Health care insurers provide incentives to hospitals and other health care providers for managing surgical patient care. These incentives include reducing the inpatient hospital length of stay (LOS) and, increasingly, controlling costs in the postacute care period. The latter includes appropriate use of home health care, avoidance of readmission, and limited use of “short-term care facilities” (eg, transfer to a skilled nursing facility).
Many hospitals and health care providers are implementing enhanced recovery type processes such as the Perioperative Surgical Home. Demonstrating the value of these initiatives requires accurate determinations of patient disposition, resources used, and outcomes. Measuring changes in LOS is relatively easy, given the ubiquitous availability of electronic hospital records. In addition, measuring clinical care during readmission can be straightforward when the patient is admitted to the same hospital where the surgery was initially performed. However, measuring the processes and resources used during the postacute phase of care can be more difficult. Skilled nursing facilities and inpatient rehabilitation hospitals in the United States often are not affiliated with the health care facilities doing surgery. Lack of integrated clinical and management information systems can make the care and resources used difficult to measure if the patient is readmitted to a different hospital. Although administrative data may be available from insurers, those data are not sufficient to model costs for individual patients (eg, as needed to provide price transparency), because the data lack sufficient detail on clinical and managerial processes.1
Previously, we examined strategies for net cost reductions with the expanded responsibility and expertise of anesthesiologists in the Perioperative Surgical Home.2 We found a large role for informatics to assist in the modeling and reduction of costs.1,2 Thus, the specific objective of our current article is to understand whether there should be a greater priority to achieve information sharing (eg, about resources used) between the hospitals where surgery occurred and (1) the skilled nursing facilities or inpatient rehabilitation hospitals to which patients are transferred upon discharge (when applicable) or (2) other hospitals where readmissions occur. Obtaining and storing data electronically from these 2 sources are dissimilar; both informatics priorities can be challenging, depending on the country, region, and health system. By knowing the answer to this question, hospitals and anesthesia groups can aim their efforts on securing access to the data needed to thrive even with risk sharing, gain sharing, and other financial incentives to control costs over episodes of care (eg, 30 days after discharge).
Our focus is on multiple surgical procedures of multiple specialties (Tables 1 and 2), not just the US Medicare’s limited focus on the Comprehensive Care for Joint Replacement.3 Both with and without patient risk adjustment, hospitals with greater incidences of patients discharged “not to home” (eg, to a skilled nursing facility) have greater rates of readmission.4 Thus, we hypothesized that there would be a positive association between (a) transfer to a skilled nursing facility or an inpatient rehabilitation hospital and (b) readmission to a hospital different from where the surgery was performed. If such an association exists, and if there were a greater incidence of disposition “not to home” than readmission to a different hospital, there would be importance for hospitals to determine postoperative activities, resources used, costs, and outcomes at the short-term facilities to which surgical patients are discharged. Such information could be obtained by entering into data-sharing agreements with the short-term care facilities.
The 2013 Nationwide Readmissions Database was used, subject to the required data use agreement.5 The database was purchased from the US Healthcare Cost and Utilization Project (HCUP) of the US Agency for Healthcare Research and Quality (AHRQ).6 The database includes 14,325,172 discharges sampled from non-Federal US hospitals (eg, US Veterans Administration and military hospitals are excluded).7 To create the public use file, AHRQ used 21 states’ data, all with unique and verified patient identifiers.8 Each pair of records representing a transfer from one acute care hospital to another was combined by AHRQ into a single record in the database.7 In other words, such occurrences were considered to represent a single index hospitalization. An encrypted identifier was created for each patient among all discharges in 2013 for the same patient.8 For each of the 612,335 discharges that we studied (last row of Table 1), we searched the 14,325,172 discharges in the database (first row of Table 1) for patient matches (ie, readmissions) using the visit linkage field.9
Table 1 shows the elective surgical discharges that we studied. Table 2 shows the resulting selection of diagnosis-related groups (DRGs) studied: common surgical DRGs with US national median LOS of 3 days or greater in 2013 (Table 1). The US national median LOSs in Table 2 were effectively measured without error, given the large (nationwide) sample sizes. Medicare Severity DRGs were studied10 since the DRGs model resource use.
The DRGs included were those with median LOS at least 3 days, because the Nationwide Readmissions Database is limited to patients undergoing inpatient surgery. The surgical DRGs with median LOS of 1 to 2 days were for procedures often performed during outpatient surgery (ie, absent from the study database). For example, DRG 352, “inguinal & femoral hernia procedures” had a median LOS of 2 days, and such procedures are often performed on an outpatient basis. DRGs could not be limited to those with median LOS at least 4 days because several of the largest surgical DRGs studied (eg, DRG 470 [major joint replacement] and DRG 766 [cesarean section]) had median LOS of 3 days (Table 2).
Discharges were included for each DRG only if there were at least 10 hospitals with at least 100 discharges for the DRG (Table 1). The minimum of 10 hospitals was based on privacy restrictions in the data use agreement of the Nationwide Readmissions Database.5 The minimum of 100 discharges was chosen so that the median LOS could be calculated reliably.a
A statistical weight was assigned by AHRQ to each of the 14,325,172 discharges in the 2013 Nationwide Readmissions Database discharges to allow estimation for the 35,580,348 discharges nationwide during this year.7,11 For example, among the 612,335 discharges that we studied (Table 1), there was a discharge with a weight of 1.861, meaning that the discharge represented 1.861 similar discharges nationally. There was another discharge with a weight of 3.099, meaning that the discharge represented 3.099 similar discharges nationally, and so forth. More precisely, the 10th percentile of the discharge weights equaled 1.425 discharges nationally; the 90th percentile equaled 3.314 discharges nationally. The studied 612,335 discharges’ weights summed to 1,429,766 discharges nationally. Note that these discharge weights7,11 have nothing to do with “DRG weights12#x201D; for measuring resource use. The data listed in the upper sections of Tables 3 and 4 (ie, with label of “observations”) are the sampled (N = 612,335) unweighted discharges. The data provided in the lower sections of Tables 3 and 4 are estimates for the United States nationwide.
Statistical Analysis for Each of 3 Questions
Question #1: Among patients having major inpatient surgical procedures on an elective basis (Table 1), what are the incidences of disposition upon discharge to locations other than home (ie, transfer to a skilled nursing facility or inpatient rehabilitation hospital) and what are the rates of readmission to a different hospital within 30 days of discharge? These are 2 separate outcome variables. Because discharge to a skilled nursing facility or an inpatient rehabilitation hospital is sometimes planned in advance of surgery (eg, after lower extremity joint arthroplasty), whereas readmission to a different hospital would generally be unexpected, substantively different incidences (eg, ≥10%) seemed possible. Confidence intervals (CIs) for incidences were calculated using STATA 14.1 (StataCorp LP, College Station, TX). Taylor linearization was used, relying on the 2-stage survey sampling (ie, the strata and weight for each discharge).b,13 For the Nationwide Readmission Database, nearly all relevant discharges were included from each of the participating states and hospitals (ie, not like the US National Inpatient Sample, where sampling occurs among discharges of the same hospital).7
Question #2: Among patients having major inpatient elective surgery of multiple common procedures, is there a positive association between (1) disposition upon discharge being not to home (ie, transfer to an inpatient rehabilitation hospital or skilled nursing facility) and (2) readmission within 30 days and to a different hospital than where the surgery was performed? The logistic regression was performed using STATA. The data used for Question #2 were the 612,335 discharges, each with its corresponding national weight (Table 1). The binary dependent variable was whether the patient had at least 1 readmission within 30 days and that readmission was to a different hospital than where the surgery was performed. The independent variable testing Question #2 was binary: whether there was disposition “not to home” (ie, transfer to skilled nursing facility or inpatient rehabilitation hospital). Because the weights were unequal among discharges for the same hospital, we used fixed effects modeling for pooling data among hospitals. The logistic regression was performed with DRG as a category representing the other fixed effect covariate (ie, independent variable); the incidence of readmission differs among hospitals based on their mixes of procedures and complications during the index admission (Table 2). The results in the forest plot (Figure 1) for each of the 15 individual DRG are presented with Dunn-Šidák correction14 (ie, 99.66% CIs where 99.66% = 100 × [1 – 0.05](1/15)). We performed sensitivity analyses including patient-specific covariates (Table 5 second and subsequent lines). The sensitivity analyses with adjustment for the patient-specific covariates were not primary because we expected that sensitivity analyses would show that the results are sensitive to the patient-specific covariates. The premise for our hypothesis was that, for each DRG, sicker patients would not be discharged to home and would have a greater chance of readmission; that is what we are testing. In addition, the patient-specific covariates are highly correlated with the DRG. For patient comorbidities, we used the van Walraven et al15 empirical combination of the Elixhauser comorbidities, as provided in the Nationwide Readmissions Database.15,16
Question #3: Would before/after studies of readmissions at individual hospitals be confounded based on association between the median LOS for the DRG at the hospital where the surgery was performed and the odds of readmission to a different hospital? The logistic regression of Question #3 used the 23,108 discharges with readmission within 30 days (Table 4), again with each discharge having its corresponding national weight. The binary dependent variable was whether the readmission was to a different hospital than where the surgery was performed. The independent variable testing Question #3 was the median hospital LOS for each patient’s DRG. The results in the forest plot (Figure 2) for each of the 15 individual DRG are Dunn-Šidák corrected. Although we also performed sensitivity analyses including covariates (Table 6, second and subsequent lines), these should be considered secondary, because our Question #3 for the logistic regression involves confounding, and that depends only on there being an association (ie, independent effect versus not independent effect does not change the conclusions).
Before the current work, we knew from previous study with the identical, public data set that there were the 23,108 unweighted readmissions within 30 days.17 In addition, for urological surgery, 20.7% of readmissions were at a different hospital than where the surgery was performed.18 Therefore, we predicted that our smallest sample size (ie, Question #3) would have a numerator for the logistic regression of approximately 4780 (ie, 20.7% of 23,108), sufficiently large to achieve a narrow CI.
Among discharges for common surgical DRGs with nationwide median LOS ≥ 3 days (Table 2), 16.15% had a disposition of “not to home” (ie, transfer to a skilled nursing facility or an inpatient rehabilitation hospital; 95% CI, 15.14%–17.22%), and 30.01% had a disposition of “home health care” (95% CI, 28.00%–32.10%). In contrast, among the 3.78% of discharges that were followed by readmission within 30 days (95% CI, 3.60%–3.99%), 0.88% were to a different hospital than the original hospital where the surgery was performed (0.82%–0.95%).
Patients readmitted to a different hospital than where the surgery was performed were 23.3% of the patients readmitted (23.3% = 0.88%/3.78%; 95% CI, 22.0%–24.6%). Adjusting for DRG, disposition “not to home” was associated with greater odds of readmission and to a different hospital than where the surgery was performed (2.11, 95% CI, 1.96–2.27; Figure 1; P < .0001, answering Question #2). This was caused by 2 associations. First, disposition “not to home” was associated with greater odds of readmission (1.90, 95% CI, 1.82–1.98; P < .0001), as expected19,20 and as shown by the sensitivity analyses to be at least in part an indirect effect of patient acuity (Table 5). Second, among the subset of patients who were readmitted, disposition “not to home” was associated with greater odds that the readmission was to a different hospital than where the surgery was performed (1.20, 95% CI, 1.11–1.31; P < .0001).
Adjusting for DRG (Table 2), there was no association between the hospitals’ median LOS for the DRG and the odds that the readmission would be at a different hospital (P = .82, answering Question #3). The odds ratio per each decrease in the hospital median LOS by 1 day was 1.01 (95% CI, 0.91–1.12). Figure 2 shows the forest plot.c,21,22 There were no significant changes in estimates for any of the 6 sensitivity analyses, including patients’ comorbidities (Table 6).23
Based on our 3 observations from the Nationwide Readmissions Database, departments and hospitals wishing to document the value of their Perioperative Surgical Home initiatives should prioritize obtaining access to accurate data on resource use at postacute care facilities such as skilled nursing facilities. This aligns with payers’ goals of reducing the cost of perioperative care, which goes beyond reducing the cost of hospitalization. In addition, approximately one-quarter of readmissions are to different hospitals than where surgery was originally performed. Provided this is known and incorporated in analyses such as risk-based contracts, obtaining information on clinical processes, workflow, and resource use from those different acute care hospitals is less important than getting the postacute care facility data.
From Question #1, many more patients receive postacute care than are readmitted and to a different hospital than where the surgery was performed. This result is likely to be reliable. One reason was that we included all causes of hospital admission within 30 days of discharge, not just readmissions due to complications of the surgery. The second reason was that we considered postdischarge care only at skilled nursing facilities and inpatient rehabilitation hospitals. Home health care can also be labor intensive, and it is used heterogeneously among hospitals.24–26 Home health care reduces readmission rates after major surgery and is net cost saving.27
Although our current findings are for all US patients, older economic studies using US Medicare data are consistent with our conclusions; there were greater total payments among all patients for postacute care facilities than for readmission.28–30 For each of several types of surgical procedures, hospitals were divided into quintiles based on Medicare’s total 30-day payments for their patients. For spine surgery in 2005 to 2007, Medicare payments for patients at the hospitals in the third quintile were greater for postacute care ($3633) than for readmission ($991).28 If we apply our finding that approximately 24.7% of these readmissions were to a different hospital than where the surgery was performed (Table 4), the readmission value would be approximately $245, an amount that is small relative to the $3633. The difference between the fifth (highest) and first (lowest) quintile among hospitals where surgery was performed (ie, an indication that informatics and change in processes can reduce costs) was greater for postacute care ($3788) than for readmission ($1086).28 For colorectal surgery, the Medicare 30-day median payment in 2004 to 2006 was also greater for postacute care ($4666) than for readmission ($730).29 For hip replacement, the fifth and first quantiles among hospitals were greater for postacute care ($3840 and $9725) than for readmission ($582 and $1052), as were the differences ($5885 vs $470).30 For coronary artery bypass grafting, the fifth and first quantiles were greater for postacute care ($2833 and $5165) than for readmission ($1810 and $2715), as were the differences ($2332 vs $905).30
Our conclusion from the answer to Question #2 should not be interpreted that hospitals should make decisions to know and reduce resource use at postacute care facilities in lieu of concern about readmissions. The hospital could promote selection of specific postacute care facilities based, in part, on the facilities’ quality at achieving low rates of readmission. From an informatics point of view, the incidence of postacute care (eg, skilled nursing facility) would not be the end point used (eg, as is currently done31). Rather, the end point would be the quality of functional recovery, resources used (eg, frequency of labor used or based on application of patient condition), and costs.32 The American Hospital Association described that the way for US hospitals to achieve the largest reduction in readmission rates is through partnership with postacute care facilities.33 From MedPAC’s recommendations to the US Congress, skilled nursing facilities’ quality can be measured by their readmission rates.34 The National Quality Forum endorsed measuring all-cause readmission rates during the first 30 days of home health care, all-cause readmission within 30 days of the preceding hospitalization among patients admitted to skilled nursing facilities, and all-cause unplanned readmission for 30 days after discharge from inpatient rehabilitation hospitals.35
Regarding Question #3, Perioperative Surgical Home and enhanced recovery programs typically are implemented at individual hospitals in an effort to reduce the average LOS of patients with specific DRGs or groups of related DRGs. Associations between hospitals’ average (median) LOS and the incidences of subsequent readmission have previously been characterized by using before/after designs for specific procedures.36–43 What has been unknown are the effects on results from absence of data from a readmission, or even knowledge that a patient was readmitted, because this occurred at a different hospital. Such confounding could be caused, for example, by the patient (or family) preferring not to return to the hospital with the relatively short LOS based on a perception that premature discharge was associated with the need for readmission. If so, each decrease in median hospital LOS would be associated with greater odds of readmission to a different hospital. Statistically speaking, readmissions to different hospitals than where the surgery was performed could result in data that are not missing completely at random. Our results in Figure 2 show that, for before/after studies of individual hospitals implementing a Perioperative Surgical Home/enhanced recovery program, the change in the readmission rate at the hospital where surgery was performed would likely remain valid, provided that control chart analysis is used, because of relatively equal incidences of missing data before and after the change in LOS. Note that our results would not apply to hospitals (rather than insurers3,35) benchmarking readmission rates if only some hospitals make efforts to know about readmissions at different hospitals, because those hospitals’ readmission rates would appear falsely higher than that of hospitals that make no effort to record such data.
We limited consideration to DRGs with median LOS nationwide of at least 3 days. We did so to avoid including procedures most commonly performed on an outpatient basis, but for which some cases would be inpatient. Otherwise, we could not study the potential confounding of heterogeneity among hospitals in median LOS on the odds that readmission would be at a different hospital than where the surgery was performed. An important consequence of the 3-day requirement was that there was substantive use of skilled nursing facilities. Medicare pays for skilled nursing facilities postoperatively only after a LOS of 3 days.44 Results would likely be different for bariatric surgery.45 Among US Medicare patients in 2011 undergoing bariatric surgery, the mean total postacute care cost (including home health care and physician visits) was only $365, with $39 being the mean skilled nursing facility cost.45 The means were so small because Medicare’s reimbursements were 0 for most patients, because most patients did not receive any skilled nursing facility care.
The Nationwide Readmissions Database is limited in not providing identifiable geographic information; this is for privacy considerations. We found unexpectedly that patients being transferred to short-term care facilities had greater odds that, if readmitted, the readmission was to a different hospital than where the surgery was performed (P < .0001). A confounding factor affecting the choice of the readmission hospital, or even perhaps the causal factor, may be the distance from the short-term care facility to the original hospital where the surgery was performed. Although we cannot study this explanation, such considerations do not affect our conclusions related to informatics.
Resource use, clinical processes and services, costs, and outcome data are not available in the Nationwide Readmissions Database for the short-term care facilities. Even if Medicare payment data were available, according to the Medicare Payment Advisory Commission, payments are inaccurate measures of costs for these facilities.34 For rural areas in the United States, accurate cost accounting would require detailed operational data,2 because federally certified skilled nursing and inpatient rehabilitation facilities can be virtualized as “swing beds” (ie, otherwise empty acute care beds dynamically designated for short-term care use).46 From the perspective of hospitals, informatics, and Perioperative Surgical Home, our study of incidence of use was reasonable because hospitals generally have 1 information system and 1 method of coordinating data flow with skilled nursing facilities, home health care organizations, and so on. Thus, informatics decisions often are weighted by numbers of patients (as we studied) because that is what influences the time spent by clinicians (ie, nurse practitioners and physicians) filling out order entry forms for postdischarge care.
Finally, for each DRG, we limited consideration to hospitals with at least 100 discharges for elective surgery over 11 months (Table 1). We do not know how this assumption would affect conclusions for smaller hospitals. We suggest treating our results as applying to hospitals with large amounts of inpatient surgery.
In conclusion, we cataloged incidences of readmissions going to different hospitals than where surgery was performed. For Perioperative Surgical Home interventions, the principal informatics focus for postdischarge care should be the resources used, costs, and outcomes related to postacute care that many patients receive, such as at skilled nursing facilities. Afterward and/or if there is a substantial budget, focus on information from the different acute care hospitals to where patients may be readmitted.
The authors thank Jennifer Espy of the University of Iowa for her assistance with editing of the manuscript and preparation for the Journal.
Name: Franklin Dexter, MD, PhD.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Name: Richard H. Epstein, MD.
Contribution: This author helped conduct the study and write the manuscript.
Name: Eric C. Sun, MD, PhD.
Contribution: This author helped analyze the data and write the manuscript.
Name: David A. Lubarsky, MD, MBA.
Contribution: This author helped write the manuscript.
Name: Elisabeth U. Dexter, MD, FACS.
Contribution: This author helped design the study and write the manuscript.
This manuscript was handled by: Nancy Borkowski, DBA, CPA, FACHE, FHFMA.
1. Dexter F, Epstein RH, Bayman EO, Ledolter J. Estimating surgical case durations and making comparisons among facilities: identifying facilities with lower anesthesia professional fees. Anesth Analg. 2013;116:1103–1115.
2. Dexter F, Wachtel RE. Strategies for net cost reductions with the expanded role and expertise of anesthesiologists in the perioperative surgical home. Anesth Analg. 2014;118:1062–1071.
3. Comprehensive Care for Joint Replacement Model. Available at: http://innovation.cms.gov/initiatives/CJR
. Accessed September 9, 2016.
4. Sacks GD, Lawson EH, Dawes AJ, et al. Variation in hospital use of postacute care after surgery and the association with care quality. Med Care. 2016;54:172–179.
13. HCUP Methods Series, Calculating Nationwide Inpatient Sample (NIS) Variances for Data Years 2011 and Earlier. Available at: http://www.hcup-us.ahrq.gov/reports/methods/2003_02.jsp
. Accessed August 9, 2016.
15. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care. 2009;47:626–633.
16. Sharabiani MT, Aylin P, Bottle A. Systematic review of comorbidity indices for administrative data. Med Care. 2012;50:1109–1118.
17. Dexter F, Epstein RH, Dexter EU, Lubarsky DA, Sun EC. Hospitals with briefer than average lengths of stays for common surgical procedures do not have greater odds of either readmission or use of short-term care facilities. Anaesth Inten Care. 2016;45:210–221.
18. Chappidi MR, Kates M, Stimson CJ, Bivalacqua TJ, Pierorazio PM. Quantifying nonindex hospital readmissions and care fragmentation after major urological oncology surgeries in a nationally representative sample. J Urol. 2017;197:235–240.
19. Damle RN, Cherng NB, Flahive JM, et al. Clinical and financial impact of hospital readmissions after colorectal resection: predictors, outcomes, and costs. Dis Colon Rectum. 2014;57:1421–1429.
20. Skolarus TA, Jacobs BL, Schroeck FR, et al. Understanding hospital readmission intensity after radical cystectomy. J Urol. 2015;193:1500–1506.
21. Lindqvist R, Stenbeck M, Diderichsen F. Does hospital discharge policy influence sick-leave patterns in the case of female breast cancer? Health Policy. 2005;72:65–71.
22. Hendren S, Morris AM, Zhang W, Dimick J. Early discharge and hospital readmission after colectomy for cancer. Dis Colon Rectum. 2011;54:1362–1367.
23. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med. 2009;360:1418–1428.
24. Stowers MD, Lemanu DP, Hill AG. Health economics in enhanced recovery after surgery programs. Can J Anesth. 2015;62:219–230.
25. Crosby G, Culley DJ, Dexter F. Cognitive outcome of surgery: is there no place like home? Anesth Analg. 2014;118:898–900.
26. Cyriac J, Garson L, Schwarzkopf R, et al. Total joint replacement perioperative surgical home program: 2-year follow-up. Anesth Analg. 2016;123:51–62.
27. Nabagiez JP, Shariff MA, Molloy WJ, Demissie S, McGinn JT Jr. Cost analysis of physician assistant home visit program to reduce readmissions after cardiac surgery. Ann Thorac Surg. 2016;102:696–702.
28. Schoenfeld AJ, Harris MB, Liu H, Birkmeyer JD. Variations in Medicare payments for episodes of spine surgery. Spine J. 2014;14:2793–2798.
29. Abdelsattar ZM, Birkmeyer JD, Wong SL. Variation in Medicare payments for colorectal cancer surgery. J Oncol Pract. 2015;11:391–395.
30. Miller DC, Gust C, Dimick JB, Birkmeyer N, Skinner J, Birkmeyer JD. Large variations in Medicare payments for surgery highlight savings potential from bundled payment programs. Health Aff (Millwood). 2011;30:2107–2115.
31. Qiu C, Cannesson M, Morkos A, et al. Practice and outcomes of the perioperative surgical home in a California integrated delivery system. Anesth Analg. 2016;123:597–606.
32. Office of Inspector General, Department of Health and Human Services, The Medicare payment system for skilled nursing facilities needs to be reevaluated, September 2015. Available at: https://oig.hhs.gov/oei/reports/oei-02-13-00610.pdf
. Accessed September 5, 2016.
33. American Hospital Association, Maximizing the value of postacute care, 2010. Available at: http://www.aha.org/research/reports/tw/10nov-tw-postacute.pdf
. Accessed September 4, 2016.
34. Medicare Payment Advisory Commission (Medpac), Report to the Congress: Medicare Payment Policy, March 2015, Chapter 8, Skilled Nursing Facilities, pp. 181–209.
35. National Quality Forum, All-cause admission and readmission measures, April 2015. Available at: http://www.qualityforum.org/Publications/2015/04/All-Cause_Admissions_and_Readmissions_Measures_-_Final_Report.aspx
. Accessed September 4, 2016.
36. Sarin A, Litonius ES, Naidu R, Yost CS, Varma MG, Chen LL. Successful implementation of an enhanced recovery after surgery program shortens length of stay and improves postoperative pain, and bowel and bladder function after colorectal surgery. BMC Anesthesiol. 2016;16:55.
37. Nelson G, Kiyang LN, Chuck A, Thanh NX, Gramlich LM. Cost impact analysis of enhanced recovery after surgery program implementation in Alberta colon cancer patients. Curr Oncol. 2016;23:e221–e227.
38. Tyson MD, Chang SS. Enhanced recovery pathways versus standard care after cystectomy: a meta-analysis of the effect on perioperative outcomes. Eur Urol. 2016;70:995–1003.
39. Song W, Wang K, Zhang RJ, Dai QX, Zou SB. The enhanced recovery after surgery (ERAS) program in liver surgery: a meta-analysis of randomized controlled trials. Springerplus. 2016;5:207.
40. Wrench IJ, Allison A, Galimberti A, Radley S, Wilson MJ. Introduction of enhanced recovery for elective caesarean section enabling next day discharge: a tertiary centre experience. Int J Obstet Anesth. 2015;24:124–130.
41. Miller TE, Thacker JK, White WD, et al.; Enhanced Recovery Study Group. Reduced length of hospital stay in colorectal surgery after implementation of an enhanced recovery protocol. Anesth Analg. 2014;118:1052–1061.
42. Zhuang CL, Ye XZ, Zhang XD, Chen BC, Yu Z. Enhanced recovery after surgery programs versus traditional care for colorectal surgery: a meta-analysis of randomized controlled trials. Dis Colon Rectum. 2013;56:667–678.
43. Adamina M, Kehlet H, Tomlinson GA, Senagore AJ, Delaney CP. Enhanced recovery pathways optimize health outcomes and resource utilization: a meta-analysis of randomized controlled trials in colorectal surgery. Surgery. 2011;149:830–840.
45. Grenda TR, Pradarelli JC, Thumma JR, Dimick JB. Variation in hospital episode costs with bariatric surgery. JAMA Surg. 2015;150:1109–1115.
46. Gale J, Croll ZT, Coburn AF, Gregg W; Why do some critical access hospitals close their skilled nursing facility services while others retain them? Policy Brief #31. Available at: http://www.flexmonitoring.org/publications/pb31/