- Email: [email protected]
Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine cancer
YGYNO-977085; No. of pages: 9; 4C: Gynecologic Oncology xxx (2018) xxx–xxx
Contents lists available at ScienceDirect
Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno
Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine cancer☆ Surbhi Agrawal a, Ling Chen a, Ana I. Tergas a,c,d,e, June Y. Hou a,d,e, Caryn M. St. Clair a,d,e, Cande V. Ananth a,c, Dawn L. Hershman a,b,d,e, Jason D. Wright a,d,e,⁎ a
Department of Obstetrics and Gynecology, Columbia University College of Physicians and Surgeons, USA Department of Medicine, Columbia University College of Physicians and Surgeons, USA Department of Epidemiology, Joseph L. Mailman School of Public Health, Columbia University, USA d Herbert Irving Comprehensive Cancer Center, Columbia University College of Physicians and Surgeons, USA e New York Presbyterian Hospital, USA b c
H I G H L I G H T S • Perioperative risk factors account for approximately 25% of the variation in prolonged LOS. • A substantial proportion of the variation in LOS remains unexplained by measurable factors. • There are similar predictors of prolonged LOS for abdominal and minimally invasive hysterectomy.
a r t i c l e
i n f o
Article history: Received 1 March 2018 Received in revised form 12 March 2018 Accepted 12 March 2018 Available online xxxx
a b s t r a c t Objective. We examined the influence of modifiable (intraoperative factors and complications) and nonmodifiable (clinical and demographic characteristics) factors on length of stay (LOS) for women who underwent hysterectomy for uterine cancer. Methods. The National Surgical Quality Improvement Program database was used to identify women who underwent hysterectomy for uterine cancer from 2006 to 2015. The association between demographic, preoperative, intraoperative, and postoperative factors and LOS was examined. The primary outcome was prolonged LOS (N75th an3 N 90th percentiles). Model fit statistics were used to assess the importance of each group of characteristics. Results. Of 19,084 women identified, 6082 (31.9%) underwent abdominal and 13,002 (68.1%) underwent minimally invasive hysterectomy. In the abdominal hysterectomy group, the 75th and 90th percentiles for LOS were 5 and 8 days, respectively. All risk factors combined accounted for 23.6% of the variation in LOS N75th percentile. Demographic characteristics explained 4.0%, preoperative factors 7.0%, intraoperative factors 7.9%, and postoperative characteristics 9.7% of variation in prolonged LOS. In the minimally invasive group, the 75th and 90th percentiles for LOS were 1 and 2 days, respectively. The combined risk factors explained 16.2% of the variation in prolonged LOS. Demographic characteristics accounted for 6.2%, preoperative factors 4.1%, intraoperative factors 6.9%, and postoperative characteristics 1.3% of variation in prolonged LOS. Similar patterns were seen when prolonged LOS was defined as N90th percentile. Conclusion. Perioperative risk factors account for approximately one quarter of the variation in prolonged LOS. Overall, a substantial proportion of the variation in LOS remains unexplained by measurable patient and hospital factors which may limit the utility of LOS as a quality metric for endometrial cancer. © 2018 Published by Elsevier Inc.
☆ Dr. Wright (NCI R01CA169121-01A1) and Dr. Hershman (NCI R01 CA166084) are recipients of grants from the National Cancer Institute. ⁎ Corresponding author at: Division of Gynecologic Oncology, Columbia University College of Physicians and Surgeons, 161 Fort Washington Ave, 8th Floor, New York, NY 10032, USA. E-mail address: [email protected] (J.D. Wright).
1. Introduction Hospital length of stay (LOS) has been proposed as a quality metric and influences reimbursement for a variety of surgical procedures. In the context of growing healthcare costs, the Centers for Medicare and
https://doi.org/10.1016/j.ygyno.2018.03.048 0090-8258/© 2018 Published by Elsevier Inc.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
2
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 1 Demographic, preoperative, intraoperative and postoperative conditions of patients who underwent abdominal hysterectomy by prolonged length of stay.
All Demographics Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Elective surgery No Yes Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Smoking No Yes Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes CHF No Yes Bleeding disorder No Yes Open wound No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes Median (IQR) Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
LOS ≤ 90th %
LOS N 90th %
N
(%)
N
(%)
4834
(79.5)
1248
(20.5)
5545
(91.2)
537
(8.8)
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
177 324 355 537 613 838 928 1004
(3.7) (6.7) (7.3) (11.1) (12.7) (17.3) (19.2) (20.8)
45 62 72 160 187 228 235 242
(3.6) (5.0) (5.8) (12.8) (15.0) (18.3) (18.8) (19.4)
201 366 392 623 717 973 1063 1141
(3.6) (6.6) (7.1) (11.2) (12.9) (17.5) (19.2) (20.6)
21 20 35 74 83 93 100 105
(3.9) (3.7) (6.5) (13.8) (15.5) (17.3) (18.6) (19.6)
607 1280 1742 1205
(12.6) (26.5) (36.0) (24.9)
97 264 412 475
(7.8) (21.2) (33.0) (38.1)
666 1431 1985 1463
(12.0) (25.8) (35.8) (26.4)
38 113 169 217
(7.1) (21.0) (31.5) (40.4)
3417 604 236 577
(70.7) (12.5) (4.9) (11.9)
843 240 59 106
(67.5) (19.2) (4.7) (8.5)
3905 741 267 632
(70.4) (13.4) (4.8) (11.4)
355 103 28 51
(66.1) (19.2) (5.2) (9.5)
134 3776 924
(2.8) (78.1) (19.1)
221 824 203
(17.7) (66.0) (16.3)
211 4294 1040
(3.8) (77.4) (18.8)
144 306 87
(26.8) (57.0) (16.2)
755 966 3098 15
(15.6) (20.0) (64.1) (0.3)
215 256 763 14
(17.2) (20.5) (61.1) (1.1)
870 1113 3544 18
(15.7) (20.1) (63.9) (0.3)
100 109 317 11
(18.6) (20.3) (59.0) (2.0)
3733 280 821
(77.2) (5.8) (17.0)
886 114 248
(71.0) (9.1) (19.9)
4234 344 967
(76.4) (6.2) (17.4)
385 50 102
(71.7) (9.3) (19.0)
4373 461
(90.5) (9.5)
1121 127
(89.8) (10.2)
5006 539
(90.3) (9.7)
488 49
(90.9) (9.1)
4728 70 17 19
(97.8) (1.4) (0.4) (0.4)
1166 70
(93.4) (5.6)
(92.2) (7.1)
a
a
a
a
a
(97.4) (1.8) (0.4) (0.4)
495 38
a
5399 102 23 21
a
a
5393 152
(97.3) (2.7)
509 28
(94.8) (5.2)
5524 21
(99.6) (0.4)
525 12
(97.8) (2.2)
5420 125
(97.7) (2.3)
501 36
(93.3) (6.7)
5512 33
(99.4) (0.6)
525 12
(97.8) (2.2)
P-value
0.050
0.17
b0.001
b0.001
b0.001
0.002
b0.001
b0.001
b0.001
b0.001
b0.001
0.01
0.50
0.66
b0.001
b0.001
b0.001
0.001
4713 121
(97.5) (2.5)
1189 59
(95.3) (4.7)
4821 13
(99.7) (0.3)
1228 20
(98.4) (1.6)
4739 95
(98.0) (2.0)
1182 66
(94.7) (5.3)
4810 24
(99.5) (0.5)
1227 21
(98.3) (1.7)
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
1963 2606 139
(40.6) (53.9) (2.9)
336 781 125
(26.9) (62.6) (10.0)
2185 3032 198
(39.4) (54.7) (3.6)
114 355 66
(21.2) (66.1) (12.3)
138
(101–189)
185
(128–242)
143
(103–196)
186
(129–250)
4531 303
(93.7) (6.3)
1056 192
(84.6) (15.4)
5159 386
(93.0) (7.0)
428 109
(79.7) (20.3)
4807 27
(99.4) (0.6)
1177 71
(94.3) (5.7)
5490 55
(99.0) (1.0)
494 43
(92.0) (8.0)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 b0.001
b0.001
b0.001
P-value
b0.001
b0.001
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
3
Table 1 (continued)
No Yes Progressive renal insufficiency No Yes UTI No Yes CVA/stroke with deficit No Yes DVT/thrombophlebitis No Yes Transfusion No Yes Sepsis No Yes Shock No Yes Clavian complication No Yes Any complication No Yes Readmission No Yes Reoperation No Yes
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
4795 39
(99.2) (0.8)
1196 52
(95.8) (4.2)
4821 13
(99.7) (0.3)
1229 19
(98.5) (1.5)
4738 96
(98.0) (2.0)
1159 89
(92.9) (7.1)
a
a
a
a
1236 12
(99.0) (1.0)
LOS ≤ 90th % P-value
LOS N 90th %
N
(%)
N
(%)
5490 55
(99.0) (1.0)
501 36
(93.3) (6.7)
5524 21
(99.6) (0.4)
526 11
(98.0) (2.0)
5411 134
(97.6) (2.4)
486 51
(90.5) (9.5)
5535 10
(99.8) (0.2)
a
a
a
a
5503 42
(99.2) (0.8)
502 35
(93.5) (6.5)
4816 729
(86.9) (13.1)
318 219
(59.2) (40.8)
5485 60
(98.9) (1.1)
488 49
(90.9) (9.1)
5526 19
(99.7) (0.3)
513 24
(95.5) (4.5)
5381 164
(97.0) (3.0)
414 123
(77.1) (22.9)
5187 358
(93.5) (6.5)
341 196
(63.5) (36.5)
5115 430
(92.2) (7.8)
464 73
(86.4) (13.6)
5458 87
(98.4) (1.6)
481 56
(89.6) (10.4)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 4807 27
(99.4) (0.6)
1198 50
(96.0) (4.0)
4322 512
(89.4) (10.6)
812 436
(65.1) (34.9)
4789 45
(99.1) (0.9)
1184 64
(94.9) (5.1)
4820 14
(99.7) (0.3)
1219 29
(97.7) (2.3)
4723 111
(97.7) (2.3)
1072 176
(85.9) (14.1)
4592 242
(95.0) (5.0)
936 312
(75.0) (25.0)
4506 328
(93.2) (6.8)
1073 175
(86.0) (14.0)
4767 67
(98.6) (1.4)
1172 76
(93.9) (6.1)
P-value
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
IQR: interquartile range. Patients with sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h and unplanned intubation were considered to have a Clavian complication. Patients with pneumonia, ARF, UTI, CVA/stroke with deficit, coma, sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h, unplanned intubation, or DVT/thrombophlebitis were considered to have any complication. a Suppressed from reporting because some cell sizes b10.
Medicaid Services (CMS) have increasingly used LOS to encourage hospitals to become more efficient at providing care [1,2]. Hysterectomy is one of the most commonly performed surgical procedures in women [3]. To date, most studies have focused on comparing LOS and overall cost differences among the different surgical techniques for hysterectomy [4–8]. A single-institution study found that the average LOS was 4.4 days for abdominal, 1.2 days for laparoscopic, and 1.0 day for robotic hysterectomy for patients with uterine cancer [9]. However, aside from surgical route, relatively little is known about the perioperative factors that drive LOS and contribute to a prolonged LOS among hysterectomy patients. Given the increased focus on using LOS as a measure of quality and driver for reimbursement, understanding the factors that affect prolonged hospitalization is imperative to develop actionable strategies to reduce LOS. We therefore performed a population-based study to determine the relative contribution of modifiable (intraoperative events and complications) and non-modifiable (clinical and demographic factors) to prolonged LOS after hysterectomy for uterine cancer. 2. Methods We used the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Participant Use Data File (PUF) for the analysis. NSQIP is a nationally validated, outcomes-based program that collects data of patients undergoing major surgical procedures from participating hospitals [10]. Preoperative, intraoperative, and 30day postoperative variables from both the inpatient and outpatient
setting are collected to measure and improve surgical quality. Data are abstracted from medical charts by trained reviewers through a systematic sampling process. Participating hospitals are required to submit data from 42 of the 46 8-day cycles. The 8-day cycle is spaced throughout the year to minimize bias in case selection. Data quality is ensured by regular audit. Women who had uterine cancer and underwent abdominal or minimally invasive hysterectomy from 2006 to 2015 were included in this analysis. Patients who had other gynecologic malignancies or missing data for length of stay were excluded. Demographic characteristics included year of operation, age (b50, 50–59, 60–69, ≥70 years), race (white, black, other, unknown), and whether the surgery was elective (yes, no, unknown). For each woman, the following preoperative conditions were recorded: body mass index (BMI, normal b25 kg/m2, overweight 25–29.9 kg/m2, obese ≥30 kg/m2, and unknown), diabetes mellitus (insulin dependent, or non-insulin dependent), tobacco use (current smoker within one year), history of severe chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF) within 30 days of surgery, a bleeding disorder, an open wound with or without infection, functional health status prior to surgery (independent, partially dependent, totally dependent, and unknown), and Society of Anesthesiology (ASA) classification score (none, 1, 2, 3, 4–5, or unknown). Total operation time (in minutes) was noted for the procedure. Postoperative conditions included wound infection, pneumonia, pulmonary embolism (PE), urinary tract infection (UTI), deep vein thrombosis (DVT), sepsis, and reoperation within 30 days.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
4
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 2 Demographic, preoperative, intraoperative and postoperative conditions of patients who underwent minimally invasive hysterectomy by prolonged length of stay. LOS ≤ 75th %
All Demographics Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Elective surgery No Yes Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Smoking No Yes Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes CHF No Yes Bleeding disorder No Yes Open wound No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes Median (IQR) Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE
LOS N 75th %
LOS ≤ 90th %
N
(%)
N
(%)
10,441
(80.3)
2561
(19.7)
P-value
a
a
a
a
a
a
a
a
a
32 77 232 812 1414 2020 2644 3200
(0.3) (0.7) (2.2) (7.8) (13.5) (19.3) (25.3) (30.6)
51 71 98 311 383 489 510 639
(2.0) (2.8) (3.8) (12.1) (15.0) (19.1) (19.9) (25.0)
71 130 295 1015 1668 2331 2984 3637
1120 3012 4038 2271
(10.7) (28.8) (38.7) (21.8)
222 595 823 921
(8.7) (23.2) (32.1) (36.0)
1281 3439 4587 2837
8432 597 578 834
(80.8) (5.7) (5.5) (8.0)
1952 239 136 234
(76.2) (9.3) (5.3) (9.1)
53 10,015 373
(0.5) (95.9) (3.6)
108 2211 242
(4.2) (86.3) (9.4)
1662 2135 6607 37
(15.9) (20.4) (63.3) (0.4)
a
a
521 1651
(20.3) (64.5)
a
a
LOS N 90th %
N
(%)
N
(%)
12,144
(93.4)
858
(6.6)
a
a
a
a
a
a
a
(0.6) (1.1) (2.4) (8.4) (13.7) (19.2) (24.6) (29.9)
12 18 35 108 129 178 170 202
(1.4) (2.1) (4.1) (12.6) (15.0) (20.7) (19.8) (23.5)
(10.5) (28.3) (37.8) (23.4)
61 168 274 355
(7.1) (19.6) (31.9) (41.4)
9725 751 674 994
(80.1) (6.2) (5.6) (8.2)
659 85 40 74
(76.8) (9.9) (4.7) (8.6)
97 11,506 541
(0.8) (94.7) (4.5)
64 720 74
(7.5) (83.9) (8.6)
1934 2492 7679 39
(15.9) (20.5) (63.2) (0.3)
a
a
164 579
(19.1) (67.5)
a
a
9653 666 1825
(79.5) (5.5) (15.0)
601 99 158
(70.0) (11.5) (18.4)
11,166 978
(91.9) (8.1)
789 69
(92.0) (8.0)
12,009 91 10 34
(98.9) (0.7) (0.1) (0.3)
813 39
(94.8) (4.5)
a
a
a
a
11,911 233
(98.1) (1.9)
824 34
(96.0) (4.0)
12,113 31
(99.7) (0.3)
845 13
(98.5) (1.5)
11,966 178
(98.5) (1.5)
821 37
(95.7) (4.3)
12,102 42
(99.7) (0.3)
a
a
a
a
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
0.57
0.003
b0.001 8386 507 1548
(80.3) (4.9) (14.8)
1868 258 435
(72.9) (10.1) (17.0)
9598 843
(91.9) (8.1)
2357 204
(92.0) (8.0)
10,345 59
(99.1) (0.6)
2477 71
(96.7) (2.8)
a
a
a
a
a
a
a
a
b0.001
0.86
0.99
b0.001
b0.001
b0.001
b0.001
10,262 179
(98.3) (1.7)
2473 88
(96.6) (3.4)
10,416 25
(99.8) (0.2)
2542 19
(99.3) (0.7)
10,313 128
(98.8) (1.2)
2474 87
(96.6) (3.4)
10,407 34
(99.7) (0.3)
2547 14
(99.5) (0.5)
a
a
a
a
a
a
a
a
a
a
a
a
989 1401 126
(38.6) (54.7) (4.9)
(2.5) (49.3) (46.3)
a
(50.4) (45.3) (1.7)
298 5990 5618
a
5262 4734 177
a
a
261 517 70
(30.4) (60.3) (8.2)
152
(117–199)
190
(139–253)
156
(119–206)
198
(143–263)
10,274 167
(98.4) (1.6)
2486 75
(97.1) (2.9)
11,950 194
(98.4) (1.6)
810 48
(94.4) (5.6)
10,426 15
(99.9) (0.1)
2539 22
(99.1) (0.9)
12,127 17
(99.9) (0.1)
838 20
(97.7) (2.3)
b0.001
b0.001
b0.001
b0.001
0.10
0.10
b0.001
b0.001
b0.001
b0.001
b0.001 b0.001
b0.001
b0.001
P-value
b0.001
b0.001
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
5
Table 2 (continued)
No Yes Progressive renal insufficiency No Yes UTI No Yes CVA/stroke with deficit No Yes DVT/thrombophlebitis No Yes Transfusion No Yes Sepsis No Yes Shock No Yes Clavian complication No Yes Any complication No Yes Readmission No Yes Reoperation No Yes
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
10,419 22
(99.8) (0.2)
2539 22
(99.1) (0.9)
a
a
a
a
a
a
10,264 177
(98.3) (1.7)
2475 86
(96.6) (3.4)
a
a
a
a
a
a
a
a
LOS ≤ 90th % P-value
LOS N 90th %
N
(%)
N
(%)
12,114 30
(99.8) (0.2)
844 14
(98.4) (1.6)
a
a
a
a
a
a
a
a
a
a
11,921 223
(98.2) (1.8)
818 40
(95.3) (4.7)
a
a
a
a
a
a
a
a
12,106 38
(99.7) (0.3)
842 16
(98.1) (1.9)
12,026 118
(99.0) (1.0)
761 97
(88.7) (11.3)
12,090 54
(99.6) (0.4)
844 14
(98.4) (1.6)
a
a
a
a
848 10
(98.8) (1.2)
12,036 108
(99.1) (0.9)
797 61
(92.9) (7.1)
11,789 355
(97.1) (2.9)
743 115
(86.6) (13.4)
11,738 406
(96.7) (3.3)
790 68
(92.1) (7.9)
12,049 95
(99.2) (0.8)
822 36
(95.8) (4.2)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 10,414 27
(99.7) (0.3)
2534 27
(98.9) (1.1)
10,385 56
(99.5) (0.5)
2402 159
(93.8) (6.2)
10,400 41
(99.6) (0.4)
2534 27
(98.9) (1.1)
a
a
a
a
2550 11
(99.6) (0.4)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 10,360 81
(99.2) (0.8)
2473 88
(96.6) (3.4)
10,166 275
(97.4) (2.6)
2366 195
(92.4) (7.6)
10,115 326
(96.9) (3.1)
2413 148
(94.2) (5.8)
10,358 83
(99.2) (0.8)
2513 48
(98.1) (1.9)
P-value
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
Three patients with missing operation time and were not included. IQR: interquartile range. Patients with sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h and unplanned intubation were considered to have a Clavian complication. Patients with pneumonia, ARF, UTI, CVA/stroke with deficit, coma, sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h, unplanned intubation, or DVT/thrombophlebitis were considered to have any complication. a Suppressed from reporting because some cell sizes b10.
The primary outcome was prolonged LOS. Patients were stratified based on the type of hysterectomy. For each stratum of hysterectomy patients, we used two cutoff points to define prolonged LOS, ≥75th percentile and ≥ 90th percentile. Categorical variables were reported as frequencies and compared using Chi-square tests. Total operation time was reported as median and interquartile range (IQR) and compared using Wilcoxon rank-sum tests. Multivariable logistic regression models were fitted to determine the association of demographic characteristics, preoperative, intraoperative, and postoperative conditions with prolonged LOS, and reported as adjusted odds ratios (aOR) and 95% confidence intervals. To examine the contribution of each group of variables (demographics, preoperative, intraoperative, and postoperative conditions) to prolonged LOS, we fitted a number of models and compared the model fit statistics. The C-statistic is equal to the area under the Receiver Operating Characteristic (ROC) curve, and measures the overall ability of a model to correctly predict prolonged LOS. The null model contains no variables and has a C-statistic of 0.5, indicating that the model is no better than random chance in predicting the outcome, whereas a Cstatistic of 1 suggests that the model is perfectly predictive of the outcome. For models including one group of variables, the increased ability to predict prolonged LOS was calculated as the relative increase of Cstatistic compared to the null model. For models omitting one group of variables, the reduction in the ability to predict prolonged LOS was calculated as the decrease in C-statistic comparing to the full model over the difference in C-statistics of the full and null model. The pseudo-R2 indicates the total observed variability explained by a
model. The model with higher pseudo-R2 explains more observed variation in prolonged LOS. A series of sensitivity analyses were performed to examine the robustness of the findings of the relative importance of the association of each characteristics and length of stay. The Akaike information criterion (AIC) measures the relative goodness of fit of a model. In models including one group of variables, a lower AIC indicates a higher importance of that group, while in models omitting one group, a higher AIC indicates a greater importance of that omitted group. The likelihood ratio test (LRT) compares the goodness of fit of two models. We compared the model including one group of variables to the null model, and the model omitting one group to the full model. When one group of variables is included, a higher LRT indicates greater importance of that group, whereas in models omitting one group, a higher LRT indicates a greater importance of the omitted variables. All analyses were performed with SAS version 9.4 (SAS Institute Inc, Cary, North Carolina). All statistical tests were two-sided. A P-value of b0.05 was considered statistically significant. 3. Results A total of 19,084 women were identified, including 6082 (31.9%) women who underwent abdominal hysterectomy and 13,002 (68.1%) women who underwent minimally invasive hysterectomy. In the abdominal hysterectomy group, the 75th percentile for LOS was 5 days while the 90th percentile for LOS was 8 days (Table 1). In the minimally
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
6
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 3 Multivariable models for predictors of prolonged length of stay. Abdominal hysterectomy
Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes Bleeding disorder No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes increase in 1 min Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE No Yes UTI No Yes DVT/thrombophlebitis No Yes Sepsis No Yes Reoperation No Yes
Minimally invasive hysterectomy
LOS N 75th %
LOS N 90th %
LOS N 75th %
LOS N 90th %
aOR
aOR
aOR
aOR
1.11 (0.11–11.15) Referent 1.01 (0.48–2.11) 0.71 (0.35–1.45) 0.66 (0.33–1.35) 0.98 (0.50–1.93) 0.99 (0.50–1.93) 0.94 (0.48–1.84) 0.80 (0.41–1.56) 0.76 (0.39–1.47)
1.36 (0.09–20.46) Referent 1.99 (0.62–6.47) 0.89 (0.27–2.89) 1.34 (0.43–4.20) 1.78 (0.59–5.38) 1.69 (0.56–5.08) 1.49 (0.50–4.48) 1.39 (0.46–4.15) 1.29 (0.43–3.87)
–a Referent 1.60 (0.50–5.10) 0.88 (0.29–2.67) 0.41 (0.14–1.22) 0.39 (0.13–1.12) 0.28 (0.10–0.81)⁎ 0.26 (0.09–0.74)⁎ 0.21 (0.07–0.62)⁎ 0.22 (0.07–0.63)⁎
–a Referent 0.33 (0.08–1.27) 0.23 (0.06–0.86)⁎ 0.23 (0.07–0.80)⁎ 0.21 (0.06–0.70)⁎ 0.15 (0.05–0.51)⁎ 0.16 (0.05–0.52)⁎ 0.13 (0.04–0.42)⁎ 0.12 (0.04–0.39)⁎
Referent 1.46 (1.11–1.93)⁎ 1.59 (1.22–2.07)⁎ 2.70 (2.05–3.54)⁎
Referent 1.67 (1.10–2.54)⁎ 1.72 (1.15–2.58)⁎ 2.84 (1.89–4.26)⁎
Referent 1.02 (0.86–1.22) 1.03 (0.87–1.23) 2.07 (1.73–2.47)⁎
Referent 1.10 (0.80–1.50) 1.27 (0.94–1.71) 2.53 (1.87–3.42)⁎
Referent 1.42 (1.17–1.71)⁎ 1.13 (0.81–1.58) 0.93 (0.73–1.18)
Referent 1.30 (1.00–1.69)⁎ 1.42 (0.91–2.23) 1.14 (0.82–1.60)
Referent 1.55 (1.30–1.84)⁎ 1.20 (0.97–1.48) 1.35 (1.15–1.59)⁎
Referent 1.40 (1.08–1.81)⁎ 1.11 (0.78–1.58) 1.21 (0.93–1.58)
Referent 0.88 (0.70–1.10) 0.69 (0.57–0.85)⁎ 2.95 (1.26–6.92)⁎
Referent 0.80 (0.59–1.10) 0.61 (0.47–0.81)⁎ 5.29 (2.11–13.27)⁎
Referent 0.98 (0.84–1.15) 0.91 (0.79–1.05) 1.00 (0.46–2.19)
Referent 1.07 (0.82–1.39) 1.04 (0.82–1.31) 3.46 (1.45–8.29)⁎
Referent 1.30 (1.00–1.69)⁎ 1.13 (0.94–1.35)
Referent 1.17 (0.82–1.67) 0.99 (0.76–1.28)
Referent 2.15 (1.81–2.57)⁎ 1.12 (0.98–1.27)
Referent 1.87 (1.46–2.41)⁎ 1.10 (0.90–1.35)
Referent 2.67 (1.82–3.91)⁎ 1.31 (0.52–3.34) 0.48 (0.11–2.17)
Referent 2.41 (1.55–3.73)⁎ 0.65 (0.18–2.30) –a
Referent 3.57 (2.45–5.20)⁎ 2.12 (0.70–6.44) 0.98 (0.41–2.33)
Referent 4.05 (2.66–6.17)⁎ 5.53 (1.70–18.03)⁎ 0.42 (0.06–3.16)
Referent 1.15 (0.80–1.67)
Referent 0.99 (0.61–1.61)
Referent 1.57 (1.18–2.08)⁎
Referent 1.31 (0.88–1.95)
Referent 2.31 (1.61–3.32)⁎
Referent 2.42 (1.57–3.74)⁎
Referent 2.22 (1.64–3.01)⁎
Referent 1.84 (1.23–2.77)⁎
–a Referent 2.22 (0.95–5.16) 3.11 (1.34–7.25)⁎
8.36 (3.45–20.26)⁎
–a Referent 1.98 (0.48–8.22) 3.88 (0.94–16.03) 8.75 (2.05–37.39)⁎
3.51 (0.61–20.13) Referent 0.96 (0.68–1.36) 1.30 (0.92–1.85) 2.82 (1.84–4.32)⁎
7.64 (0.79–73.67) Referent 1.08 (0.54–2.16) 1.80 (0.90–3.61) 5.69 (2.69–12.03)⁎
1.007 (1.006–1.008)⁎
1.006 (1.005–1.007)⁎
1.007 (1.006–1.007)⁎
1.007 (1.006–1.007)⁎
Referent 1.67 (1.32–2.10)⁎
Referent 1.74 (1.30–2.33)⁎
Referent 1.28 (0.92–1.78)
Referent 2.35 (1.57–3.53)⁎
Referent 7.44 (4.55–12.16)⁎
Referent 5.15 (3.19–8.30)⁎
Referent 3.93 (1.91–8.08)⁎
Referent 10.47 (5.03–21.79)⁎
Referent 2.49 (1.51–4.10)⁎
Referent 3.28 (1.92–5.61)⁎
Referent 2.87 (1.48–5.59)⁎
Referent 4.06 (1.91–8.62)⁎
Referent 2.50 (1.77–3.52)⁎
Referent 2.33 (1.56–3.49)⁎
Referent 1.61 (1.20–2.14)⁎
Referent 1.95 (1.32–2.88)⁎
Referent 4.65 (2.70–8.01)⁎
Referent 4.96 (2.86–8.60)⁎
Referent 2.03 (1.10–3.74)⁎
Referent 2.07 (1.01–4.26)⁎
Referent 2.50 (1.58–3.95)⁎
Referent 3.66 (2.27–5.90)⁎
Referent 1.30 (0.72–2.33)
Referent 0.83 (0.39–1.75)
Referent 2.11 (1.42–3.16)⁎
Referent 3.28 (2.14–5.04)⁎
Referent 1.70 (1.13–2.54)⁎
Referent 3.61 (2.29–5.70)⁎
Logistic regression models included year of operation, age, race, preoperative BMI, diabetes, functional status, COPD, bleeding disorder, ASA class, total operation time, postoperative wound infection, pneumonia, PE, UTI, DVT/thrombophlebitis, sepsis, and reoperation. In the minimally invasive hysterectomy group, three patients with missing operation time were not included in the model. aOR: adjusted odds ratio. a Unestimable. ⁎ P-value b 0.05.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
invasive group, the 75th percentile for LOS was 1 day while the 90th percentile was 2 days (Table 2). Among women who underwent abdominal hysterectomy, older age, black race, insulin-dependent diabetes mellitus, partially dependent functional status, an underlying bleeding disorder and a higher ASA classification were all associated with a LOS N75th percentile (P b 0.05 for all) (Table 3). Likewise, a longer operative time and the occurrence of any of the postoperative complications were associated with a LOS N75th percentile (P b 0.05 for all). Similar patterns were identified for LOS N90th percentile for women who underwent abdominal hysterectomy. When minimally invasive hysterectomy was examined, advanced age, non-white race, comorbidities including insulin-dependent diabetes mellitus, non-independent functional status, COPD, bleeding disorders and a high ASA classification were associated with a LOS N75th percentile (P b 0.05 for all) (Table 3). A longer operative time and occurrence of the many of the postoperative complications were associated with a LOS of N75th percentile (P b 0.05 for all). In contrast, more recent year of surgery was associated with a lower likelihood of a prolonged hospitalization. The trends were relatively similar when LOS N90th percentile was examined. For abdominal hysterectomy, the examined factors were able to explain 23.6% of the variation in LOS N75th percentile (Pseudo-R2) (Table 4). Models including only one group of variables demonstrated a pseudo-R2 of 4.0% for demographic, 7.0% for preoperative, 7.9% for intraoperative, and 9.7% for postoperative characteristics (C-statistic: 0.61, 0.63, 0.65, 0.62, respectively). When LOS N90th percentile was examined, the full model explained 23.3% of the variation in LOS. Models including only one group of variables demonstrated a pseudo-R2 of 2.9% for demographic, 7.4% for preoperative, 4.8% for intraoperative, and 12.8% for postoperative characteristics (C-statistic: 0.62, 0.66, 0.63, 0.67, respectively). Table 4 Significance of inclusion or omission of 4 groups of clinical variables from a multivariable logistic regression model for prolonged length of stay in uterine cancer patients undergoing abdominal hysterectomy. C-statistic Pseudo-R2 Length of stay N 75th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions Length of stay N 90th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions
0.61 0.63 0.65 0.62
4.0% 7.0% 7.9% 9.7%
0.75 0.74 0.72 0.73
21.4% 19.4% 17.2% 17.4%
0.77
23.6%
0.62 0.66 0.63 0.67
2.9% 7.4% 4.8% 12.8%
0.78 0.76 0.77 0.74
21.6% 18.6% 19.9% 14.0%
0.79
23.3%
Logistic regression models were fitted and compared. Demographic factors included year, age, and race. Preoperative conditions included BMI, diabetes, functional status, COPD, bleeding disorders, and ASA class. Intraoperative conditions included total operation time. Postoperative complications included wound infection, pneumonia, PE, UTI, DVT, sepsis, and reoperation.
7
For minimally invasive hysterectomy, the examined factors were able to explain 16.2% of the variation in LOS N75th percentile (PseudoR2) (Table 5). Models including only one group of variables demonstrated a pseudo-R2 of 6.2% for demographic, 4.1% for preoperative, 6.9% for intraoperative, and 1.3% for postoperative characteristics (Cstatistic: 0.63, 0.60, 0.64, 0.53, respectively). When LOS N90th percentile was examined, the full model explained 16.2% of the variation in LOS. Models including only one group of variables demonstrated a pseudoR2 of 4.3% for demographic, 5.5% for preoperative, 5.1% for intraoperative, and 3.6% for postoperative characteristics (C-statistic: 0.64, 0.65, 0.64, 0.56, respectively). For both abdominal and minimally invasive hysterectomy, a series of sensitivity analyses indicated that the perioperative factors examined had similar importance to prolonged LOS (supplemental tables). 4. Discussion We found substantial variation in the LOS of women who underwent hysterectomy for uterine cancer. While postoperative complications explain a portion of the risk, non-modifiable demographic and clinical factors also account for a significant portion of the explained variation in prolonged LOS. Measurable demographic, preoperative, intraoperative, and postoperative factors explain a smaller percentage of the variation in LOS for women undergoing minimally invasive procedures. Overall, both procedures had a significant portion of variation in LOS that could not be accounted for by measurable patient, hospital or clinical factors. To date, studies analyzing the factors that drive LOS for patients undergoing hysterectomy have been limited. A few small, single-center studies have analyzed LOS and overall gynecologic cancer hospital
Table 5 Significance of inclusion or omission of 4 groups of clinical variables from a multivariable logistic regression model for prolonged length of stay in uterine cancer patients undergoing minimally invasive hysterectomy. C-statistic Pseudo-R2 Length of stay N 75th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions Length of stay N 90th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions
0.63 0.60 0.64 0.53
6.2% 4.1% 6.9% 1.3%
0.69 0.70 0.68 0.72
11.9% 12.8% 10.6% 15.5%
0.72
16.2%
0.64 0.65 0.64 0.56
4.3% 5.5% 5.1% 3.6%
0.74 0.72 0.73 0.75
13.5% 11.6% 12.1% 13.6%
0.76
16.2%
Three patients missing operation time were not included. Logistic regression models were fitted and compared. Demographic factors included year, age, and race. Preoperative conditions included BMI, diabetes, functional status, COPD, bleeding disorders, and ASA class. Intraoperative conditions included total operation time. Postoperative complications included wound infection, pneumonia, PE, UTI, DVT, sepsis, and reoperation.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
8
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
admissions and have found that older age, multiple co-morbidities, lower preoperative albumin and hemoglobin levels, multiple postoperative complications, a higher ASA class, and prior chemotherapy were predictive of extended LOS [11,12]. A 2010 prospective cohort study of 157 patients at a tertiary center found that malnutrition, low quality of life scores, and a diagnosis of advanced stage ovarian cancer were associated with prolonged LOS [13]. Examination of LOS for specific gynecologic cancers demonstrates similar risk factors. Among epithelial ovarian cancer patients, older age, higher ASA score, worse Eastern Cooperative Oncology Group (ECOG) status, and higher CA-125 levels were predictive of non-home discharges at one center [14]. For patients undergoing robotic surgery for endometrial cancer, factors such as older age, prior myocardial infarction, and the need for perioperative blood transfusions were associated with a longer hospital stay [15]. Our data is consistent with the above studies showing that factors from each stage of the perioperative period influence LOS for patients undergoing both abdominal and minimally invasive hysterectomy. However, if LOS is to be utilized as a quality metric, hospitals and physicians must have the ability to modify LOS. Examination of risk factors for partial nephrectomy, colectomy, and trauma patients have shown that non-modifiable patient traits often play an important role in determining prolonged LOS [16–18]. For patients undergoing hysterectomy, our data suggest first, that a large portion of the variation in LOS is not explained by measurable clinical and demographic characteristics. There are likely a number of factors that contribute to this unexplained variation including socioeconomic factors, a patient's functional status, a patient's social support network as well as hospital characteristics and practice patterns. Second, much of the variation that can be explained is due to underlying demographic characteristics that are not modifiable. These factors raise concern about how actionable LOS is as a quality metric. While understanding drivers of LOS is critical for hospital reimbursement, LOS also contributes to patient satisfaction. Studies in other surgical fields have found that patient satisfaction is higher in those who undergo procedures with shorter LOS [19,20]. Recently, a large national, aggregate study demonstrated that shorter LOS after six of the most common major surgeries, not including hysterectomy, was associated with higher patient satisfaction scores on the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey [21]. HCAHPS scores can have a large impact on public perception as well as hospital reimbursement [22,23]. One major concern with the widespread efforts to reduce hospital LOS is that successful reduction of LOS may contribute to increasing readmission rates, another CMS quality metric [24–27]. However, there is no strong evidence to suggest that a negative relationship exists between these metrics. An analysis of the Veterans Affairs hospital system over a 14-year period showed reduced LOS for medical diagnoses were not associated with an increase in 30-day readmission rates, but rather in a reduction in the readmission rate as well [28]. Previous work done by our group revealed that 30-day readmission for patients undergoing hysterectomy for benign indications and uterine cancer was driven mainly by postoperative complications [29]. The implications of the previous work and the current study are that improving LOS through reduction in modifiable factors, such as postoperative complications, would decrease both LOS and 30-day readmission rates for patients undergoing hysterectomy. We acknowledge a number of important limitations. Although an effort was made to capture a wide array of hospitals, NSQIP does not include data from every hospital in the U.S. and therefore, the conclusions may not be generalizable. We were also unable to distinguish between laparoscopic and robotic hysterectomy from the database, which are grouped together as minimally invasive hysterectomy. These procedures have unique risk profiles and may affect LOS differently. NSQIP lacks data on hospital characteristics which would be of great interest in the analysis. NSQIP does not capture some
complications that are specific to hysterectomy; thus, our analysis is based on commonly reported perioperative complications. In conclusion, only a portion of the variation in LOS after hysterectomy for endometrial cancer is explained by measurable clinical and demographic factors. While some of the explained variation is accounted for by modifiable factors (intraoperative and postoperative), nonmodifiable factors (demographic and preoperative) also contribute to prolonged LOS after hysterectomy. In the context of continuously evolving reimbursement systems and the desire for improved patient satisfaction, specifically measuring and addressing modifiable risk factors should be a focus.
Disclosure Dr. Wright has served as a consultant for Tesaro and Clovis Oncology. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.ygyno.2018.03.048. References [1] D. Blumenthal, K. Davis, S. Guterman, Medicare at 50—origins and evolution, N. Engl. J. Med. 372 (2015) 479–486. [2] D. Blumenthal, K. Davis, S. Guterman, Medicare at 50—moving forward, N. Engl. J. Med. 372 (2015) 671–677. [3] M.K. Whiteman, S.D. Hillis, D.J. Jamieson, et al., Inpatient hysterectomy surveillance in the United States, 2000–2004, Am. J. Obstet. Gynecol. 198 (34) (2008) e1–7. [4] C.M. Farquhar, C.A. Steiner, Hysterectomy rates in the United States 1990–1997, Obstet. Gynecol. 99 (2002) 229–234. [5] S.R. Kovac, Hysterectomy outcomes in patients with similar indications, Obstet. Gynecol. 95 (2000) 787–793. [6] E.S. Campbell, H. Xiao, M.K. Smith, Types of hysterectomy. Comparison of characteristics, hospital costs, utilization and outcomes, J. Reprod. Med. 48 (2003) 943–949. [7] R. Garry, J. Fountain, S. Mason, et al., The eVALuate study: two parallel randomised trials, one comparing laparoscopic with abdominal hysterectomy, the other comparing laparoscopic with vaginal hysterectomy, BMJ 328 (2004) 129. [8] J.M. Wu, M.E. Wechter, E.J. Geller, T.V. Nguyen, A.G. Visco, Hysterectomy rates in the United States, 2003, Obstet. Gynecol. 110 (2007) 1091–1095. [9] J.F. Boggess, P.A. Gehrig, L. Cantrell, et al., A comparative study of 3 surgical methods for hysterectomy with staging for endometrial cancer: robotic assistance, laparoscopy, laparotomy, Am. J. Obstet. Gynecol. 199 (360) (2008) e1–9. [10] ACS NSQIP participant use data file, Program ACoSNSQI, 2015. [11] L.S. Massad, G. Vogler, T.J. Herzog, D.G. Mutch, Correlates of length of stay in gynecologic oncology patients undergoing inpatient surgery, Gynecol. Oncol. 51 (1993) 214–218. [12] M.M. Dean, M.A. Finan, R.C. Kline, Predictors of complications and hospital stay in gynecologic cancer surgery, Obstet. Gynecol. 97 (2001) 721–724. [13] B. Laky, M. Janda, S. Kondalsamy-Chennakesavan, G. Cleghorn, A. Obermair, Pretreatment malnutrition and quality of life - association with prolonged length of hospital stay among patients with gynecological cancer: a cohort study, BMC Cancer 10 (2010) 232. [14] M.M. AlHilli, C.W. Tran, C.L. Langstraat, et al., Risk-scoring model for prediction of non-home discharge in epithelial ovarian cancer patients, J. Am. Coll. Surg. 217 (2013) 507–515. [15] M.I. Liang, M.A. Rosen, K.S. Rath, et al., Predicting inpatient stay lasting 2 midnights or longer after robotic surgery for endometrial cancer, J. Minim. Invasive Gynecol. 22 (2015) 583–589. [16] J.A. Larson, J.H. Kaouk, M.D. Stifelman, et al., Nonmodifiable factors and complications contribute to length of stay in robot-assisted partial nephrectomy, J. Endourol. 29 (2015) 422–429. [17] K.G. Cologne, S. Byers, D.R. Rosen, et al., Factors associated with a short (b2 days) or long (N10 days) length of stay after colectomy: a multivariate analysis of over 400 patients, Am. Surg. 82 (2016) 960–963. [18] K.J. Brasel, H.J. Lim, R. Nirula, J.A. Weigelt, Length of stay: an appropriate quality measure? Arch. Surg. 142 (2007) 461–465 (discussion 5–6). [19] L.H. Cohn, D.H. Adams, G.S. Couper, et al., Minimally invasive cardiac valve surgery improves patient satisfaction while reducing costs of cardiac valve replacement and repair, Ann. Surg. 226 (1997) 421–426 (discussion 7–8). [20] H. Husted, G. Holm, S. Jacobsen, Predictors of length of stay and patient satisfaction after hip and knee replacement surgery: fast-track experience in 712 patients, Acta Orthop. 79 (2008) 168–173. [21] T.C. Tsai, E.J. Orav, A.K. Jha, Patient satisfaction and quality of surgical care in US hospitals, Ann. Surg. 261 (2015) 2–8. [22] L.A. Giordano, M.N. Elliott, E. Goldstein, W.G. Lehrman, P.A. Spencer, Development, implementation, and public reporting of the HCAHPS survey, Med. Care Res. Rev. 67 (2010) 27–37.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx [23] Hospital Value-Based Purchasing Program: Centers for Medicare and Medicaid Services (CMS)2015. [24] A. Clarke, R. Rosen, Length of stay. How short should hospital care be? Eur. J. Pub. Health 11 (2001) 166–170. [25] P.A. Cowper, E.R. DeLong, E.L. Hannan, et al., Is early too early? Effect of shorter stays after bypass surgery, Ann. Thorac. Surg. 83 (2007) 100–107. [26] D.W. Baker, D. Einstadter, S.S. Husak, R.D. Cebul, Trends in postdischarge mortality and readmissions: has length of stay declined too far? Arch. Intern. Med. 164 (2004) 538–544.
9
[27] S. Hendren, A.M. Morris, W. Zhang, J. Dimick, Early discharge and hospital readmission after colectomy for cancer, Dis. Colon Rectum 54 (2011) 1362–1367. [28] P.J. Kaboli, J.T. Go, J. Hockenberry, et al., Associations between reduced hospital length of stay and 30-day readmission rate and mortality: 14-year experience in 129 veterans affairs hospitals, Ann. Intern. Med. 157 (2012) 837–845. [29] K. Dessources, J.Y. Hou, A.I. Tergas, et al., Factors associated with 30-day hospital readmission after hysterectomy, Obstet. Gynecol. 125 (2015) 461–470.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
Contents lists available at ScienceDirect
Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno
Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine cancer☆ Surbhi Agrawal a, Ling Chen a, Ana I. Tergas a,c,d,e, June Y. Hou a,d,e, Caryn M. St. Clair a,d,e, Cande V. Ananth a,c, Dawn L. Hershman a,b,d,e, Jason D. Wright a,d,e,⁎ a
Department of Obstetrics and Gynecology, Columbia University College of Physicians and Surgeons, USA Department of Medicine, Columbia University College of Physicians and Surgeons, USA Department of Epidemiology, Joseph L. Mailman School of Public Health, Columbia University, USA d Herbert Irving Comprehensive Cancer Center, Columbia University College of Physicians and Surgeons, USA e New York Presbyterian Hospital, USA b c
H I G H L I G H T S • Perioperative risk factors account for approximately 25% of the variation in prolonged LOS. • A substantial proportion of the variation in LOS remains unexplained by measurable factors. • There are similar predictors of prolonged LOS for abdominal and minimally invasive hysterectomy.
a r t i c l e
i n f o
Article history: Received 1 March 2018 Received in revised form 12 March 2018 Accepted 12 March 2018 Available online xxxx
a b s t r a c t Objective. We examined the influence of modifiable (intraoperative factors and complications) and nonmodifiable (clinical and demographic characteristics) factors on length of stay (LOS) for women who underwent hysterectomy for uterine cancer. Methods. The National Surgical Quality Improvement Program database was used to identify women who underwent hysterectomy for uterine cancer from 2006 to 2015. The association between demographic, preoperative, intraoperative, and postoperative factors and LOS was examined. The primary outcome was prolonged LOS (N75th an3 N 90th percentiles). Model fit statistics were used to assess the importance of each group of characteristics. Results. Of 19,084 women identified, 6082 (31.9%) underwent abdominal and 13,002 (68.1%) underwent minimally invasive hysterectomy. In the abdominal hysterectomy group, the 75th and 90th percentiles for LOS were 5 and 8 days, respectively. All risk factors combined accounted for 23.6% of the variation in LOS N75th percentile. Demographic characteristics explained 4.0%, preoperative factors 7.0%, intraoperative factors 7.9%, and postoperative characteristics 9.7% of variation in prolonged LOS. In the minimally invasive group, the 75th and 90th percentiles for LOS were 1 and 2 days, respectively. The combined risk factors explained 16.2% of the variation in prolonged LOS. Demographic characteristics accounted for 6.2%, preoperative factors 4.1%, intraoperative factors 6.9%, and postoperative characteristics 1.3% of variation in prolonged LOS. Similar patterns were seen when prolonged LOS was defined as N90th percentile. Conclusion. Perioperative risk factors account for approximately one quarter of the variation in prolonged LOS. Overall, a substantial proportion of the variation in LOS remains unexplained by measurable patient and hospital factors which may limit the utility of LOS as a quality metric for endometrial cancer. © 2018 Published by Elsevier Inc.
☆ Dr. Wright (NCI R01CA169121-01A1) and Dr. Hershman (NCI R01 CA166084) are recipients of grants from the National Cancer Institute. ⁎ Corresponding author at: Division of Gynecologic Oncology, Columbia University College of Physicians and Surgeons, 161 Fort Washington Ave, 8th Floor, New York, NY 10032, USA. E-mail address: [email protected] (J.D. Wright).
1. Introduction Hospital length of stay (LOS) has been proposed as a quality metric and influences reimbursement for a variety of surgical procedures. In the context of growing healthcare costs, the Centers for Medicare and
https://doi.org/10.1016/j.ygyno.2018.03.048 0090-8258/© 2018 Published by Elsevier Inc.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
2
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 1 Demographic, preoperative, intraoperative and postoperative conditions of patients who underwent abdominal hysterectomy by prolonged length of stay.
All Demographics Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Elective surgery No Yes Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Smoking No Yes Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes CHF No Yes Bleeding disorder No Yes Open wound No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes Median (IQR) Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
LOS ≤ 90th %
LOS N 90th %
N
(%)
N
(%)
4834
(79.5)
1248
(20.5)
5545
(91.2)
537
(8.8)
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
177 324 355 537 613 838 928 1004
(3.7) (6.7) (7.3) (11.1) (12.7) (17.3) (19.2) (20.8)
45 62 72 160 187 228 235 242
(3.6) (5.0) (5.8) (12.8) (15.0) (18.3) (18.8) (19.4)
201 366 392 623 717 973 1063 1141
(3.6) (6.6) (7.1) (11.2) (12.9) (17.5) (19.2) (20.6)
21 20 35 74 83 93 100 105
(3.9) (3.7) (6.5) (13.8) (15.5) (17.3) (18.6) (19.6)
607 1280 1742 1205
(12.6) (26.5) (36.0) (24.9)
97 264 412 475
(7.8) (21.2) (33.0) (38.1)
666 1431 1985 1463
(12.0) (25.8) (35.8) (26.4)
38 113 169 217
(7.1) (21.0) (31.5) (40.4)
3417 604 236 577
(70.7) (12.5) (4.9) (11.9)
843 240 59 106
(67.5) (19.2) (4.7) (8.5)
3905 741 267 632
(70.4) (13.4) (4.8) (11.4)
355 103 28 51
(66.1) (19.2) (5.2) (9.5)
134 3776 924
(2.8) (78.1) (19.1)
221 824 203
(17.7) (66.0) (16.3)
211 4294 1040
(3.8) (77.4) (18.8)
144 306 87
(26.8) (57.0) (16.2)
755 966 3098 15
(15.6) (20.0) (64.1) (0.3)
215 256 763 14
(17.2) (20.5) (61.1) (1.1)
870 1113 3544 18
(15.7) (20.1) (63.9) (0.3)
100 109 317 11
(18.6) (20.3) (59.0) (2.0)
3733 280 821
(77.2) (5.8) (17.0)
886 114 248
(71.0) (9.1) (19.9)
4234 344 967
(76.4) (6.2) (17.4)
385 50 102
(71.7) (9.3) (19.0)
4373 461
(90.5) (9.5)
1121 127
(89.8) (10.2)
5006 539
(90.3) (9.7)
488 49
(90.9) (9.1)
4728 70 17 19
(97.8) (1.4) (0.4) (0.4)
1166 70
(93.4) (5.6)
(92.2) (7.1)
a
a
a
a
a
(97.4) (1.8) (0.4) (0.4)
495 38
a
5399 102 23 21
a
a
5393 152
(97.3) (2.7)
509 28
(94.8) (5.2)
5524 21
(99.6) (0.4)
525 12
(97.8) (2.2)
5420 125
(97.7) (2.3)
501 36
(93.3) (6.7)
5512 33
(99.4) (0.6)
525 12
(97.8) (2.2)
P-value
0.050
0.17
b0.001
b0.001
b0.001
0.002
b0.001
b0.001
b0.001
b0.001
b0.001
0.01
0.50
0.66
b0.001
b0.001
b0.001
0.001
4713 121
(97.5) (2.5)
1189 59
(95.3) (4.7)
4821 13
(99.7) (0.3)
1228 20
(98.4) (1.6)
4739 95
(98.0) (2.0)
1182 66
(94.7) (5.3)
4810 24
(99.5) (0.5)
1227 21
(98.3) (1.7)
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
a
1963 2606 139
(40.6) (53.9) (2.9)
336 781 125
(26.9) (62.6) (10.0)
2185 3032 198
(39.4) (54.7) (3.6)
114 355 66
(21.2) (66.1) (12.3)
138
(101–189)
185
(128–242)
143
(103–196)
186
(129–250)
4531 303
(93.7) (6.3)
1056 192
(84.6) (15.4)
5159 386
(93.0) (7.0)
428 109
(79.7) (20.3)
4807 27
(99.4) (0.6)
1177 71
(94.3) (5.7)
5490 55
(99.0) (1.0)
494 43
(92.0) (8.0)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 b0.001
b0.001
b0.001
P-value
b0.001
b0.001
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
3
Table 1 (continued)
No Yes Progressive renal insufficiency No Yes UTI No Yes CVA/stroke with deficit No Yes DVT/thrombophlebitis No Yes Transfusion No Yes Sepsis No Yes Shock No Yes Clavian complication No Yes Any complication No Yes Readmission No Yes Reoperation No Yes
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
4795 39
(99.2) (0.8)
1196 52
(95.8) (4.2)
4821 13
(99.7) (0.3)
1229 19
(98.5) (1.5)
4738 96
(98.0) (2.0)
1159 89
(92.9) (7.1)
a
a
a
a
1236 12
(99.0) (1.0)
LOS ≤ 90th % P-value
LOS N 90th %
N
(%)
N
(%)
5490 55
(99.0) (1.0)
501 36
(93.3) (6.7)
5524 21
(99.6) (0.4)
526 11
(98.0) (2.0)
5411 134
(97.6) (2.4)
486 51
(90.5) (9.5)
5535 10
(99.8) (0.2)
a
a
a
a
5503 42
(99.2) (0.8)
502 35
(93.5) (6.5)
4816 729
(86.9) (13.1)
318 219
(59.2) (40.8)
5485 60
(98.9) (1.1)
488 49
(90.9) (9.1)
5526 19
(99.7) (0.3)
513 24
(95.5) (4.5)
5381 164
(97.0) (3.0)
414 123
(77.1) (22.9)
5187 358
(93.5) (6.5)
341 196
(63.5) (36.5)
5115 430
(92.2) (7.8)
464 73
(86.4) (13.6)
5458 87
(98.4) (1.6)
481 56
(89.6) (10.4)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 4807 27
(99.4) (0.6)
1198 50
(96.0) (4.0)
4322 512
(89.4) (10.6)
812 436
(65.1) (34.9)
4789 45
(99.1) (0.9)
1184 64
(94.9) (5.1)
4820 14
(99.7) (0.3)
1219 29
(97.7) (2.3)
4723 111
(97.7) (2.3)
1072 176
(85.9) (14.1)
4592 242
(95.0) (5.0)
936 312
(75.0) (25.0)
4506 328
(93.2) (6.8)
1073 175
(86.0) (14.0)
4767 67
(98.6) (1.4)
1172 76
(93.9) (6.1)
P-value
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
IQR: interquartile range. Patients with sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h and unplanned intubation were considered to have a Clavian complication. Patients with pneumonia, ARF, UTI, CVA/stroke with deficit, coma, sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h, unplanned intubation, or DVT/thrombophlebitis were considered to have any complication. a Suppressed from reporting because some cell sizes b10.
Medicaid Services (CMS) have increasingly used LOS to encourage hospitals to become more efficient at providing care [1,2]. Hysterectomy is one of the most commonly performed surgical procedures in women [3]. To date, most studies have focused on comparing LOS and overall cost differences among the different surgical techniques for hysterectomy [4–8]. A single-institution study found that the average LOS was 4.4 days for abdominal, 1.2 days for laparoscopic, and 1.0 day for robotic hysterectomy for patients with uterine cancer [9]. However, aside from surgical route, relatively little is known about the perioperative factors that drive LOS and contribute to a prolonged LOS among hysterectomy patients. Given the increased focus on using LOS as a measure of quality and driver for reimbursement, understanding the factors that affect prolonged hospitalization is imperative to develop actionable strategies to reduce LOS. We therefore performed a population-based study to determine the relative contribution of modifiable (intraoperative events and complications) and non-modifiable (clinical and demographic factors) to prolonged LOS after hysterectomy for uterine cancer. 2. Methods We used the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Participant Use Data File (PUF) for the analysis. NSQIP is a nationally validated, outcomes-based program that collects data of patients undergoing major surgical procedures from participating hospitals [10]. Preoperative, intraoperative, and 30day postoperative variables from both the inpatient and outpatient
setting are collected to measure and improve surgical quality. Data are abstracted from medical charts by trained reviewers through a systematic sampling process. Participating hospitals are required to submit data from 42 of the 46 8-day cycles. The 8-day cycle is spaced throughout the year to minimize bias in case selection. Data quality is ensured by regular audit. Women who had uterine cancer and underwent abdominal or minimally invasive hysterectomy from 2006 to 2015 were included in this analysis. Patients who had other gynecologic malignancies or missing data for length of stay were excluded. Demographic characteristics included year of operation, age (b50, 50–59, 60–69, ≥70 years), race (white, black, other, unknown), and whether the surgery was elective (yes, no, unknown). For each woman, the following preoperative conditions were recorded: body mass index (BMI, normal b25 kg/m2, overweight 25–29.9 kg/m2, obese ≥30 kg/m2, and unknown), diabetes mellitus (insulin dependent, or non-insulin dependent), tobacco use (current smoker within one year), history of severe chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF) within 30 days of surgery, a bleeding disorder, an open wound with or without infection, functional health status prior to surgery (independent, partially dependent, totally dependent, and unknown), and Society of Anesthesiology (ASA) classification score (none, 1, 2, 3, 4–5, or unknown). Total operation time (in minutes) was noted for the procedure. Postoperative conditions included wound infection, pneumonia, pulmonary embolism (PE), urinary tract infection (UTI), deep vein thrombosis (DVT), sepsis, and reoperation within 30 days.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
4
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 2 Demographic, preoperative, intraoperative and postoperative conditions of patients who underwent minimally invasive hysterectomy by prolonged length of stay. LOS ≤ 75th %
All Demographics Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Elective surgery No Yes Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Smoking No Yes Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes CHF No Yes Bleeding disorder No Yes Open wound No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes Median (IQR) Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE
LOS N 75th %
LOS ≤ 90th %
N
(%)
N
(%)
10,441
(80.3)
2561
(19.7)
P-value
a
a
a
a
a
a
a
a
a
32 77 232 812 1414 2020 2644 3200
(0.3) (0.7) (2.2) (7.8) (13.5) (19.3) (25.3) (30.6)
51 71 98 311 383 489 510 639
(2.0) (2.8) (3.8) (12.1) (15.0) (19.1) (19.9) (25.0)
71 130 295 1015 1668 2331 2984 3637
1120 3012 4038 2271
(10.7) (28.8) (38.7) (21.8)
222 595 823 921
(8.7) (23.2) (32.1) (36.0)
1281 3439 4587 2837
8432 597 578 834
(80.8) (5.7) (5.5) (8.0)
1952 239 136 234
(76.2) (9.3) (5.3) (9.1)
53 10,015 373
(0.5) (95.9) (3.6)
108 2211 242
(4.2) (86.3) (9.4)
1662 2135 6607 37
(15.9) (20.4) (63.3) (0.4)
a
a
521 1651
(20.3) (64.5)
a
a
LOS N 90th %
N
(%)
N
(%)
12,144
(93.4)
858
(6.6)
a
a
a
a
a
a
a
(0.6) (1.1) (2.4) (8.4) (13.7) (19.2) (24.6) (29.9)
12 18 35 108 129 178 170 202
(1.4) (2.1) (4.1) (12.6) (15.0) (20.7) (19.8) (23.5)
(10.5) (28.3) (37.8) (23.4)
61 168 274 355
(7.1) (19.6) (31.9) (41.4)
9725 751 674 994
(80.1) (6.2) (5.6) (8.2)
659 85 40 74
(76.8) (9.9) (4.7) (8.6)
97 11,506 541
(0.8) (94.7) (4.5)
64 720 74
(7.5) (83.9) (8.6)
1934 2492 7679 39
(15.9) (20.5) (63.2) (0.3)
a
a
164 579
(19.1) (67.5)
a
a
9653 666 1825
(79.5) (5.5) (15.0)
601 99 158
(70.0) (11.5) (18.4)
11,166 978
(91.9) (8.1)
789 69
(92.0) (8.0)
12,009 91 10 34
(98.9) (0.7) (0.1) (0.3)
813 39
(94.8) (4.5)
a
a
a
a
11,911 233
(98.1) (1.9)
824 34
(96.0) (4.0)
12,113 31
(99.7) (0.3)
845 13
(98.5) (1.5)
11,966 178
(98.5) (1.5)
821 37
(95.7) (4.3)
12,102 42
(99.7) (0.3)
a
a
a
a
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
0.57
0.003
b0.001 8386 507 1548
(80.3) (4.9) (14.8)
1868 258 435
(72.9) (10.1) (17.0)
9598 843
(91.9) (8.1)
2357 204
(92.0) (8.0)
10,345 59
(99.1) (0.6)
2477 71
(96.7) (2.8)
a
a
a
a
a
a
a
a
b0.001
0.86
0.99
b0.001
b0.001
b0.001
b0.001
10,262 179
(98.3) (1.7)
2473 88
(96.6) (3.4)
10,416 25
(99.8) (0.2)
2542 19
(99.3) (0.7)
10,313 128
(98.8) (1.2)
2474 87
(96.6) (3.4)
10,407 34
(99.7) (0.3)
2547 14
(99.5) (0.5)
a
a
a
a
a
a
a
a
a
a
a
a
989 1401 126
(38.6) (54.7) (4.9)
(2.5) (49.3) (46.3)
a
(50.4) (45.3) (1.7)
298 5990 5618
a
5262 4734 177
a
a
261 517 70
(30.4) (60.3) (8.2)
152
(117–199)
190
(139–253)
156
(119–206)
198
(143–263)
10,274 167
(98.4) (1.6)
2486 75
(97.1) (2.9)
11,950 194
(98.4) (1.6)
810 48
(94.4) (5.6)
10,426 15
(99.9) (0.1)
2539 22
(99.1) (0.9)
12,127 17
(99.9) (0.1)
838 20
(97.7) (2.3)
b0.001
b0.001
b0.001
b0.001
0.10
0.10
b0.001
b0.001
b0.001
b0.001
b0.001 b0.001
b0.001
b0.001
P-value
b0.001
b0.001
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
5
Table 2 (continued)
No Yes Progressive renal insufficiency No Yes UTI No Yes CVA/stroke with deficit No Yes DVT/thrombophlebitis No Yes Transfusion No Yes Sepsis No Yes Shock No Yes Clavian complication No Yes Any complication No Yes Readmission No Yes Reoperation No Yes
LOS ≤ 75th %
LOS N 75th %
N
(%)
N
(%)
10,419 22
(99.8) (0.2)
2539 22
(99.1) (0.9)
a
a
a
a
a
a
10,264 177
(98.3) (1.7)
2475 86
(96.6) (3.4)
a
a
a
a
a
a
a
a
LOS ≤ 90th % P-value
LOS N 90th %
N
(%)
N
(%)
12,114 30
(99.8) (0.2)
844 14
(98.4) (1.6)
a
a
a
a
a
a
a
a
a
a
11,921 223
(98.2) (1.8)
818 40
(95.3) (4.7)
a
a
a
a
a
a
a
a
12,106 38
(99.7) (0.3)
842 16
(98.1) (1.9)
12,026 118
(99.0) (1.0)
761 97
(88.7) (11.3)
12,090 54
(99.6) (0.4)
844 14
(98.4) (1.6)
a
a
a
a
848 10
(98.8) (1.2)
12,036 108
(99.1) (0.9)
797 61
(92.9) (7.1)
11,789 355
(97.1) (2.9)
743 115
(86.6) (13.4)
11,738 406
(96.7) (3.3)
790 68
(92.1) (7.9)
12,049 95
(99.2) (0.8)
822 36
(95.8) (4.2)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 10,414 27
(99.7) (0.3)
2534 27
(98.9) (1.1)
10,385 56
(99.5) (0.5)
2402 159
(93.8) (6.2)
10,400 41
(99.6) (0.4)
2534 27
(98.9) (1.1)
a
a
a
a
2550 11
(99.6) (0.4)
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001 10,360 81
(99.2) (0.8)
2473 88
(96.6) (3.4)
10,166 275
(97.4) (2.6)
2366 195
(92.4) (7.6)
10,115 326
(96.9) (3.1)
2413 148
(94.2) (5.8)
10,358 83
(99.2) (0.8)
2513 48
(98.1) (1.9)
P-value
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
b0.001
Three patients with missing operation time and were not included. IQR: interquartile range. Patients with sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h and unplanned intubation were considered to have a Clavian complication. Patients with pneumonia, ARF, UTI, CVA/stroke with deficit, coma, sepsis, shock, cardiac arrest, MI, PE, ventilation N48 h, unplanned intubation, or DVT/thrombophlebitis were considered to have any complication. a Suppressed from reporting because some cell sizes b10.
The primary outcome was prolonged LOS. Patients were stratified based on the type of hysterectomy. For each stratum of hysterectomy patients, we used two cutoff points to define prolonged LOS, ≥75th percentile and ≥ 90th percentile. Categorical variables were reported as frequencies and compared using Chi-square tests. Total operation time was reported as median and interquartile range (IQR) and compared using Wilcoxon rank-sum tests. Multivariable logistic regression models were fitted to determine the association of demographic characteristics, preoperative, intraoperative, and postoperative conditions with prolonged LOS, and reported as adjusted odds ratios (aOR) and 95% confidence intervals. To examine the contribution of each group of variables (demographics, preoperative, intraoperative, and postoperative conditions) to prolonged LOS, we fitted a number of models and compared the model fit statistics. The C-statistic is equal to the area under the Receiver Operating Characteristic (ROC) curve, and measures the overall ability of a model to correctly predict prolonged LOS. The null model contains no variables and has a C-statistic of 0.5, indicating that the model is no better than random chance in predicting the outcome, whereas a Cstatistic of 1 suggests that the model is perfectly predictive of the outcome. For models including one group of variables, the increased ability to predict prolonged LOS was calculated as the relative increase of Cstatistic compared to the null model. For models omitting one group of variables, the reduction in the ability to predict prolonged LOS was calculated as the decrease in C-statistic comparing to the full model over the difference in C-statistics of the full and null model. The pseudo-R2 indicates the total observed variability explained by a
model. The model with higher pseudo-R2 explains more observed variation in prolonged LOS. A series of sensitivity analyses were performed to examine the robustness of the findings of the relative importance of the association of each characteristics and length of stay. The Akaike information criterion (AIC) measures the relative goodness of fit of a model. In models including one group of variables, a lower AIC indicates a higher importance of that group, while in models omitting one group, a higher AIC indicates a greater importance of that omitted group. The likelihood ratio test (LRT) compares the goodness of fit of two models. We compared the model including one group of variables to the null model, and the model omitting one group to the full model. When one group of variables is included, a higher LRT indicates greater importance of that group, whereas in models omitting one group, a higher LRT indicates a greater importance of the omitted variables. All analyses were performed with SAS version 9.4 (SAS Institute Inc, Cary, North Carolina). All statistical tests were two-sided. A P-value of b0.05 was considered statistically significant. 3. Results A total of 19,084 women were identified, including 6082 (31.9%) women who underwent abdominal hysterectomy and 13,002 (68.1%) women who underwent minimally invasive hysterectomy. In the abdominal hysterectomy group, the 75th percentile for LOS was 5 days while the 90th percentile for LOS was 8 days (Table 1). In the minimally
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
6
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
Table 3 Multivariable models for predictors of prolonged length of stay. Abdominal hysterectomy
Year of operation 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Age b50 50–59 60–69 ≥70 Race White Black Other Unknown Preoperative conditions BMI Normal Overweight Obese Unknown Diabetes No Insulin dependent Type 2 Functional status Independent Partially dependent Totally dependent Unknown COPD No Yes Bleeding disorder No Yes ASA class None 1 2 3 4–5 Intraoperative conditions Total operation time in minutes increase in 1 min Postoperative conditions Any wound infection No Yes Pneumonia No Yes PE No Yes UTI No Yes DVT/thrombophlebitis No Yes Sepsis No Yes Reoperation No Yes
Minimally invasive hysterectomy
LOS N 75th %
LOS N 90th %
LOS N 75th %
LOS N 90th %
aOR
aOR
aOR
aOR
1.11 (0.11–11.15) Referent 1.01 (0.48–2.11) 0.71 (0.35–1.45) 0.66 (0.33–1.35) 0.98 (0.50–1.93) 0.99 (0.50–1.93) 0.94 (0.48–1.84) 0.80 (0.41–1.56) 0.76 (0.39–1.47)
1.36 (0.09–20.46) Referent 1.99 (0.62–6.47) 0.89 (0.27–2.89) 1.34 (0.43–4.20) 1.78 (0.59–5.38) 1.69 (0.56–5.08) 1.49 (0.50–4.48) 1.39 (0.46–4.15) 1.29 (0.43–3.87)
–a Referent 1.60 (0.50–5.10) 0.88 (0.29–2.67) 0.41 (0.14–1.22) 0.39 (0.13–1.12) 0.28 (0.10–0.81)⁎ 0.26 (0.09–0.74)⁎ 0.21 (0.07–0.62)⁎ 0.22 (0.07–0.63)⁎
–a Referent 0.33 (0.08–1.27) 0.23 (0.06–0.86)⁎ 0.23 (0.07–0.80)⁎ 0.21 (0.06–0.70)⁎ 0.15 (0.05–0.51)⁎ 0.16 (0.05–0.52)⁎ 0.13 (0.04–0.42)⁎ 0.12 (0.04–0.39)⁎
Referent 1.46 (1.11–1.93)⁎ 1.59 (1.22–2.07)⁎ 2.70 (2.05–3.54)⁎
Referent 1.67 (1.10–2.54)⁎ 1.72 (1.15–2.58)⁎ 2.84 (1.89–4.26)⁎
Referent 1.02 (0.86–1.22) 1.03 (0.87–1.23) 2.07 (1.73–2.47)⁎
Referent 1.10 (0.80–1.50) 1.27 (0.94–1.71) 2.53 (1.87–3.42)⁎
Referent 1.42 (1.17–1.71)⁎ 1.13 (0.81–1.58) 0.93 (0.73–1.18)
Referent 1.30 (1.00–1.69)⁎ 1.42 (0.91–2.23) 1.14 (0.82–1.60)
Referent 1.55 (1.30–1.84)⁎ 1.20 (0.97–1.48) 1.35 (1.15–1.59)⁎
Referent 1.40 (1.08–1.81)⁎ 1.11 (0.78–1.58) 1.21 (0.93–1.58)
Referent 0.88 (0.70–1.10) 0.69 (0.57–0.85)⁎ 2.95 (1.26–6.92)⁎
Referent 0.80 (0.59–1.10) 0.61 (0.47–0.81)⁎ 5.29 (2.11–13.27)⁎
Referent 0.98 (0.84–1.15) 0.91 (0.79–1.05) 1.00 (0.46–2.19)
Referent 1.07 (0.82–1.39) 1.04 (0.82–1.31) 3.46 (1.45–8.29)⁎
Referent 1.30 (1.00–1.69)⁎ 1.13 (0.94–1.35)
Referent 1.17 (0.82–1.67) 0.99 (0.76–1.28)
Referent 2.15 (1.81–2.57)⁎ 1.12 (0.98–1.27)
Referent 1.87 (1.46–2.41)⁎ 1.10 (0.90–1.35)
Referent 2.67 (1.82–3.91)⁎ 1.31 (0.52–3.34) 0.48 (0.11–2.17)
Referent 2.41 (1.55–3.73)⁎ 0.65 (0.18–2.30) –a
Referent 3.57 (2.45–5.20)⁎ 2.12 (0.70–6.44) 0.98 (0.41–2.33)
Referent 4.05 (2.66–6.17)⁎ 5.53 (1.70–18.03)⁎ 0.42 (0.06–3.16)
Referent 1.15 (0.80–1.67)
Referent 0.99 (0.61–1.61)
Referent 1.57 (1.18–2.08)⁎
Referent 1.31 (0.88–1.95)
Referent 2.31 (1.61–3.32)⁎
Referent 2.42 (1.57–3.74)⁎
Referent 2.22 (1.64–3.01)⁎
Referent 1.84 (1.23–2.77)⁎
–a Referent 2.22 (0.95–5.16) 3.11 (1.34–7.25)⁎
8.36 (3.45–20.26)⁎
–a Referent 1.98 (0.48–8.22) 3.88 (0.94–16.03) 8.75 (2.05–37.39)⁎
3.51 (0.61–20.13) Referent 0.96 (0.68–1.36) 1.30 (0.92–1.85) 2.82 (1.84–4.32)⁎
7.64 (0.79–73.67) Referent 1.08 (0.54–2.16) 1.80 (0.90–3.61) 5.69 (2.69–12.03)⁎
1.007 (1.006–1.008)⁎
1.006 (1.005–1.007)⁎
1.007 (1.006–1.007)⁎
1.007 (1.006–1.007)⁎
Referent 1.67 (1.32–2.10)⁎
Referent 1.74 (1.30–2.33)⁎
Referent 1.28 (0.92–1.78)
Referent 2.35 (1.57–3.53)⁎
Referent 7.44 (4.55–12.16)⁎
Referent 5.15 (3.19–8.30)⁎
Referent 3.93 (1.91–8.08)⁎
Referent 10.47 (5.03–21.79)⁎
Referent 2.49 (1.51–4.10)⁎
Referent 3.28 (1.92–5.61)⁎
Referent 2.87 (1.48–5.59)⁎
Referent 4.06 (1.91–8.62)⁎
Referent 2.50 (1.77–3.52)⁎
Referent 2.33 (1.56–3.49)⁎
Referent 1.61 (1.20–2.14)⁎
Referent 1.95 (1.32–2.88)⁎
Referent 4.65 (2.70–8.01)⁎
Referent 4.96 (2.86–8.60)⁎
Referent 2.03 (1.10–3.74)⁎
Referent 2.07 (1.01–4.26)⁎
Referent 2.50 (1.58–3.95)⁎
Referent 3.66 (2.27–5.90)⁎
Referent 1.30 (0.72–2.33)
Referent 0.83 (0.39–1.75)
Referent 2.11 (1.42–3.16)⁎
Referent 3.28 (2.14–5.04)⁎
Referent 1.70 (1.13–2.54)⁎
Referent 3.61 (2.29–5.70)⁎
Logistic regression models included year of operation, age, race, preoperative BMI, diabetes, functional status, COPD, bleeding disorder, ASA class, total operation time, postoperative wound infection, pneumonia, PE, UTI, DVT/thrombophlebitis, sepsis, and reoperation. In the minimally invasive hysterectomy group, three patients with missing operation time were not included in the model. aOR: adjusted odds ratio. a Unestimable. ⁎ P-value b 0.05.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
invasive group, the 75th percentile for LOS was 1 day while the 90th percentile was 2 days (Table 2). Among women who underwent abdominal hysterectomy, older age, black race, insulin-dependent diabetes mellitus, partially dependent functional status, an underlying bleeding disorder and a higher ASA classification were all associated with a LOS N75th percentile (P b 0.05 for all) (Table 3). Likewise, a longer operative time and the occurrence of any of the postoperative complications were associated with a LOS N75th percentile (P b 0.05 for all). Similar patterns were identified for LOS N90th percentile for women who underwent abdominal hysterectomy. When minimally invasive hysterectomy was examined, advanced age, non-white race, comorbidities including insulin-dependent diabetes mellitus, non-independent functional status, COPD, bleeding disorders and a high ASA classification were associated with a LOS N75th percentile (P b 0.05 for all) (Table 3). A longer operative time and occurrence of the many of the postoperative complications were associated with a LOS of N75th percentile (P b 0.05 for all). In contrast, more recent year of surgery was associated with a lower likelihood of a prolonged hospitalization. The trends were relatively similar when LOS N90th percentile was examined. For abdominal hysterectomy, the examined factors were able to explain 23.6% of the variation in LOS N75th percentile (Pseudo-R2) (Table 4). Models including only one group of variables demonstrated a pseudo-R2 of 4.0% for demographic, 7.0% for preoperative, 7.9% for intraoperative, and 9.7% for postoperative characteristics (C-statistic: 0.61, 0.63, 0.65, 0.62, respectively). When LOS N90th percentile was examined, the full model explained 23.3% of the variation in LOS. Models including only one group of variables demonstrated a pseudo-R2 of 2.9% for demographic, 7.4% for preoperative, 4.8% for intraoperative, and 12.8% for postoperative characteristics (C-statistic: 0.62, 0.66, 0.63, 0.67, respectively). Table 4 Significance of inclusion or omission of 4 groups of clinical variables from a multivariable logistic regression model for prolonged length of stay in uterine cancer patients undergoing abdominal hysterectomy. C-statistic Pseudo-R2 Length of stay N 75th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions Length of stay N 90th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions
0.61 0.63 0.65 0.62
4.0% 7.0% 7.9% 9.7%
0.75 0.74 0.72 0.73
21.4% 19.4% 17.2% 17.4%
0.77
23.6%
0.62 0.66 0.63 0.67
2.9% 7.4% 4.8% 12.8%
0.78 0.76 0.77 0.74
21.6% 18.6% 19.9% 14.0%
0.79
23.3%
Logistic regression models were fitted and compared. Demographic factors included year, age, and race. Preoperative conditions included BMI, diabetes, functional status, COPD, bleeding disorders, and ASA class. Intraoperative conditions included total operation time. Postoperative complications included wound infection, pneumonia, PE, UTI, DVT, sepsis, and reoperation.
7
For minimally invasive hysterectomy, the examined factors were able to explain 16.2% of the variation in LOS N75th percentile (PseudoR2) (Table 5). Models including only one group of variables demonstrated a pseudo-R2 of 6.2% for demographic, 4.1% for preoperative, 6.9% for intraoperative, and 1.3% for postoperative characteristics (Cstatistic: 0.63, 0.60, 0.64, 0.53, respectively). When LOS N90th percentile was examined, the full model explained 16.2% of the variation in LOS. Models including only one group of variables demonstrated a pseudoR2 of 4.3% for demographic, 5.5% for preoperative, 5.1% for intraoperative, and 3.6% for postoperative characteristics (C-statistic: 0.64, 0.65, 0.64, 0.56, respectively). For both abdominal and minimally invasive hysterectomy, a series of sensitivity analyses indicated that the perioperative factors examined had similar importance to prolonged LOS (supplemental tables). 4. Discussion We found substantial variation in the LOS of women who underwent hysterectomy for uterine cancer. While postoperative complications explain a portion of the risk, non-modifiable demographic and clinical factors also account for a significant portion of the explained variation in prolonged LOS. Measurable demographic, preoperative, intraoperative, and postoperative factors explain a smaller percentage of the variation in LOS for women undergoing minimally invasive procedures. Overall, both procedures had a significant portion of variation in LOS that could not be accounted for by measurable patient, hospital or clinical factors. To date, studies analyzing the factors that drive LOS for patients undergoing hysterectomy have been limited. A few small, single-center studies have analyzed LOS and overall gynecologic cancer hospital
Table 5 Significance of inclusion or omission of 4 groups of clinical variables from a multivariable logistic regression model for prolonged length of stay in uterine cancer patients undergoing minimally invasive hysterectomy. C-statistic Pseudo-R2 Length of stay N 75th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions Length of stay N 90th percentile One group of variables included in model Demographic factors Preoperative conditions Intraoperative conditions Postoperative complications One group of variable omitted from model Demographic factors omitted Preoperative conditions omitted Intraoperative conditions omitted Postoperative conditions omitted Full model Demographics, preoperative, intraoperative, and postoperative conditions
0.63 0.60 0.64 0.53
6.2% 4.1% 6.9% 1.3%
0.69 0.70 0.68 0.72
11.9% 12.8% 10.6% 15.5%
0.72
16.2%
0.64 0.65 0.64 0.56
4.3% 5.5% 5.1% 3.6%
0.74 0.72 0.73 0.75
13.5% 11.6% 12.1% 13.6%
0.76
16.2%
Three patients missing operation time were not included. Logistic regression models were fitted and compared. Demographic factors included year, age, and race. Preoperative conditions included BMI, diabetes, functional status, COPD, bleeding disorders, and ASA class. Intraoperative conditions included total operation time. Postoperative complications included wound infection, pneumonia, PE, UTI, DVT, sepsis, and reoperation.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
8
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx
admissions and have found that older age, multiple co-morbidities, lower preoperative albumin and hemoglobin levels, multiple postoperative complications, a higher ASA class, and prior chemotherapy were predictive of extended LOS [11,12]. A 2010 prospective cohort study of 157 patients at a tertiary center found that malnutrition, low quality of life scores, and a diagnosis of advanced stage ovarian cancer were associated with prolonged LOS [13]. Examination of LOS for specific gynecologic cancers demonstrates similar risk factors. Among epithelial ovarian cancer patients, older age, higher ASA score, worse Eastern Cooperative Oncology Group (ECOG) status, and higher CA-125 levels were predictive of non-home discharges at one center [14]. For patients undergoing robotic surgery for endometrial cancer, factors such as older age, prior myocardial infarction, and the need for perioperative blood transfusions were associated with a longer hospital stay [15]. Our data is consistent with the above studies showing that factors from each stage of the perioperative period influence LOS for patients undergoing both abdominal and minimally invasive hysterectomy. However, if LOS is to be utilized as a quality metric, hospitals and physicians must have the ability to modify LOS. Examination of risk factors for partial nephrectomy, colectomy, and trauma patients have shown that non-modifiable patient traits often play an important role in determining prolonged LOS [16–18]. For patients undergoing hysterectomy, our data suggest first, that a large portion of the variation in LOS is not explained by measurable clinical and demographic characteristics. There are likely a number of factors that contribute to this unexplained variation including socioeconomic factors, a patient's functional status, a patient's social support network as well as hospital characteristics and practice patterns. Second, much of the variation that can be explained is due to underlying demographic characteristics that are not modifiable. These factors raise concern about how actionable LOS is as a quality metric. While understanding drivers of LOS is critical for hospital reimbursement, LOS also contributes to patient satisfaction. Studies in other surgical fields have found that patient satisfaction is higher in those who undergo procedures with shorter LOS [19,20]. Recently, a large national, aggregate study demonstrated that shorter LOS after six of the most common major surgeries, not including hysterectomy, was associated with higher patient satisfaction scores on the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey [21]. HCAHPS scores can have a large impact on public perception as well as hospital reimbursement [22,23]. One major concern with the widespread efforts to reduce hospital LOS is that successful reduction of LOS may contribute to increasing readmission rates, another CMS quality metric [24–27]. However, there is no strong evidence to suggest that a negative relationship exists between these metrics. An analysis of the Veterans Affairs hospital system over a 14-year period showed reduced LOS for medical diagnoses were not associated with an increase in 30-day readmission rates, but rather in a reduction in the readmission rate as well [28]. Previous work done by our group revealed that 30-day readmission for patients undergoing hysterectomy for benign indications and uterine cancer was driven mainly by postoperative complications [29]. The implications of the previous work and the current study are that improving LOS through reduction in modifiable factors, such as postoperative complications, would decrease both LOS and 30-day readmission rates for patients undergoing hysterectomy. We acknowledge a number of important limitations. Although an effort was made to capture a wide array of hospitals, NSQIP does not include data from every hospital in the U.S. and therefore, the conclusions may not be generalizable. We were also unable to distinguish between laparoscopic and robotic hysterectomy from the database, which are grouped together as minimally invasive hysterectomy. These procedures have unique risk profiles and may affect LOS differently. NSQIP lacks data on hospital characteristics which would be of great interest in the analysis. NSQIP does not capture some
complications that are specific to hysterectomy; thus, our analysis is based on commonly reported perioperative complications. In conclusion, only a portion of the variation in LOS after hysterectomy for endometrial cancer is explained by measurable clinical and demographic factors. While some of the explained variation is accounted for by modifiable factors (intraoperative and postoperative), nonmodifiable factors (demographic and preoperative) also contribute to prolonged LOS after hysterectomy. In the context of continuously evolving reimbursement systems and the desire for improved patient satisfaction, specifically measuring and addressing modifiable risk factors should be a focus.
Disclosure Dr. Wright has served as a consultant for Tesaro and Clovis Oncology. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.ygyno.2018.03.048. References [1] D. Blumenthal, K. Davis, S. Guterman, Medicare at 50—origins and evolution, N. Engl. J. Med. 372 (2015) 479–486. [2] D. Blumenthal, K. Davis, S. Guterman, Medicare at 50—moving forward, N. Engl. J. Med. 372 (2015) 671–677. [3] M.K. Whiteman, S.D. Hillis, D.J. Jamieson, et al., Inpatient hysterectomy surveillance in the United States, 2000–2004, Am. J. Obstet. Gynecol. 198 (34) (2008) e1–7. [4] C.M. Farquhar, C.A. Steiner, Hysterectomy rates in the United States 1990–1997, Obstet. Gynecol. 99 (2002) 229–234. [5] S.R. Kovac, Hysterectomy outcomes in patients with similar indications, Obstet. Gynecol. 95 (2000) 787–793. [6] E.S. Campbell, H. Xiao, M.K. Smith, Types of hysterectomy. Comparison of characteristics, hospital costs, utilization and outcomes, J. Reprod. Med. 48 (2003) 943–949. [7] R. Garry, J. Fountain, S. Mason, et al., The eVALuate study: two parallel randomised trials, one comparing laparoscopic with abdominal hysterectomy, the other comparing laparoscopic with vaginal hysterectomy, BMJ 328 (2004) 129. [8] J.M. Wu, M.E. Wechter, E.J. Geller, T.V. Nguyen, A.G. Visco, Hysterectomy rates in the United States, 2003, Obstet. Gynecol. 110 (2007) 1091–1095. [9] J.F. Boggess, P.A. Gehrig, L. Cantrell, et al., A comparative study of 3 surgical methods for hysterectomy with staging for endometrial cancer: robotic assistance, laparoscopy, laparotomy, Am. J. Obstet. Gynecol. 199 (360) (2008) e1–9. [10] ACS NSQIP participant use data file, Program ACoSNSQI, 2015. [11] L.S. Massad, G. Vogler, T.J. Herzog, D.G. Mutch, Correlates of length of stay in gynecologic oncology patients undergoing inpatient surgery, Gynecol. Oncol. 51 (1993) 214–218. [12] M.M. Dean, M.A. Finan, R.C. Kline, Predictors of complications and hospital stay in gynecologic cancer surgery, Obstet. Gynecol. 97 (2001) 721–724. [13] B. Laky, M. Janda, S. Kondalsamy-Chennakesavan, G. Cleghorn, A. Obermair, Pretreatment malnutrition and quality of life - association with prolonged length of hospital stay among patients with gynecological cancer: a cohort study, BMC Cancer 10 (2010) 232. [14] M.M. AlHilli, C.W. Tran, C.L. Langstraat, et al., Risk-scoring model for prediction of non-home discharge in epithelial ovarian cancer patients, J. Am. Coll. Surg. 217 (2013) 507–515. [15] M.I. Liang, M.A. Rosen, K.S. Rath, et al., Predicting inpatient stay lasting 2 midnights or longer after robotic surgery for endometrial cancer, J. Minim. Invasive Gynecol. 22 (2015) 583–589. [16] J.A. Larson, J.H. Kaouk, M.D. Stifelman, et al., Nonmodifiable factors and complications contribute to length of stay in robot-assisted partial nephrectomy, J. Endourol. 29 (2015) 422–429. [17] K.G. Cologne, S. Byers, D.R. Rosen, et al., Factors associated with a short (b2 days) or long (N10 days) length of stay after colectomy: a multivariate analysis of over 400 patients, Am. Surg. 82 (2016) 960–963. [18] K.J. Brasel, H.J. Lim, R. Nirula, J.A. Weigelt, Length of stay: an appropriate quality measure? Arch. Surg. 142 (2007) 461–465 (discussion 5–6). [19] L.H. Cohn, D.H. Adams, G.S. Couper, et al., Minimally invasive cardiac valve surgery improves patient satisfaction while reducing costs of cardiac valve replacement and repair, Ann. Surg. 226 (1997) 421–426 (discussion 7–8). [20] H. Husted, G. Holm, S. Jacobsen, Predictors of length of stay and patient satisfaction after hip and knee replacement surgery: fast-track experience in 712 patients, Acta Orthop. 79 (2008) 168–173. [21] T.C. Tsai, E.J. Orav, A.K. Jha, Patient satisfaction and quality of surgical care in US hospitals, Ann. Surg. 261 (2015) 2–8. [22] L.A. Giordano, M.N. Elliott, E. Goldstein, W.G. Lehrman, P.A. Spencer, Development, implementation, and public reporting of the HCAHPS survey, Med. Care Res. Rev. 67 (2010) 27–37.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048
S. Agrawal et al. / Gynecologic Oncology xxx (2018) xxx–xxx [23] Hospital Value-Based Purchasing Program: Centers for Medicare and Medicaid Services (CMS)2015. [24] A. Clarke, R. Rosen, Length of stay. How short should hospital care be? Eur. J. Pub. Health 11 (2001) 166–170. [25] P.A. Cowper, E.R. DeLong, E.L. Hannan, et al., Is early too early? Effect of shorter stays after bypass surgery, Ann. Thorac. Surg. 83 (2007) 100–107. [26] D.W. Baker, D. Einstadter, S.S. Husak, R.D. Cebul, Trends in postdischarge mortality and readmissions: has length of stay declined too far? Arch. Intern. Med. 164 (2004) 538–544.
9
[27] S. Hendren, A.M. Morris, W. Zhang, J. Dimick, Early discharge and hospital readmission after colectomy for cancer, Dis. Colon Rectum 54 (2011) 1362–1367. [28] P.J. Kaboli, J.T. Go, J. Hockenberry, et al., Associations between reduced hospital length of stay and 30-day readmission rate and mortality: 14-year experience in 129 veterans affairs hospitals, Ann. Intern. Med. 157 (2012) 837–845. [29] K. Dessources, J.Y. Hou, A.I. Tergas, et al., Factors associated with 30-day hospital readmission after hysterectomy, Obstet. Gynecol. 125 (2015) 461–470.
Please cite this article as: S. Agrawal, et al., Identifying modifiable and non-modifiable risk factors associated with prolonged length of stay after hysterectomy for uterine can..., Gynecol Oncol (2018), https://doi.org/10.1016/j.ygyno.2018.03.048