Actual versus estimated length of stay after colorectal surgery: which factors influence a deviation?

The American Journal of Surgery (2014) 208, 663-669

Clinical Science

Actual versus estimated length of stay after colorectal surgery: which factors influence a deviation? Usama Ahmed Ali, M.D., Tony Dunne, M.D., Brooke Gurland, M.D., Jon D. Vogel, M.D., Ravi P. Kiran, M.D.* Department of Colorectal Surgery, Cleveland Clinic, Mail Code A30, 9500 Euclid Avenue, Cleveland, OH 44195, USA

KEYWORDS: Diagnosis-related group; Length of hospital stay; Colorectal surgery

Abstract BACKGROUND: The aim of this study was to determine factors associated with deviation in length of hospital stay (LOS) from that determined by diagnosis-related groups. METHODS: A cohort study from a prospectively collected database was conducted, including consecutive patients undergoing surgery in a high-volume colorectal surgery department in 2009. RESULTS: For 1,461 included patients, average expected and actual LOS were 8.17 days (interquartile range, 4.7 to 11.9 days) and 8.31 days (interquartile range, 4 to 10 days), respectively. The most prominent factors associated with an increase of LOS from expected were parenteral nutrition (5.11 days), emergency room admittance (3.67 days), and ileus (3.45 days) (P % .001 for all). Other independently associated factors included blood transfusion, anastomotic leak, sepsis, pulmonary embolism, and surgeon. Patients with higher severity illness indexes and longer postoperative intensive care stay had lower than expected LOS. CONCLUSIONS: After colorectal surgery, several modifiable factors are associated with deviation of LOS from expected. An opportunity hence exists to reduce both LOS and financial burden for hospitals in an era of pay for performance. Ó 2014 Elsevier Inc. All rights reserved.

Colorectal surgical procedures carry an increased risk for certain complications, such as postoperative ileus and anastomotic leakage. These complications are known risk factors for prolonged hospital stay.1–4 Studies have looked at the influence of these factors on the absolute length of hospital stay5–7 (LOS), but correlating these differences with deviation from what is expected by risk estimation models, such as diagnosis-related groups (DRGs), has yet to be done. DRGs are a classification system intended to The authors declare no conflicts of interest. * Corresponding author. Tel.: 11-216-444-6930; fax: 11-216-445-8627. E-mail address: [email protected] Manuscript received December 22, 2012; revised manuscript June 3, 2013 0002-9610/$ - see front matter Ó 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjsurg.2013.06.004

relate the type of patients a hospital treats to the costs incurred by the hospital.8 In the United States, DRGs are used to determine hospital payments for Medicare patients.9 Other similar systems exist around the world.10–12 DRGs offer an estimation of the expected LOS on the basis of preoperative patient characteristics, type of surgery, and postoperative complications. Trying to predict deviation from this estimate has certain advantages. Such analysis offers insights to physicians and departments as to which factors result in the greatest increases in cost. This allows prioritization of research and management goals.13–16 As an example, a previous study from our group showed that mean direct hospital costs were significantly lower for laparoscopic compared with open colectomy, although DRG classifications (and thus payments to hospitals) were similar.17

664 Additionally, the DRG classification offers a reference value that is more reliable than absolute LOS alone. Because the DRG classification corrects for the complexity of patient population (ie, case mix), results can be better interpreted across different sources and are thus more generalizable. Therefore, the aim of this study was to evaluate factors associated with deviation of actual LOS from expected LOS on the basis of the DRG classification.

Methods Study population After study approval was obtained from the institutional review board, all consecutive patients operated on from January 12 to December 17, 2009, in the Department of Colorectal Surgery at the Cleveland Clinic were included. Patients were selected on the basis of procedural codes for colorectal surgical procedures. Patients had to be R18 years of age. No exclusion criteria were used.

Study variables Clinical and administrative data were obtained from the prospectively collected clinical ‘‘outcomes database’’ and from departmental financial records, respectively. Clinical data collected included patient characteristics, medical histories, comorbidities, smoking, laboratory profiles, operative details, and postoperative outcome (actual LOS and 30 daypostoperative complications). Patients were accrued into the database on the day of discharge. A predetermined data set was collected, including diagnosis, laboratory tests, surgical details, and in-hospital complications. Anastomotic leak was defined as any clinically evident leak or extravasation of contrast from the anastomosis detected by radiologic studies. Prolonged postoperative ileus was defined as absence of gastrointestinal function for .5 days or the need for placement of a nasogastric tube. Other complications were also based on specific definitions. All patients were contacted after discharge to obtain complete 30-day follow-up. Administrative data included DRG classification, expected LOS, severity of illness index (from 1 to 4 indicating, respectively, minor, moderate, major, and extreme severity of illness),8,18 and source of admittance (ie, clinic referral, emergency room, or transfer).

Data and statistical analysis Difference between actual and expected LOS was calculated for all patients. Variables associated on univariate analysis with deviation of LOS were included in a multivariate model. Wilcoxon’s rank-sum test, the KruskalWallis test, and Spearman’s correlation test were used for univariate analysis as appropriate. Linear regression analysis with forward stepwise inclusion was used for the multivariate model. A P value of .05 was used as the cutoff point for statistical significance.

The American Journal of Surgery, Vol 208, No 4, October 2014

Results A total of 1,461 patients (mean age, 50 6 18 years; 755 [52%] men) were included (Table 1). Stoma takedown was the indication for surgery in 361 patients (25%), cancer in 257 (18%), Crohn’s disease in 204 (14%), and ulcerative colitis in 200 (14%). Patients were coded to 76 different surgical procedures. Most frequently performed operations were enterostomy closure, laparoscopic and open partial colectomy, exploratory laparotomy, and laparoscopic total colectomy. A total of 1,196 patients (82%) were assigned to 5 distinct DRG codes (Table 1). The average expected LOS on the basis of the DRG classification was 8.17 days (interquartile range [IQR], 4.7 to 11.9 days), while

Table 1

Patient characteristics

Characteristic

Value

Age (y) Men BMI (kg/m2) Expected LOS (d) Actual LOS (d) Actual LOS 2 expected LOS (d) Diagnosis/indication for surgery Stoma removal Cancer Crohn’s disease Ulcerative colitis Diverticulitis Polyp Other ASA classification 1 (no disturbance) 2 (mild disturbance) 3 (moderate disturbance) 4 (severe disturbance) Surgical procedure Closure enterostomy Laparoscopic partial colectomy Partial colectomy Exploratory laparotomy Laparoscopic total colectomy Adominoperineal resection Ileal pouch anal anastomosis Other DRG codes Major bowel procedure with CC (330) Major bowel procedure without CC (331) Major bowel procedure with MCC (329) Minor bowel procedure without CC (346) Minor bowel procedure with CC (345) Rectal resection with CC (333) Other

50 6 18 755 (52%) 26.3 6 6.2 8.17 6 4.97 8.31 6 6.58 .14 6 4.69 361 257 204 200 93 58 288

(25%) (18%) (14%) (14%) (6%) (4%) (20%)

31 828 541 58

(2%) (57%) (37%) (4%)

339 253 152 88 79 66 59 425

(23%) (17%) (10%) (6%) (5%) (5%) (4%) (29%)

476 245 238 131 106 42 223

(33%) (17%) (16%) (9%) (7%) (3%) (15%)

Data are expressed as mean 6 SD or as number (percentage). ASA 5 American Society of Anesthesiologists; BMI 5 body mass index; CC 5 complications and comorbidities; LOS 5 length of hospital stay; MCC 5 major complications and comorbidities.

U. Ahmed Ali et al.

Estimation of length of hospital stay

665

Table 2 Univariate analysis of preoperative and intraoperative factors potentially associated with differences between actual and expected length of hospital stay Factor General characteristics Sex (male) Age at surgery BMI Diagnosis† ASA classification Preoperative characteristics Diabetes COPD Dialysis Smoking (current smokers) Creatinine levels Albumin levels White blood cell count Hematocrit Platelet count History of chemotherapy History of pelvic radiation Surgical details Surgical procedure Surgical approach‡ Operative time Type of intestinal anastomosisx Creation of stoma Blood transfusions (intraoperative) Intraoperative complications Any complication Small bowel puncture or laceration

N

n (%)*

P

1,461 1,455 1,450 1,458 1,460

755 (52%) d d d d

.54 .74 .61 ,.001 .59

1,460 1,460 1,460 1,460 1,458 1,037 1,458 1,457 1,457 1,460 1,460

120 24 9 280 d d d d d 61 43

1,460 1,460 1,461 1,442 1,459 1,457

d d d d 597 (41%) 163 (11%)

.01 ,.001 .08 .09 .07 .74

1,461 1,460

197 (13%) 158 (11%)

.56 .05

(8%) (2%) (1%) (19%)

(4%) (3%)

.10 .35 .77 .046 .54 .96 .01 .76 .41 .22 .007

ASA 5 American Society of Anesthesiologists; BMI 5 body mass index; COPD 5 chronic obstructive pulmonary disease. *Number of patients with a certain characteristic (for dichotomous factors only). † Diagnosis of cancer was associated with an increase in length of hospital stay (LOS) from expected, while diverticulitis and ulcerative colitis were associated with a decrease in LOS from expected. ‡ Surgical approach was classified as open, laparoscopic, hand assisted, or converted. Open surgery and conversion were associated with increase LOS from expected, and laparoscopic and hand-assisted surgery were associated with a decrease from expected. x Type of intestinal anastomosis was classified as stapled anastomosis, hand-sewn anastomosis, other techniques, or no anastomosis.

the average actual LOS was 8.31 days (IQR, 4 to 10 days). The average actual LOS was higher than expected by .14 days (IQR, 22.2 to 1.8 days) per patient. Actual hospital stay was shorter than or equal to expected in 804 patients (55%) and longer than expected in the remaining 657 (45%). The proportions of patients who stayed R5, R7, or R10 days longer than expected were 136 (9.3%), 91 (6.2%), and 41 (2.8%), respectively.

Risk factors for deviation of hospital stay Preoperative and postoperative factors that were associated on univariate analysis with deviation of LOS from expected are shown in Tables 2 and 3, respectively. Diagnosis, type of surgical approach (laparoscopic, open, or hand assisted), postoperative parenteral nutrition, multiple postoperative complications, DRG classification, and severity of illness were all strongly associated with deviation from

expected LOS (P ,.001). Additionally, other factors showed sufficient association to be included in multivariate analysis. On multivariate analysis, diagnosis, surgeon, postoperative recovery unit, use of parenteral nutrition, several postoperative complications, DRG classification, source of admittance, and severity of illness were independently associated with deviation of LOS from expected (Table 4). Postoperative factors were generally more strongly associated with a deviation of LOS than preoperative factors. Need for parenteral nutrition, prolonged ileus, elevated creatinine (serum creatinine .1 mg/dL), need for blood transfusions, and anastomotic leaks were each associated with an increase of LOS of .2 days from expected, compared with not having these complications (Table 4). Additionally, sepsis, venous thromboembolism, and wound infections were independently associated with increases of LOS from expected. Patients admitted from the emergency room and from other hospitals had LOS that were longer than expected by

666

The American Journal of Surgery, Vol 208, No 4, October 2014

Table 3 Univariate analysis of administrative and postoperative factors potentially associated with difference between actual and expected length of hospital stay Factor Administrative characteristics DRG code Admitting source† Severity of illness index‡ Postoperative recovery Initial postoperative facilityx Need for parenteral nutrition Postoperative complications Urinary tract infection Pulmonary embolism/DVT Prolonged ileus Blood transfusion New arrhythmias Deep SSI (organ space) Wound infection Anastomotic leak Elevated creatinine (.1 mg/dL) Sepsis

N

n (%)*

P

1,461 1,461 1,461

d d d

,.001 .001 ,.001

1,459 1,460

d 181 (12%)

.01 ,.001

1,459 1,460 1,460 1,460 1,460 1,460 1,459 1,461 1,460 1,460

42 43 223 268 38 103 139 43 53 113

.002 .001 ,.001 .003 .85 ,.001 .001 ,.001 .002 ,.001

(3%) (3%) (15%) (18%) (3%) (7%) (10%) (3%) (4%) (8%)

DRG 5 diagnosis related group; DVT 5 deep venous thrombosis; SSI 5 surgical site infection. *Number of patients with a certain characteristic (for dichotomous factors only). † Admitting source scored as outpatient clinic referral, emergency room, transfers from other hospital, transfers from other health care facility, or unknown. ‡ Severity of illness index classified into 4 categories: minor, moderate, major, and extreme severity. x Initial postoperative facility classified as regular ward, postanesthesia care unit (overnight stay), and intensive care unit.

3.67 days (standard error, .53 days) and 1.32 days (standard error, .48 days), respectively, compared with patients admitted through clinic referral. Of the 58 patients admitted through the emergency room, only 3 (5%) had previous surgery in our department. Different surgeons were also independently associated with a deviation of LOS from expected of up to 61.25 days from the mean deviation of LOS observed for all surgeons. Patients with increasing severity of illness had shorter than expected LOS. Actual LOS for moderate, major and extreme severity was 21.06 days (standard error, .25 days), 24.57 days (standard error, .34 days), and 27.56 days (standard error, .48 days) shorter than expected, respectively, compared with minor severity. Patients admitted to the intensive care unit (ICU) as the initial postoperative facility also stayed shorter than expected by 21.25 days (standard error, 0.60 days) compared with those discharged to the regular nursing floor. To evaluate whether the use of laparoscopic surgery was responsible for LOS differences among surgeons, data were also specifically examined for patients who underwent laparoscopic surgery (n 5 315). For these patients, the influence of surgeon was not statistically significant, on both univariate and multivariate analysis. However, when data were further scrutinized, there was a large variation between different surgeons. Most of this variation was between surgeons who regularly performed laparoscopic surgery and those who did not. Of a total of 12 surgeons, 5 had performed ,10 laparoscopic surgeries each, while 7 performed .20 laparoscopic surgeries during the study

period. When the 7 surgeons with .20 laparoscopic surgeons were analyzed (n 5 292 patients), a variation of only about 60.5 LOS days was observed.

Comments In this study, we examined the difference between actual and expected LOS on the basis of the DRG classification in a large colorectal surgical department. Overall, the DRG classification accurately indicated the actual LOS, with an average difference of .14 days (IQR, 22.2 to 1.8 days). We did identify several important factors that were associated with significant deviation of LOS from expected. In this regard, postoperative complications were found to be more influential than preoperative or intraoperative risk factors. Thorough examination of the identified factors is necessary to evaluate potential ways of reducing their burden on patients’ health and hospital costs. The need for parenteral nutrition and admittance from the emergency room resulted in the largest increases from expected, with average covariate-adjusted increases of 5.1 and 3.5 days, respectively. Parenteral nutrition has previously been associated with postoperative morbidity because of hyperglycemia and infectious complications.19–21 A consequent reduction in parenteral nutrition use in some disciplines demonstrated decreased complications and hospital resource utilization.22 At our institution, parenteral nutrition is administered only when necessitated by the patient’s

U. Ahmed Ali et al. Table 4

Estimation of length of hospital stay

667

Independent factors associated with differences between actual and expected length of hospital stay

Factor General characteristics Diagnosis Ulcerative colitis Surgical details Surgeon‡ Postoperative recovery Initial postoperative facility Regular nursing ward ICU PACU overnight admission Use of parenteral nutrition Postoperative complications DVT/PE Prolonged ileus Blood transfusion (postoperatively) Wound infection Elevated creatinine (.1 mg/dL) Anastomotic leak Sepsis Administrative characteristics DRG Rectal resection with CC (code 333) Admitting Source Outpatient clinic Emergency room Other hospital Other health care facility Severity Minor Moderate Major Extreme

N/n (%)*

200 (14%) 1,461

Estimate (standard error)†

P

2.77 (.30)

.010

Up to 61.25

.017 .04

1,386 59 14 181

(95%) (4%) (1%) (12%)

43 223 268 139 53 43 113

(3%) (15%) (18%) (10%) (4%) (3%) (8%)

42 (3%)

Reference 21.25 (.60) 1.17 (1.04) 5.11 (.37) 1.90 3.45 2.26 .79 2.49 2.15 1.97

.04 .27 ,.001

(.62) (.30) (.31) (.37) (.58) (.65) (.44)

.002 ,.001 ,.001 .03 ,.001 .001 ,.01

1.38 (.60)

.02 ,.001

1,305 58 86 10

(89%) (4%) (6%) (1%)

Reference 3.67 (.53) 1.32 (.48) 21.10 (1.22)

417 593 282 169

(29%) (41%) (19%) (12%)

Reference 21.06 (.25) 24.57 (.34) 27.56 (.48)

,.001 .006 .36 ,.001 ,.001 ,.001 ,.001

CC 5 complications and comorbidities; DRG 5 diagnosis-related group; DVT 5 deep venous thrombosis; ICU 5 intensive care unit; PACU 5 postanesthesia care unit; PE 5 pulmonary embolism; SSI 5 surgical site infection. *For continuous outcomes, the number of patients valid for analysis. For dichotomous factors, number and percentage of patients with a certain characteristic. † The estimate indicates the covariate-adjusted deviation in days of length of stay (LOS) from expected for the respective factor compared with the deviation in the group not having the factor (or reference group in case of categorical variables). ‡ Different surgeons were responsible for a deviation of up to 61.25 days from the mean deviation of LOS observed in this cohort.

condition. In the postoperative phase, our policy is to initiate parenteral nutrition in patients with delayed recovery of gastrointestinal function or to continue this until the reestablishment of an adequate and consistent oral intake that can maintain nutrition can be ascertained when patients on this preoperatively undergo surgery. It is therefore likely a surrogate of protracted recovery. Consequently, directly influencing this factor is probably difficult. A potential limitation of this study is that we were unable to separate out if the timing of total parenteral nutrition, whether preoperative or postoperative, influenced its associated with outcomes in terms of LOS. Although it may be expected that certain diagnoses, such as inflammatory bowel disease, might result in increased use of preoperative parenteral nutrition, in a post hoc analysis, no associations were seen between these diagnoses and longer than expected LOS. Furthermore, postoperative parenteral nutrition remained

an important factor on multivariate analysis even when accounting for both diagnosis and ileus. Thus, these results indirectly suggest that preoperative parenteral nutrition may have had a limited influence, although this may also have been due to the relatively small number of patients receiving the therapy preoperatively in our study. In a similar way, emergency room admittance is not an easy target for care providers to reduce LOS. In tertiary centers serving a large urban area, emergency admittance is known to contribute to increased LOS and expenses.23,24 The increased severity of patients presenting in the emergency room and time needed to perform tests that are routinely performed outside the hospital are factors that may lead to increased LOS.24 Prolonged ileus was the postoperative complication associated with the largest increase in LOS, averaging .3 days from expected. Because of the nature of colorectal

668 surgery, prolonged ileus is a fairly common complication.1,5 It also is particularly difficult to prevent. We previously showed that 2 of the 3 independent predictors of ileus were unmodifiable (ie, age and previous abdominal surgery).1 The third factor (preoperative narcotic use) is a potential target for LOS reduction. A recent double-blind randomized trial showed that a short course of systemic lidocaine postoperatively was successful in improving gastrointestinal motility and shortening of the LOS compared with placebo.25 Another double-blind randomized trial showed that intravenous ketorolac was able to reduce the rate of postoperative ileus in patients undergoing laparoscopic colon surgery.16 The study also showed ‘‘promise in reducing hospital stay,’’ although the outcome was not statistically significant, probably because of a relatively small sample size (n 5 44).16 Other complications, such as elevated creatinine, need for blood transfusion, anastomotic leak, sepsis, and deep venous thrombosis, were also associated with a substantial increases in LOS. These postoperative complications are known for their deleterious influence on clinical outcome, and several studies have investigated factors contributing to their occurrence.26–32 Our study highlights the importance of addressing these specific factors from a cost perspective. Measures aimed at reducing these complications has the largest potential impact for hospitals reimbursed through a DRG-based system. Therefore, departments and hospitals can prioritize and select their goals in a more informed manner. Another potential factor that emerged was the role of the individual surgeon. Individual surgeons had an important influence on deviation of LOS of up to 61.25 days compared with the mean of all surgeons. This suggests that modifiable factors likely exist and further supports a strategy of identifying and targeting potential factors that influence outcomes and LOS on the basis of best evidence. Enhanced recovery (fast-track) programs may offer a way of reducing differences in LOS between care providers. Such programs help decrease variations in care by different physicians and have shown to reduce LOS.33–37 Although our patients are managed using standardized care pathways, individual variations exist amongst surgeons for preoperative and postoperative care and decisions in addition to any potential intraoperative differences. Although the inclusion of a large group of patients undergoing different operations by several surgeons serves the current aim of the study well, these very factors limit our ability to study the specific role of surgeon on outcome. In particular, such factors as experience, certain variations in preoperative and postoperative care decisions among surgeons, and other specific factors such as surgical technique were not recorded. Although the use of laparoscopic surgery seemed to minimize the influence of surgeon on LOS, differences among surgeons persisted, with surgeons who regularly performed laparoscopic surgery tending to have less variation in LOS than those who performed laparoscopy less commonly.

The American Journal of Surgery, Vol 208, No 4, October 2014 Future studies that prospectively collect such differences in preoperative, intraoperative, and postoperative factors among surgeons and their influence on LOS may help elucidate the reasons for these differences in LOS. Patients with severe illnesses and those initially admitted to the ICU after surgery (ie, patients identified as potentially at high risk for complications) stayed at our hospital for shorter periods than expected. The DRG classification allocates a larger margin of fluctuation in the expected LOS for severe patients, which may in part be responsible for these results. In a post hoc analysis, we observed a strong relation between the presence of preoperative comorbidities and direct postoperative admittance to the ICU. First, we found a highly significant association between American Society of Anesthesiologists classification and rate of ICU admittance, with 0%, 1.2%, 5.0%, and 36% of patients being directly admitted to the ICU for American Society of Anesthesiologists classes 1, 2, 3, and 4, respectively (P , .001). We also found a significant association between direct postoperative ICU admittance and preoperative workup for comorbidities, including having had a cardiac consult (P , .001) or stress test (P , .001). These results suggest that admittance directly to the ICU was indeed highly associated with the presence of known preoperative comorbid states and thus supports the explanation that DRG classification currently may overestimate the LOS in patients with such comorbidities. These results also suggests that our department performed better than expected in this complex category of patients. This is in line with previous evidence showing that tertiary high-volume centers generally perform better than lower volume centers.5,38 In addition to identifying factors that are important for LOS and costs, this study also provides a potential way to compare results from various institutions more reliably. Rhodes et al7 showed that patient mix is probably the most important factor in determining LOS. Schoetz et al39 warned against accepting rigid criteria for LOS for patients undergoing colorectal surgery, because differences in clinical characteristics can require differences in approach. Although the DRG classification cannot explain individual patient variation of LOS and thus is not suitable for detailed subgroup analysis,7 it does provide a more reliable way for comparing the overall LOS between different hospital. For example, in a study by Faiz et al5 comparing LOS after colorectal surgery in various British National Health Service hospitals, it was observed that wide variation existed in the rates of prolonged stay at all levels of hospital volume. Correlating such findings to a case mix–adjusted expectation can provide insight as to whether these variations were mainly driven by differences in the patients case mix or more by different practices between hospitals. A potential limitation of this study is the retrospective nature of the analysis. Thus, some potentially relevant factors identified in previous studies, such as delays in time of presentation and socioeconomic status of patients, were not available for this analysis.5,6 Despite these drawbacks,

U. Ahmed Ali et al.

Estimation of length of hospital stay

669

this study measures the relative effects of comorbidities and complications on LOS, providing another dimension to the importance of reduction of these events on health costs not previously provided by estimates of absolute LOS alone.

17. Senagore AJ, Brannigan A, Kiran RP, et al. Diagnosis-related group assignment in laparoscopic and open colectomy: financial implications for payer and provider. Dis Colon Rectum 2005;48:1016–20. 18. Horn SD, Sharkey PD, Chambers AF, et al. Severity of illness within DRGs: impact on prospective payment. Am J Public Health 1985;75: 1195–9. 19. Pasquel FJ, Smiley D, Spiegelman R, et al. Hyperglycemia is associated with increased hospital complications and mortality during parenteral nutrition. Hosp Pract (Minneap) 2011;39:81–8. 20. Sena MJ, Utter GH, Cuschieri J, et al. Early supplemental parenteral nutrition is associated with increased infectious complications in critically ill trauma patients. J Am Coll Surg 2008;207:459–67. 21. Ishizuka M, Nagata H, Takagi K, et al. Total parenteral nutrition is a major risk factor for central venous catheter-related bloodstream infection in colorectal cancer patients receiving postoperative chemotherapy. Eur Surg Res 2008;41:341–5. 22. Rhee P, Hadjizacharia P, Trankiem C, et al. What happened to total parenteral nutrition? The disappearance of its use in a trauma intensive care unit. J Trauma 2007;63:1215–22. 23. Grannemann TW, Brown RS, Pauly MV. Estimating hospital costs. A multiple-output analysis. J Health Econ 1986;5:107–27. 24. Thorpe KE. Why are urban hospital costs so high? The relative importance of patient source of admission, teaching, competition, and case mix. Health Serv Res 1988;22:821–36. 25. Herroeder S, Pecher S, Schonherr ME, et al. Systemic lidocaine shortens length of hospital stay after colorectal surgery: a doubleblinded, randomized, placebo-controlled trial. Ann Surg 2007;246: 192–200. 26. Causey MW, Maykel JA, Hatch Q, et al. Identifying risk factors for renal failure and myocardial infarction following colorectal surgery. J Surg Res 2011;170:32–7. 27. El-Gazzaz G, Geisler D, Hull T. Risk of clinical leak after laparoscopic versus open bowel anastomosis. Surg Endosc 2010;24:1898–903. 28. Kiran RP, da Luz MA, Remzi FH, et al. Factors associated with septic complications after restorative proctocolectomy. Ann Surg 2010;251: 436–40. 29. Kooby DA, Stockman J, Ben-Porat L, et al. Influence of transfusions on perioperative and long-term outcome in patients following hepatic resection for colorectal metastases. Ann Surg 2003;237:860–9. 30. Ondrula DP, Nelson RL, Prasad ML, et al. Multifactorial index of preoperative risk factors in colon resections. Dis Colon Rectum 1992;35: 117–22. 31. Ortega-Deballon P, Radais F, Facy O, et al. C-reactive protein is an early predictor of septic complications after elective colorectal surgery. World J Surg 2010;34:808–14. 32. Stender MT, Nielsen TS, Frokjaer JB, et al. High preoperative prevalence of deep venous thrombosis in patients with colorectal cancer. Br J Surg 2007;94:1100–3. 33. Lovely JK, Maxson PM, Jacob AK, et al. Case-matched series of enhanced versus standard recovery pathway in minimally invasive colorectal surgery. Br J Surg 2012;99:120–6. 34. Lee Y, Fleming FJ, Deeb AP, et al. A laparoscopic approach reduces short-term complications and length of stay following ileocolic resection in Crohn’s disease: an analysis of outcomes from the NSQIP database. Colorectal Dis 2012;14:572–7. 35. Feroci F, Kroning KC, Lenzi E, et al. Laparoscopy within a fast-track program enhances the short-term results after elective surgery for resectable colorectal cancer. Surg Endosc 2011;25:2919–25. 36. Muller S, Zalunardo MP, Hubner M, et al. A fast-track program reduces complications and length of hospital stay after open colonic surgery. Gastroenterology 2009;136:842–7. 37. Kehlet H. Fast-track colorectal surgery. Lancet 2008;371:791–3. 38. Mehta RH, Liang L, Karve AM, et al. Association of patient case-mix adjustment, hospital process performance rankings, and eligibility for financial incentives. JAMA 2008;300:1897–903. 39. Schoetz Jr DJ, Bockler M, Rosenblatt MS, et al. ‘‘Ideal’’ length of stay after colectomy: whose ideal? Dis Colon Rectum 1997;40: 806–10.

Conclusions Identifying predictors of deviation of LOS offers an opportunity to identify maximally cost-effective targets for hospital cost savings. We identified multiple factors that may serve as targets for specific cost-saving measures. DRG outcomes also offer a tool to more accurately compare results from hospitals with different case mixes.

References 1. Kronberg U, Kiran RP, Soliman MS, et al. A characterization of factors determining postoperative ileus after laparoscopic colectomy enables the generation of a novel predictive score. Ann Surg 2011;253: 78–81. 2. Retchin SM, Penberthy L, Desch C, et al. Perioperative management of colon cancer under Medicare risk programs. Arch Intern Med 1997;157:1878–84. 3. Bokey EL, Chapuis PH, Fung C, et al. Postoperative morbidity and mortality following resection of the colon and rectum for cancer. Dis Colon Rectum 1995;38:480–6. 4. Colorectal Cancer Collaborative Group. Surgery for colorectal cancer in elderly patients: a systematic review. Lancet 2000;356:968–74. 5. Faiz O, Haji A, Burns E, et al. Hospital stay amongst patients undergoing major elective colorectal surgery: predicting prolonged stay and readmissions in NHS hospitals. Colorectal Dis 2011;13:816–22. 6. Reddy KM, Meyer CE, Palazzo FF, et al. Postoperative stay following colorectal surgery: a study of factors associated with prolonged hospital stay. Ann R Coll Surg Engl 2003;85:111–4. 7. Rhodes RS, Sharkey PD, Horn SD. Effect of patient factors on hospital costs for major bowel surgery: implications for managed health care. Surgery 1995;117:443–50. 8. Averill RF, Goldfield N, Hughes JS, et al. All patient refined diagnosis related groups (APR-DRGs). Methodology overview. Wallingford, CT: 3M Health Information System; 2003. 9. Averill RF, Kalison MJ, Vertrees JC, et al. Achieving short-term Medicare savings through the expansion of the prospective payment system. Health Care Manage Rev 1996;21:18–25. 10. Pirson M, Schenker L, Martins D, et al. What can we learn from international comparisons of costs by DRG? Eur J Health Econ 2013;14: 67–73. 11. Trocchi P, Kluttig A, Dralle H, et al. Thyroid cancer surgery in Germany: an analysis of the nationwide DRG statistics 2005-2006. Langenbecks Arch Surg 2012;397:421–8. 12. Oostenbrink JB, Rutten FF. Cost assessment and price setting of inpatient care in The Netherlands. The DBC case-mix system. Health Care Manag Sci 2006;9:287–94. 13. Christensen HK, Thaysen HV, Rodt SA, et al. Short hospital stay and low complication rate are possible with a fully implemented fast-track model after elective colonic surgery. Eur Surg Res 2011;46:156–61. 14. Kehlet H, Mogensen T. Hospital stay of 2 days after open sigmoidectomy with a multimodal rehabilitation programme. Br J Surg 1999;86: 227–30. 15. Murphy ME, Noetscher CM. Reducing hospital inpatient lengths of stay. J Nurs Care Qual 1999;Spec No:40–54. 16. Schlachta CM, Burpee SE, Fernandez C, et al. Optimizing Recovery After Laparoscopic Colon Surgery (ORAL-CS): effect of intravenous ketorolac on length of hospital stay. Surg Endosc 2007;21:2212–9.