Perioperative Fluid Utilization Variability and Association With Outcomes

ORIGINAL ARTICLE Perioperative Fluid Utilization Variability and Association With Outcomes Considerations for Enhanced Recovery Efforts in Sample US Surgical Populations Julie K. M. Thacker, MD, William K. Mountford, PhD,y Frank R. Ernst, PharmD, MS,z Michelle R. Krukas, MA,z and Michael (Monty) G. Mythen, MBBS, MD, FRCA, FFICM, FCAI (Hon) Objectives: To study current perioperative fluid administration and associated outcomes in common surgical cohorts in the United States. Background: An element of enhanced recovery care protocols, optimized perioperative fluid administration may be associated with improved outcomes; however, there is currently no consensus in the United States on fluid use or the effects on outcomes of this use. Methods: The study included all inpatients receiving colon, rectal, or primary hip or knee surgery, 18 years of age or older, who were discharged from a hospital between January 1, 2008 and June, 30 2012 in the Premier Research Database. Patient outcomes and intravenous fluid utilization on the day of surgery were summarized for each surgical cohort. Regression models were developed to evaluate associations of high or low day-of-surgery fluids with the likelihood of increased hospital length of stay (LOS), total costs, or postoperative ileus. Results: The study showed significant associations between high fluid volume given on the day of surgery with both increased LOS (odds ratio 1.10 1.40) and increased total costs (odds ratio 1.10 1.50). High fluid utilization was associated with increased presence of postoperative ileus for both rectal and colon surgery patients. Low fluid utilization was also associated with worse outcomes. Conclusions: According to results from this review of current practice in US hospitals, fluid optimization would likely lead to decreased variability and improved outcomes. Keywords: enhanced recovery, intraoperative fluid use, perioperative fluid, surgical costs, surgical outcomes (Ann Surg 2016;263:502 510) Despite improvements in surgical technique, anesthesia, and perioperative care, colorectal operations are plagued with morbidity and complicated postsurgical inpatient courses. In addition, orthopedic procedures are carried out more frequently each year, with significant costs. Any effort to avoid complications or unnecessary hospital days in these populations is warranted. Throughout the 1990s and early 2000s, the benefits of concerted From the Department of Surgery, Duke University School of Medicine, Durham, NC; ydepartment of Research Quintiles, Durham, NC; zdepartment of Research Premier, Inc, Charlotte, NC; and Surgical Outcomes Research Centre (SOuRCe), University College London Hospitals National Institute of Health Research Biomedical Research Centre, London, UK. Disclosure: Supported by funding from Deltex Medical (to W.M., F.E., and M.K.). The authors declare no conflicts of interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal s Web site (www.annalsofsurgery.com). Reprints: Julie K. M. Thacker, MD, Department of Surgery, Duke University School of Medicine, Durham, NC, USA 27710. E-mail: julie.thacker@duke.edu. Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. ISSN: 0003-4932/14/26105-0821 DOI: 10.1097/SLA.0000000000001402 efforts to implement multimodal perioperative care methods were reported. 1 Described as Enhanced Recovery After Surgery (ERAS), fast-track surgery, perioperative goal-directed therapy, or perioperative care mapping, these efforts represent multidisciplinary preoperative, intraoperative, and postoperative care pathways. 2,3 The primary aim of this multimodal approach to perioperative care is to reduce the length of hospitalization following elective surgery by maintaining preoperative organ function and reducing postsurgical stress. The key elements of evidence-based, multimodal, perioperative care in Europe have been summarized in Consensus Guidelines established in 2009, 4 and subsequently updated in 2012. 5 The 20 or so elements defined in these consensus statements include preoperative assessment and education, optimal perioperative fluid maintenance, standardized regimens for analgesia and anesthesia, and early mobilization. Although such a perioperative regimen has been shown to decrease length of stay (LOS) and improve outcomes, the successful implementation of these protocols requires patient engagement and multidisciplinary teamwork. 6 Acceptance and implementation of enhanced recovery principles has not been swift or widespread in the United States. The team approach to managing outcomes with multiple causative factors has been a challenge to gaining wide adoption of these principles. Elsewhere, however, the payment structure of the healthcare system has facilitated change. Implementation in England, for example, was first established as a National Health Service (NHS) project. 7 The success of the British effort is measured as a marker of acceptance; those centers with proof of adoption have shown expected improvements. 8 Postoperative elements have been electively supported at random US centers, although no nation-wide analysis or implementation effort has occurred. Evidence-based postoperative care programs have been defined in the US literature as enhanced recovery, whereas the intraoperative elements are less often reported. The anesthesia literature abounds with discussion of intraoperative fluid management regimens, but these are largely divorced from postoperative fluid management strategies or enhanced recovery efforts. When both the perioperative surgeon-driven elements and the intraoperative pain and fluid management anesthesia-driven elements are combined, the benefits of enhanced recovery are maximized. 4,5 One outcome addressed by enhanced recovery efforts in colorectal surgery (CRS) is postoperative ileus (POI). Often a driver of postoperative LOS, POI develops less frequently in CRS populations with directed perioperative fluid management protocols. 6 Postoperative ileus also occurs after orthopedic operations. Although POI is considered a medical complication to the orthopedic patient and a surgical complication to the gastrointestinal surgery patient, the effect of POI on LOS and postoperative recovery is the same. Attributing complications of an orthopedic patient or bowel surgery patient to perioperative fluid management is not definitively 502 www.annalsofsurgery.com Annals of Surgery Volume 263, Number 3, March 2016

Annals of Surgery Volume 263, Number 3, March 2016 Perioperative Fluid Utilization Variability supported in the literature; however, a general understanding that algorithm-led hemodynamic management including individualized goal-directed fluid management may yield improved outcomes is reported. 3 To best extrapolate a proposed perioperative fluid management protocol, an understanding of current perioperative fluid management is necessary. With acknowledgment of the paucity of published data on the impact of intraoperative and perioperative fluid management on surgical outcomes from these common procedures in the United States, this study was designed to summarize intravenous fluid utilization, in particular on the day of surgery (DOS), and patient outcomes, including the presence of POI, hospital LOS, in-hospital mortality, total hospital costs, and hospital readmissions (30, 60, and 90 days following index discharge) among adult inpatients receiving elective colon surgery, rectal surgery, or hip or knee replacement procedures. We hypothesized that there would be substantial variation in fluid utilization on the DOS and associations between both high- and low-volume usage and increased LOS, hospital costs, and POI. METHODS Data Source The Premier Research Database 9 contains information on over 67 million inpatient hospital discharges (6 million discharges per year) one-fifth of all annual US acute care hospitalizations from the year 2000 to the present. The database contains standard hospital discharge files that include patient demographic information. Patients can be tracked across the inpatient and hospital-based outpatient settings, as well as across visits with a unique identifier. In addition to the data elements available in most standard hospital discharge files, the Premier database contains a date-stamped record of all charge master items, including procedures, medications, laboratory and diagnostic services, and therapeutic services at the individual patient level. Drug utilization information is available by day of stay and includes quantity, dosing, strength, and cost. Costs are those reported by the hospital, and represent the hospital s internal perspective; they are not the charge to a payer and are not the amount reimbursed to the hospital. Patient Population This retrospective, observational, database analysis included inpatients, 18 years of age or older, who were discharged from 524 contributing hospitals between January 1, 2008, and June 30, 2012. Patients of interest had procedure codes from the International Classification of Diseases 9th Revision, Clinical Modification (ICD-9) signifying colon, rectal, or hip or knee replacement procedures (see Supplemental Digital Content Appendix, available at http://links.lww.com/sla/a837). To increase the likelihood that the hospital visit was directly related to the surgical procedure, patients were required to have the identifying ICD-9 procedure code(s) of interest occur on the day of hospital admission. Furthermore, patients with hospital admissions that were not listed as elective were excluded. The colon surgery population was subcategorized by open versus laparoscopic procedures, which was defined by the ICD-9 code used to identify the patient. When both open and laparoscopic ICD-9 procedure codes for the same surgical procedure were present on the day of admission, the patient was removed from the subgroup evaluations to maintain clear separation between the subgroups. Patient Outcomes The main outcomes of interest were hospital LOS, in-hospital mortality, total hospital costs, intravenous fluid utilization, presence of POI (identified by ICD-9 diagnosis codes 560.1 and 997.4x), and all-cause hospital readmissions (30, 60, and 90 days following index discharge, where the patient was alive at discharge). Fluid utilization was defined using the charge master data, which were provided at the day of service level. Therefore, fluids were summarized for the total duration of the index hospital stay and stratified by time period: during the DOS or on days postsurgery (DPS). Considerable variation was defined as being below the 25th percentile (low volume quartile) or above the 75th percentile (high volume quartile). Thus, within each surgery type, patients in the middle between the 25th and 75th percentiles (reference volume) were compared to those in the lowest and highest quartiles. Statistical Analysis All analyses were stratified by surgical cohort: colon, rectal, and hip/knee replacement. Descriptive analyses were conducted to characterize the patient population by patient, clinical, and hospital attributes. Data measured on a continuous scale were expressed as mean, standard deviation, median, and interquartile range (IQR) (25th percentile [Q1] to 75th percentile [Q3]). Categorical data were expressed as counts/percentages in the categories. Student t tests were used for comparisons of continuous measures such as age, and x 2 tests were used for categorical variables. All statistical tests were two sided and significance was indicated by P-values less than 0.05. Logistic regression modeling was conducted to evaluate the relationship between the volume of fluids received by patients on the DOS and key outcomes, including LOS, total hospital costs, and occurrence of POI. The volume of fluids received on the DOS was categorized into 3 levels: within the IQR (values lying between the 25th and 75th percentile); within Quartile 1 (Q1, defined as being in the lower 25th percentile of volume amounts), referred to as low; and within Quartile 4 (Q4, defined as being in the upper 25th percentile), referred to as high. The moderate IQR was designated as the referent level in all logistic models. Separate logistic regression models were developed for the following outcomes: excess LOS (top 25%), excess total hospital costs (top 25%), and presence of POI. Models were created to compute both unadjusted and adjusted results where, in the latter case, multivariable adjustment was made for patient and hospital covariates. The covariates varied by model, but generally included age, race, sex, Charlson Comorbidity Index (CCI), Medicare Severity Diagnosis Related Groups (MS-DRG), 10 and hospital teaching status, location (urban versus rural), US Census region, and number of beds. For each adjusted model, covariates eligible for inclusion in the models were selected using a stepwise selection method until all variables remaining in the model were significant at P 0.1. RESULTS Patient Characteristics The study included 3 groups: 84,722 colon, 22,178 rectal, and 548,526 primary hip or knee replacement surgical patients (Table 1). Overall, the study population had a mean age >60 years, with more females as compared to males, and a larger percentage of white race for all surgical cohorts. The most common insurance providers were Medicare (43%) and managed care (30%). The study population had geographic representation from all US regions. Furthermore, the study population included admissions from urban (89%) and rural (7%) hospitals, teaching and nonteaching (55%) hospitals, and small (<100 hospital beds), medium, and large (500þ hospital beds) hospitals. Unadjusted Outcomes Mean (SD) DOS fluid use was 3.5 2.4 L (median 3.1, IQR 1.7 5) for patients undergoing colon surgery, 3.8 2.7 L (median ß 2015 Wolters Kluwer Health, Inc. All rights reserved. www.annalsofsurgery.com 503

Thacker et al Annals of Surgery Volume 263, Number 3, March 2016 TABLE 1. Patient and Hospital Characteristics by Surgical Cohort Surgical Cohort Colon Rectal Hip/Knee N (%) N (%) N (%) Total patients 84,722 100.0 22,178 100.0 548,526 100.0 Patient demographics Age, yr 18 44 9355 11.0 2524 11.4 15,432 2.8 45 64 35,776 42.2 10,009 45.1 229,004 41.8 65 74 21,410 25.3 5321 24.0 177,592 32.4 75 84 14,426 17.0 3345 15.1 108,545 19.8 85þ 3755 4.4 979 4.4 17,953 3.3 Mean (SD) 62.3 (14.2) 61.4 (14.2) 65.7 (11) Median (IQR) 63 (53 73) 62 (52 72) 66 (58 74) Sex Female 45,868 54.1 11,762 53.0 331,591 60.5 Male 38,843 45.9 10,415 47.0 216,890 39.5 Race/ethnicity Black 7128 8.4 1387 6.3 38,864 7.1 Hispanic 2497 3.0 561 2.5 8916 1.6 Other 12,891 15.2 3841 17.3 81,282 14.8 White 62,206 73.4 16,389 73.9 419,464 76.5 Primary payer Commercial 6808 8.0 1800 8.1 38,094 6.9 Managed care 30,865 36.4 8546 38.5 166,756 30.4 Medicaid 3297 3.9 1061 4.8 14,052 2.6 Medicare 39,290 46.4 9547 43.1 303,114 55.3 Other 4462 5.3 1224 5.5 26,510 4.8 Charlson Comorbidity Index Mean (SD) 3.8 (3) 4.1 (3) 1.9 (1.5) Median (IQR) 3 (2 5) 3 (2 8) 1 (1 2) Hospital characteristics Provider region Midwest 16,271 19.2 3728 16.8 107,506 19.6 Northeast 15,659 18.5 4667 21.0 108,008 19.7 South 36,110 42.6 9340 42.1 224,289 40.9 West 16,682 19.7 4443 20.0 108,723 19.8 Population served Rural 9348 11.0 1548 7.0 62,261 11.4 Urban 75,374 89.0 20,630 93.0 486,265 88.7 Teaching status y No 51,487 60.8 12,206 55.0 335,898 61.2 Yes 33,235 39.2 9972 45.0 212,628 38.8 Hospital bed size 001 099 2871 3.4 428 1.9 18,428 3.4 100 199 8139 9.6 1357 6.1 75,085 13.7 200 299 14,806 17.5 3353 15.1 107,618 19.6 300 499 31,318 37.0 7906 35.7 203,891 37.2 500þ 27,588 32.6 9134 41.2 143,504 26.2 Provider region as defined by US Census: https://www.census.gov/geo/www/us_regdiv.pdf1.4. yteaching hospital or nonteaching hospital, as defined by CMS to mean, a hospital engaged in an approved graduate medical examination (GME) residency program in medicine, osteopathy, dentistry, or podiatry. IQR indicates interquartile range; SD, standard deviation. 3.3, IQR 1.5 5.4) for rectal surgery patients, and 3.0 2.0 L (median 3, IQR 1.3 4.1) for hip/knee replacement patients (Table 2). Average (SD) DPS fluid use was 8.4 7.9 L (median 6.6, IQR 3.2 11) for patients receiving colon surgery, 9.3 9.1 L (median 7.1, IQR 3.1 12.3) for rectal surgery patients, and 2.2 2.4 L (median 1.5, IQR 1 3) for hip/knee replacement patients. Mean LOS was 6.3 days (median 5, IQR 4 7) for patients receiving colon surgery, 7.0 days (median 6, IQR 4 8) for rectal surgery patients, and 3.3 days (median 3, IQR 3 4) for hip/knee replacement patients. Average total hospital costs were $15,069 (median $12,485, IQR $9576 $16,822) for colon surgery, $17,284 (median $14,401, IQR $10,821 $19,882) for rectal surgery, and $16,397 (median $15,287, IQR $12,632 $18,898) for hip/ knee replacement. Postoperative ileus was diagnosed in approximately 18% of both colon surgery and rectal surgery cohorts, but was found in <1% of hip/knee replacement patients. The rate of colon surgery patients with readmissions for colon surgery patients ranged from 10.2% within 30 days to 14.7% within 90 days. For rectal surgery patients, readmissions ranged from 14.5% within 30 days to 24.6% within 90 days. Hip/knee replacement patients were readmitted within 30 days 9.5% of the time and 12.8% of the time within 90 days. Death during the index hospitalization was rare (<1%) for all surgical groups. 504 www.annalsofsurgery.com ß 2015 Wolters Kluwer Health, Inc. All rights reserved.

Annals of Surgery Volume 263, Number 3, March 2016 Perioperative Fluid Utilization Variability TABLE 2. Patient Outcomes and Fluid Utilization by Surgical Cohort Surgical Cohort Colon Rectal Hip/Knee N (%) N (%) N (%) Total patients 84,722 100.0 22,178 100.0 548,526 100.0 Length of stay (d) Mean (SD) 6.3 (5.8) 7 (5.8) 3.3 (1.6) Median (IQR) 5 (4 7) 6 (4 8) 3 (3 4) Inhospital mortality Yes 676 0.8 128 0.6 416 0.1 Postoperative ileus Yes 14,972 17.7 4017 18.1 3479 0.6 Inpatient hospital readmissions Readmission 30-d 8613 10.2 3207 14.5 51,869 9.5 60-d 10,744 12.7 4473 20.2 61,473 11.2 90-d 12,492 14.7 5458 24.6 70,292 12.8 Total length of stay (index plus readmit within 30 d) N 8613 10.2 3207 14.5 51,869 9.5 Mean (SD) 16.3 (13.3) 16.9 (13.2) 11.7 (6.4) Median (IQR) 13 (9 19) 14 (9 20) 10 (8 14) Total length of stay (index plus readmit within 60 d) N 10,744 12.7 4473 20.2 61,473 11.2 Mean (SD) 16.7 (13.8) 16.6 (13.1) 11.3 (7) Median (IQR) 13 (9 20) 13 (9 20) 10 (7 14) Total length of stay (index plus readmit within 90 d) N 12,492 14.7 5458 24.6 70,292 12.8 Mean (SD) 16.9 (14.5) 16.7 (13.4) 11 (7.4) Median (IQR) 13 (9 20) 13 (9 20) 9 (6 14) Hospitalization cost outcomes ($) Total cost N 83,835 99.0 21,953 99.0 542,677 98.9 Mean (SD) $15,069 ($10,036) $17,284 ($11,350) $16,397 ($5,897) Median (IQR) $12,485 ($9,576 $16,822) $14,401 ($10,821 $19,882) $15,287 ($12,632 $18,898) ICU day(s) cost N 11,870 14.0 3722 16.8 11,510 2.1 Mean (SD) $18,159 ($19,021) $18,438 ($16,979) $16,835 ($13,092) Median (IQR) $12,524 ($8,418 $19,895) $13,869 ($9,284 $20,824) $14,536 ($7,804 $21,301) Non-ICU day(s) cost N 83,733 98.8 21,920 98.8 542,574 98.9 Mean (SD) $12,760 ($6,813) $14,373 ($8,311) $16,083 ($5,736) Median (IQR) $11,478 ($8,659 $15,263) $12,937 ($9,303 $17,566) $15,148 ($12,491 $18,695) Fluid utilization outcomes (L) Total fluid use N 77,140 91.1 19,792 89.2 479,202 87.4 Mean (SD) 11.5 (14.2) 12.7 (13.8) 4.6 (14.1) Median (IQR) 9.6 (5 14.6) 10.3 (5 16.5) 4 (2 6.1) Day of surgery fluid use N 70,920 83.7 17,916 80.8 453,445 82.7 Mean (SD) 3.5 (2.4) 3.8 (2.7) 3 (2) Median (IQR) 3.1 (1.7 5) 3.3 (1.5 5.4) 3 (1.3 4.1) Day(s) postsurgery fluid use total N 70,436 83.1 18,153 81.9 336,388 61.3 Mean (SD) 8.4 (7.9) 9.3 (9.1) 2.2 (2.4) Median (IQR) 6.6 (3.2 11) 7.1 (3.1 12.3) 1.5 (1 3) IQR indicates interquartile range; L, liters, SD, standard deviation. Compared with patients having laparoscopic colon surgery, patients with open colon procedures tended to have greater unadjusted average LOS (7.1 6.6 vs 4.9 3.8 days; P < 0.0001), greater unadjusted average total costs ($15,675 10,980 vs $13,582 7890; P < 0.0001), and poorer patient-level outcomes, as measured by POI (19.8% vs 13.2%; P < 0.0001) and readmissions (11.6% vs 7.5% at 30 days; P < 0.0001) (Table 3). Despite statistically significant differences in all unadjusted outcomes between open and laparoscopic colon surgeries, the DOS fluid volume did not substantially vary between open and laparoscopic cases (median 3.1 L, mean 3.5 L for both groups; P ¼ 0.6975). ß 2015 Wolters Kluwer Health, Inc. All rights reserved. www.annalsofsurgery.com 505

Thacker et al Annals of Surgery Volume 263, Number 3, March 2016 TABLE 3. Patient Outcomes and Fluid Utilization by Surgery Type for Colon Procedures Colon Surgeries Open Laparoscopic N (%) N (%) P Total patients 48,066 100.0 30,443 100.0 Percentage of surgery type 56.7 35.9 Length of stay (d) Mean (SD) 7.1 (6.6) 4.9 (3.8) <0.0001 Median (IQR) 6 (4 8) 4 (3 5) Inhospital mortality Yes 537 1.1 86 0.3 <0.0001 Postoperative ileus Yes 9511 19.8 4025 13.2 <0.0001 Inpatient hospital readmissions Readmission 30-d 5575 11.6 2280 7.5 <0.0001 60-d 6987 14.5 2831 9.3 <0.0001 90-d 8150 17.0 3265 10.7 <0.0001 Total length of stay (index plus readmit within 30 d) N 5575 11.6 2280 7.5 Mean (SD) 17.4 (13.8) 13.5 (10.9) <0.0001 Median (IQR) 13 (9 21) 10 (7 16) Total length of stay (index plus readmit within 60 d) N 6987 14.5 2831 9.3 Mean (SD) 17.7 (14.3) 13.9 (11.4) <0.0001 Median (IQR) 13 (9 21) 10 (7 16) Total length of stay (index plus readmit within 90 d) N 8150 17.0 3265 10.7 Mean (SD) 18 (15.2) 14 (11.6) <0.0001 Median (IQR) 13 (9 21) 10 (7 17) Hospitalization cost outcomes ($) Total cost N 47,546 98.9 30,138 99.0 Mean (SD) $15,675 ($10,980) $13,582 ($7890) <0.0001 Median (IQR) $12,617 ($9563 $17,503) $11,900 ($9330 $15,377) ICU day(s) cost N 8626 18.0 2236 7.3 Mean (SD) $18,398 ($19,523) $16,584 ($17,162) <0.0001 Median (IQR) $12,502 ($8302 $20,315) $11,830 ($8456 $17,862) Non-ICU day(s) cost N 47,402 98.6 30,197 99.2 Mean (SD) $12,729 ($7148) $12,395 ($5988) <0.0001 Median (IQR) $11,294 ($8409 $15,302) $11,383 ($8809 $14,690) Fluid utilization outcomes (ml) Total fluid use N 43,911 91.4 27,505 90.4 Mean (SD) 12.7 (16) 9.4 (9.8) <0.0001 Median (IQR) 10.8 (5.7 16.1) 8.1 (4.4 12.1) Day of surgery fluid use total N 40,133 83.5 25,572 84.0 Mean (SD) 3.5 (2.4) 3.5 (2.3) 0.6975 Median (IQR) 3.1 (1.5 5) 3.1 (2 5) Day(s) postsurgery fluid use total N 40,781 84.8 24,397 80.1 Mean (SD) 9.3 (8.4) 6.6 (6.4) <0.0001 Median (IQR) 7.5 (4 12.2) 5 (3 8.2) IQR indicates interquartile range; L, liters; SD, standard deviation. Among the colon surgical cohort, we found a statistically significant association between fluid volumes received after the first postoperative day (Table 2, DPS) and the volume of fluids received on the DOS (Table 2, DOS). In particular, the mean DPS fluid volume on DOS increased from 6.7 L among patients in the lowest quartile of DOS fluids to 10.0 L in DPS fluids among patients in the highest quartile of DOS fluids (Table 4; P < 0.0001). In other words, patients who tended to receive lesser DOS fluids also tended to receive lesser DPS fluids and those with higher DOS fluids also tended to receive higher DPS fluids. 506 www.annalsofsurgery.com ß 2015 Wolters Kluwer Health, Inc. All rights reserved.

Annals of Surgery Volume 263, Number 3, March 2016 Perioperative Fluid Utilization Variability TABLE 4. Postsurgical Fluid Utilization by Day-of-surgery Quartile for Colon Surgery Cohort DOS Fluid Quartile Mean DPS Fluids (L) Q1 6.71 Q2 7.97 Q3 8.70 Q4 10.01 Linear trend test P < 0.0001. Modeled Outcomes The multivariable (adjusted) odds ratios (ORs) and corresponding 95% confidence intervals (CIs) describing the relationship between DOS fluids with LOS, total hospital costs, and POI are displayed in Figure 1. ORs from the univariable models were consistently similar to those in the multivariable models and, therefore, are not presented (available upon request). In colon surgery patients, both low and high fluid volumes were often associated with worse outcomes. Low fluid volume (Q1) was associated with 9% greater odds (OR 1.09, 95% CI 1.04 1.14) of prolonged LOS than moderate fluids (Q2 Q3), when controlling for patient and hospital covariates, whereas high fluid volume (Q4) was associated with 25% greater odds (OR 1.25, 95% CI 1.20 1.31) of longer LOS. Low fluid volume (Q1) was also associated with 17% greater odds (OR 1.17, 95% CI 1.12 1.23) of higher hospitalization costs, and high fluid volume (Q4) was associated with 30% greater odds (OR 1.30, 95% CI 1.24 1.36) of higher costs. In addition, low fluid volume (Q1) was associated with 12% greater odds (OR 1.12, 95% CI 1.06 1.18) of having POI, whereas high fluid volume (Q4) was associated with 10% greater odds (OR 1.10, 95% CI 1.05 1.16) of POI. In rectal surgery patients, both low and high fluid volumes were often associated with worse outcomes. Specifically, low fluid volume (Q1) was associated with 7% greater odds (OR 1.07, 95% CI 0.98 1.17) of prolonged LOS than moderate fluids (Q2 Q3), which was not statistically significant when controlling for patient and hospital covariates. By contrast, high fluid volume (Q4) was associated with 41% greater odds (OR 1.41, 95% CI 1.29 1.53) of longer LOS. Low fluid volume (Q1) was also associated with 3% greater odds (OR 1.03, 95% CI 0.94 1.13) of higher hospitalization costs, which was not statistically significant. However, high fluid volume (Q4) was associated with 53% greater odds (OR 1.53, 95% CI 1.40 1.67) of higher costs, which was significant. In addition, low fluid volume (Q1) was associated with 16% greater odds (OR 1.16, 95% CI 1.05 1.28) of having POI, whereas high fluid volume (Q4) was associated with 26% greater odds (OR 1.26, 95% CI 1.14 1.38) of POI. In hip/knee surgery patients as well, both low and high fluid volumes were often associated with worse outcomes. In particular, low fluid volume (Q1) was associated with 14% greater odds (OR 1.14, 95% CI 1.12 1.16) of prolonged LOS than moderate fluids (Q2 Q3), when controlling for patient and hospital covariates. High fluid volume (Q4) was associated with 3% greater odds (OR 1.03, 95% CI 1.02 1.05) of longer LOS. Low fluid volume (Q1) was also associated with 12% greater odds (OR 1.12, 95% CI 1.10 1.14) of higher hospitalization costs, and high fluid volume (Q4) was associated with 26% greater odds (OR 1.26, 95% CI 1.23 1.28) of higher costs. Low fluid volume (Q1) was associated with 10% greater odds (OR 1.10, 95% CI 1.00 1.20) of having POI, whereas high fluid volume (Q4) was associated with 4% greater odds (OR 1.04, 95% CI 0.95 1.14) of POI, but both were insignificant differences. Open Versus Laparoscopic Colon Surgery Additional models were developed to evaluate the relationship between DOS fluids and outcomes by open versus laparoscopic surgeries. No statistically significant differences in LOS, costs, or risk of POI were detected between open and laparoscopic surgical approaches, as shown by the overlapping 95% CIs at the bottom of Figure 1. In particular, low fluid volume (Q1) was associated with 10% greater (OR 1.1010, 95% CI 1.04 1.16) odds of prolonged LOS among the open colon surgery patients compared to 9% greater odds (OR 1.09, 95% CI 0.9999 1.20) among the laparoscopic colon surgery patients. For high fluid-volume colon surgery patients, increased LOS was 26% more likely (OR 1.26, 95% CI 1.19 1.33) for open surgery, versus 23% more likely (OR 1.23, 95% CI 1.12 1.35) for laparoscopic surgery. Low fluid volume (Q1) in open colon surgery cases was associated with 16% greater odds (OR 1.16, 95% CI 1.10 1.23) of higher costs, compared to 23% greater odds (OR 1.23, 95% CI 1.13 1.33) among the laparoscopic colon surgery patients. High fluid volume (Q4) in open colon surgery cases was associated with 35% greater odds (OR 1.35, 95% CI 1.27 1.43) of higher costs, compared to 19% greater odds (OR 1.19, 95% CI 1.10 1.30) among the laparoscopic colon surgery patients. Then, finally, low fluid volume (Q1) in open colon surgery cases was associated with 14% greater odds (OR 1.1414, 95% CI 1.0707 1.22) of POI, compared to 12% greater odds (OR 1.12, 95% CI 1.02 1.23) among the laparoscopic colon surgery patients. High fluid volume (Q4) in open colon surgery cases was associated with 13% greater odds (OR 1.1313, 95% CI 1.0606 1.20) of POI, compared to 8% greater odds (OR 1.08, 95% CI 0.99 1.18) among the laparoscopic colon surgery patients. DISCUSSION The fluid utilization and associated outcomes of 3 common US surgical patient populations were evaluated by retrospective analysis of the Premier database. From this approach, we studied fluid given on DOS in a demographically diverse patient cohort, well represented in our large patient population from a range of hospital types. The first aim of this study was to characterize the current DOS intravenous fluid use during colon, rectal, and hip/knee operations. A wide range of fluid given on DOS is observed, as shown in Table 2. For example, 25% of patients received less than the following DOS amounts: 1.7 L for colon, 1.5 L for rectal, and 1.3 L for hip/knee surgeries. Another 25% of patients received more than 5.0 L for colon, 5.4 L for rectal, and 4.1 L for hip/knee surgeries on the DOS. Furthermore, 25% of patients received less than the following DPS amounts: 3.2 L for colon, 3.1 L for rectal, and 1.0 L for hip/knee surgeries. Moreover, 25% of patients received more than 11.0 L for colon, 12.3 L for rectal, and 3.0 L for hip/knee surgery patients in the days following surgery. The significant variability of DOS fluid administration corresponds to our background impression that protocoled/goal-directed fluid management and enhanced recovery pathways are not routinely followed in the United States. The second aim was to define associations between DOS fluid use and outcomes. In every group, we identified an association between receiving the greatest amount of fluid and worse outcomes. High (Q4) DOS fluid patients had increased likelihood of longer LOS, increased POI, and increased cost compared to patients with DOS fluids in the second and third quartiles. Although the overall LOS and POI incidence are consistent with those reported in the literature, 11 we have shown an association with the amount of fluid given on the DOS. In all instances, adjusted and unadjusted, worse outcomes are seen in the extremes of fluid administration. By defining quartiles 2 and 3 as presumptively optimal with improved ß 2015 Wolters Kluwer Health, Inc. All rights reserved. www.annalsofsurgery.com 507

Thacker et al Annals of Surgery Volume 263, Number 3, March 2016 FIGURE 1. Odds of increased/decreased (% and 95% confidence intervals) outcomes (length of stay, cost, or ileus) comparing reference group of patients (quartiles 2 and 3) to those who received lower (Q1) and higher (Q4) volumes of fluid on the day of surgery. 508 www.annalsofsurgery.com ß 2015 Wolters Kluwer Health, Inc. All rights reserved.

Annals of Surgery Volume 263, Number 3, March 2016 Perioperative Fluid Utilization Variability FIGURE 1. (Continued). outcomes over the extremes, we conclude that variation in intraoperative fluid delivery impacts surgical outcomes. It is important to note the independence of the impact of fluids given and MS-DRG and CCI. The association of improved outcomes with our defined optimal range of fluid given is maintained while controlling for every level of patient severity of illness and baseline comorbidity status. Regardless of procedure category or comorbidity index, patients receiving fluids in the second and third quartile ranges had better outcomes and lower costs. Our results suggest that an applicable, protocoled approach toward optimal fluid management may improve outcomes. In the orthopedic cohort, the variability of fluid use was less than in the colorectal cohorts. Despite this, and the categorization by the Charlson index and MS-DRG, the orthopedic cohort experienced the best outcomes in our defined optimal fluid categories. The association of improved surgical outcomes with amount of intraoperative fluid is independent of MS-DRG. In other words, even in the defined healthier populations, intraoperative optimized fluid strategies may positively impact outcomes. Similarly, we note a difference in outcomes between the laparoscopic and open colorectal groups. Consistent with the previous literature, surgical outcomes are improved with laparoscopic approaches to surgery. Again, we report the association between fluid administration and improved outcomes is seen in each group, with multivariable analysis. Within the laparoscopic group and within the open group, patients with high or low DOS fluids tended to have poorer outcomes, regardless of surgical approach. Adoption and application of patient-centered care pathways and protocols can reduce variability and improve outcomes. Recent meta-analyses emphasized that goal-directed fluid protocols are associated with reduced variability in fluid management and better outcomes. 3,12 Goal-directed fluid therapy (GDFT) utilizes a cardiac output monitor to individualize fluid therapy and aims to restrict volume therapy to the minimum required to maintain central euvolemia. Background crystalloid infusion rates ( maintenance fluid ) may also be dramatically reduced. The relative contributions of fluid restriction and the use of vasoactive drugs to increase oxygen delivery remain uncertain, but flow-guided algorithms for fluid management are associated with improved outcomes with little or any evidence of harm. 13 Although some degree of uncertainty exists about the need for GDFT when considering low-risk patients undergoing laparoscopic procedures in enhanced recovery care programs, our observations support the importance of avoiding either excessive fluid restriction or overload in all our patients. 6,13 The primary strengths of our study are the size of our patient sample and the availability of data allowing multivariable analysis. With the diversity of hospital size, patient characteristics, and patient comorbidities represented in our sample, we combat the biases of individual or hospital patient case complexity and outlier provider variables. We did not account for biases introduced by nonrandomization of treatment, such as with propensity score methods. Instead, we observed variability in real-world use of fluids in our study and applied multivariable regression techniques to account for confounding factors. Because the granularity of the Premier database permits no finer analyses of timing than a calendar day, our study was not able to distinguish the fluids used during surgery (intraoperative time) versus perioperatively (preoperative, intraoperative, and postoperative time). However, we were able to separate the day of a surgical procedure from days following surgery. The hospital cost data in our analysis were captured from financial information gathered by the hospitals during the course of treatment. The limitations of this approach are the possible underreporting of clinical outcomes of interest, and the reliance on ICD-9 coding to identify clinically related information such as procedures and diagnosed comorbid conditions. For example, a diagnosis of POI may be underreported due to the lack of consistent documentation in patient medical records, whereas a readmission is increasingly unlikely to be missed. However, this limitation should have minimal impact on our conclusion given that over- or under-reporting would likely be consistent within institutions and, therefore, consistent in effect. In conclusion, we analyzed differences in outcomes as they may relate to the volume of fluid given on the DOS. Based on the poorer outcomes observed in patients receiving both restricted and liberal amounts of fluid, we advocate continued examination of the effect of patient-centered care pathways that incorporate protocolized optimal intraoperative intravenous fluid management to enhance surgical recovery. REFERENCES 1. Kehlet H. Multimodal approach to control postoperative pathophysiology and rehabilitation. Br J Anaesth. 1997;78:606 617. 2. Grocott MP, Dushianthan A, Hamilton MA, et al. Perioperative increase in global blood flow to explicit defined goals and outcomes following surgery. Cochrane Database Syst Rev. 2012;11:CD004082. 3. Spanjersberg WR, Reurings J, Keus F, et al. Fast track surgery versus conventional recovery strategies for colorectal surgery. Cochrane Database Syst Rev. 2011;2:CD007635. 4. Lassen K, Soop M, Nygren J, et al. Consensus review of optimal perioperative care in colorectal surgery: Enhanced Recovery After Surgery (ERAS) Group recommendations. Arch Surg. 2009;144:961 969. 5. Gustafsson UO, Scott MJ, Schwenk W, et al. Guidelines for perioperative care in elective colonic surgery: Enhanced Recovery After Surgery (ERAS(R)) Society recommendations. Clin Nutr. 2012;31:783 800. ß 2015 Wolters Kluwer Health, Inc. All rights reserved. www.annalsofsurgery.com 509

Thacker et al Annals of Surgery Volume 263, Number 3, March 2016 6. Miller TE, Thacker JK, White WD, et al. Reduced length of hospital stay in colorectal surgery after implementation of an enhanced recovery protocol. Anesth Analg. 2014;118:1052 1061. 7. Mythen MG. Spread and adoption of enhanced recovery from elective surgery in the English National Health Service. Can J Anaesth. 2015;62: 105 109. 8. Knott A, Pathak S, McGrath JS, et al. Consensus views on implementation and measurement of enhanced recovery after surgery in England: Delphi study. BMJ open. 2012;2:e001878. 9. Premier database. Premier Inc. (Charlotte, NC). Available at: http://www.pre mierinc.com/transforming-healthcare/healthcare-performance-improvement/ premier-research-services/. Accessed 5 February, 2015. 10. Medicare Program; Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2008 Rates; Final Rule. Available at: http:// www.cms.gov/medicare/medicare-fee-for-service-payment/acuteinpatient PPS/IPPS-Regulations-and-Notices-Items/CMS1228401.html. 11. Chapuis PH, Bokey L, Keshava A, et al. Risk factors for prolonged ileus after resection of colorectal cancer: an observational study of 2400 consecutive patients. Ann Surg. 2013;257:909 915. 12. Pearse RM, Harrison DA, MacDonald N, et al. Effect of a perioperative, cardiac output-guided hemodynamic therapy algorithm on outcomes following major gastrointestinal surgery: a randomized clinical trial and systematic review. JAMA. 2014;311:2181 2190. 13. Bennett-Guerrero E. Hemodynamic goal-directed therapy in high-risk surgical patients. JAMA. 2014;311:2177 2178. 510 www.annalsofsurgery.com ß 2015 Wolters Kluwer Health, Inc. All rights reserved.