
Corresponding Author: Daniel A. Anaya, MD, Michael E. DeBakey Veterans Affairs Medical Center, Health Services Research and Development Center of Excellence (Room 152), 2002 Holcombe Blvd, Houston, TX 77030 (email@example.com).
Published Online: January 15, 2014. doi:10.1001/jamasurg.2013.3838.
Author Contributions: Drs Berger and Anaya had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Berger and Anaya contributed equally to this work.
Study concept and design: Balentine, Chen, Berger, Anaya.
Acquisition of data: Balentine, Robinson, Anaya.
Analysis and interpretation of data: Balentine, Naik, Petersen, Chen, Berger, Anaya.
Drafting of the manuscript: Balentine, Chen, Berger, Anaya.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Balentine, Petersen, Chen.
Obtained funding: Anaya.
Administrative, technical, or material support: Naik, Chen, Berger.
Study supervision: Naik, Berger, Anaya.
Conflict of Interest Disclosures: Dr Naik is supported by grant K23AG027144 from the National Institute on Aging and by a Doris Duke Charitable Foundation Clinical Scientist Development Award. Dr Robinson has received support from Baylor College of Medicine Comprehensive Cancer Training Program Grant Cancer Prevention & Research Institute of Texas, Research Project 101499. No other disclosures were reported.
Funding/Support: The authors received financial support and resources for the preparation of the manuscript by grant HF P90-020 from the Michael E. DeBakey Veterans Affairs Medical Center, Health Services Research and Development Houston Center of Excellence. This work was funded in part by an American Society of Clinical Oncology Career Development Award from the Conquer Cancer Foundation (Dr Anaya).
Role of the Sponsor: The funding sources 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.
Disclaimers: Any opinions, findings, and conclusions expressed in this material are those of the authors and do not necessarily reflect those of the American Society of Clinical Oncology or the Conquer Cancer Foundation. The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs or Baylor College of Medicine.