Surgery for Obesity and Related Diseases 11 (2015) 866–873
Original article
Predicting potentially preventable hospital readmissions following bariatric surgery Wendy L. Patterson, M.P.H.*, Brittany D. Peoples, M.S., Foster C. Gesten, M.D. Office of Quality and Patient Safety, New York State Department of Health, Albany, NY Received September 23, 2014; accepted December 17, 2014
Abstract
Background: Using hospital readmissions as a quality of care measure predicates that some readmissions were preventable. Objectives: This study identifies predictors of potentially preventable readmissions (PPR) within 30 days of bariatric surgery discharge. Setting: New York State acute care hospitals. Methods: Adult inpatient surgical discharges, during 2012, with a principal diagnosis of overweight or obesity and a principal procedure for bariatric surgery were identified. Logistic regression was used to evaluate surgical approach, sex, age, race/ethnicity, payor, body mass index, complications and co-morbidities recorded during the surgical admission. Results: There were 10,448 surgeries studied for readmission of which 552 were followed by a PPR, for a statewide rate of 5.3 per 100 surgeries. Laparoscopic Roux-en-Y Gastric Bypass (LRYGB) was the most common surgical approach (46.0%), then Sleeve Gastrectomy (SG) (41.3%), Laparoscopic Adjustable Gastric Band (LAGB) (8.1%), and Open Roux-en-Y Gastric Bypass (RYGB) (4.6%). RYGB had the highest PPR rate (8.8), followed by LRYGB (6.1), SG (4.3) and LAGB (3.3). Compared to LAGB, the odds of a PPR in patients with RYGB, LRYGB, and SG increased by 2.4 fold, 1.8 fold and 1.2 fold respectively. Black, non-Hispanic patients were at a greater risk of PPR (odds-ratio 2.0, P o .0001) compared to White, non-Hispanic patients while the risk of a PPR increased by 2-fold in patients with a surgical complication. Conclusions: Taking all patient risk factors into account, the most significant predictors of a PPR were surgical approach, race and the presence of a surgical complication. (Surg Obes Relat Dis 2015;11:866–873.) r 2015 American Society for Metabolic and Bariatric Surgery. All rights reserved.
Keywords:
Bariatric surgery; Gastric bypass; Laparoscopy; Sleeve resection; Co-morbidity; Risk factors; Logistic regression; Readmission; Preventable; PPR
While there are benefits of bariatric surgery, there are potential consequences as well. These include postoperative complications before discharge [1–8] and hospital readmissions [9–22]. As the total volume of bariatric surgery continues to increase, and the link between quality of inpatient care, inpatient costs and readmissions receives continued attention [19,23], it becomes increasingly * Correspondence: Wendy L. Patterson, New York State Department of Health, 878 Corning Tower, Empire State Plaza, Albany, NY 12237. Phone: þ1-518-486-9012. E-mail: [email protected]
important to better understand outcomes associated with this intervention. Previous studies have used any type of hospital readmission within a specified time frame [24]. However, all readmissions are not related to inadequate care and using them in the calculation of readmission rates may be misleading. The usefulness of readmissions as a measure of quality of care is predicated on the notion that the readmissions might have been prevented. It is essential that credible and consistent clinical criteria be used to define that subset of readmissions that were potentially preventable [25].
http://dx.doi.org/10.1016/j.soard.2014.12.019 1550-7289/r 2015 American Society for Metabolic and Bariatric Surgery. All rights reserved.
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873
The purpose of this study is to identify predictors of a potentially preventable readmission after bariatric surgery. This study is the first to evaluate readmissions after bariatric surgery using the Potentially Preventable Readmission (PPR) Classification System Software developed by 3 M Health Information Systems [26]. This algorithm uses precisely defined clinical criteria to identify those readmissions that are clinically related to previous hospitalizations and thus may have been preventable. In previous research, Saunders et al. [10,27] distinguished among the various bariatric surgical approaches and compared readmission rates between approaches. Lindsey et al. [8] found that complication rates varied dramatically according to the bariatric surgical approach. In this paper, we will explore potentially preventable readmission rates by surgical approach and develop a multivariate logistic model to calculate the odds of a potentially preventable readmission based on surgical approach and patient risk factors, including patient demographic characteristics, postoperative complications and co-morbidities. Methods Hospital inpatient discharge data from the New York State Statewide Planning and Research Cooperative System (SPARCS) were used in this study. SPARCS collects patient level detail regarding patient demographic characteristics, diagnoses, procedures, discharge status, and charges for every inpatient discharge from New York State hospitals. The study population was limited to adult (age 18 or older) inpatient discharges between January 1, 2012 and December 31, 2012 from all hospitals in New York State with an International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) principal diagnosis code for overweight or obesity (278.00, 278.01, or 278.02) and an ICD-9-CM principal procedure code for bariatric surgery. Bariatric surgery was defined as Open Roux-en-Y Gastric Bypass (RYGB) (44.31, 44.39, 44.69), Laparoscopic Roux-en-Y Gastric Bypass (LRYGB) (44.38), Laparoscopic Adjustable Gastric Band (LAGB) (44.68, 44.95), and Sleeve Gastrectomy (SG) (43.82, 43.89). SPARCS data for all of 2012 were analyzed for bariatric surgery, with inpatient readmissions extending through January 31, 2013, to allow 30 days for readmission. The age and diagnostic restrictions ensured the bariatric surgery was for weight loss. The dependent variable was whether or not the bariatric surgical inpatient admission was followed by a potentially preventable inpatient admission within 30 days of discharge using the PPR software (Version 30.0 [3M, Salt Lake City, UT]) [25,26]. We used the 30 day timeframe because readmissions during that time are more likely to be related to the initial hospitalization and the transition of care to the outpatient setting. In addition, this timeframe is used in other
867
publicly reported measures from the National Quality Forum (NQF) and the Centers for Medicare and Medicaid Services (CMS). There were no restrictions that the readmission had to occur at the same hospital as the bariatric surgery. The PPR software first identifies globally excluded inpatient admissions and removes them from analysis. These include; most types of major metastatic malignancies, trauma, burns, many types of obstetric admissions and newborns, as well as patients whose treatment abruptly ended (the patient left against medical advice or the patient was transferred to another hospital). If suitable for analysis, the admission is classified as an initial admission, readmission, or only admission. The PPR software determines if an admission is clinically related to a prior admission and occurred in the readmission time interval. Clinically related means the underlying reason for admission is plausibly related to the care rendered during or immediately following the prior hospital admission and may have resulted from the process of care and treatment during the prior admission or from a lack of postadmission follow up. An initial admission is an inpatient admission followed by at least one clinically related inpatient admission within the readmission timeframe. A readmission is a clinically related inpatient admission that followed the initial inpatient admission within the readmission timeframe. An initial admission and all the readmissions associated with that initial admission form a PPR chain. If the admission was suitable to be followed by a readmission (not excluded) but was not, it is classified an only admission. The patient characteristics evaluated included gender, age, race/ethnicity, payor, and body mass index (BMI). Age was calculated at the time of admission and categorized into 4 groups: 18–29 years of age, 30–39 years of age, 40–49 years of age, and 50 years and older. Race/ethnicity was categorized into the following 4 groups: white, non-Hispanic; black, non-Hispanic; Hispanic; and Other. The Other group included Asian, Native Hawaiian/Pacific Islander, Native American, and those for whom race was recorded as other or unknown. Payor was identified by the source of payment listed on the surgical admission. Payors were split into 4 major groups: Medicaid, Medicare, Commercial (workers compensation, insurance company, self-pay, Blue Cross) and Other (Other federal program, CHAMPUS, other nonfederal program). BMI was identified by the diagnosis code recorded on the surgical admission and was grouped into 5 groups: BMI of 19.0–39.9 kg/m2, 40.0–44.9 kg/m2, 45.0– 49.9 kg/m2, 50.0 kg/m2 and over, and other (where BMI was either missing or miscoded as a pediatric code). In addition, we included whether or not the patient experienced a postoperative complication during the bariatric surgery admission [8]. Patient co-morbidities were defined using the approach developed by Elixhauser et al. [8,28]. We excluded a
868
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
number of these co-morbidities for the following reasons: those which defined the condition we were investigating (obesity and weight loss), those which had 6 or fewer patients that had a PPR (congestive heart failure, pulmonary circulation disease, peripheral vascular disease, paralysis, peptic ulcer disease, AIDS, lymphoma, metastatic cancer, solid tumor w/out metastasis, coagulopathy, chronic blood loss anemia, alcohol abuse, and drug abuse), and those that could be considered postoperative bariatric surgical complications (renal failure, fluid and electrolyte disorders, and deficiency anemias). Bivariate analyses were performed to investigate the relationship between each patient risk factor present at the time of surgery and whether or not the surgery was followed by a PPR for each of the surgical approaches. We used χ2 tests to examine the relationship between nonordered categorical variables (race/ethnicity, gender, the presence or absence of each of the risk factors) and the presence or absence of a PPR. For age, the CochraneArmitage test for trend was used to assess the existence of any linear trend in PPR frequency across the 4 age categories. Data for all 4 surgical approaches were pooled and a multivariate logistic regression model was constructed that contained indicators for all patient risk factors. While surgical approach may be influenced by the patient’s overall health status and surgical preference of the physician, creating some preselection of risk, we tried to control for much of this with our statistical model. This model permitted us to calculate the odds of a PPR following bariatric surgery associated with each of these patient risk factors. All analyses were conducted using SAS version 9.4 (Cary, NC). Results A total of 10,489 inpatient bariatric surgeries for obesity were performed in New York hospitals during 2012. After globally excluded admissions were removed, 10,448 discharges remained for analysis. The most common bariatric surgical approach was LRYGB, 4,807 (46.0%), while 4,319 surgeries (41.3%) were SG, 846 (8.1%) were LAGB, and 476 (4.6%) were RYGB. Overall, there were 552 bariatric surgeries that were followed by at least one PPR, for a statewide PPR rate of 5.3 per 100 discharges. The PPR rate was calculated by dividing the number of surgeries that were followed by at least one PPR by the total number of bariatric surgeries (552/10,448). These 552 surgeries were followed by 711 potentially preventable readmissions, an average of 1.3 readmissions per surgery among patients with a readmission. Of those patients experiencing a readmission, the vast majority, 69%, had only a single readmission, 20% had 2, and 11% had 3 or more readmissions. The most common primary diagnosis codes
for readmissions were ICD-9 539.89 (other complication after bariatric surgery, 16%), and ICD-9 276.51 (dehydration, 12%). Demographic characteristics and co-morbidities for the bariatric surgical population are displayed in Table 1. The population was predominantly female (79.2%); over 40 years of age (58.5%); white, non-Hispanic (53.4%); and covered by commercial insurance (84.4%). Nearly half (45.0%) of the patients had a BMI of 45 kg/m2 or greater and 5.4% of the patients had a complication during the surgical admission. The most common co-morbidities were hypertension (51.9%), diabetes without chronic complications (28.7%), chronic pulmonary disease (22.1%), and depression (19.4%). Table 1 also demonstrates substantial variation in PPR rates among the 4 surgical approaches. RYGB discharges had the highest PPR rate, 8.8 per 100 discharges, followed by LRYGB (6.1 per 100), SG (4.3 per 100), and LAGB (3.3 per 100). The table indicates there are differences among the 4 surgical approaches in terms of the patient characteristics and co-morbidities associated with PPRs. Black, non-Hispanic patients experienced a PPR more often when they had RYGB compared to any other surgical approach. LRYGB patients experiencing a complication had PPRs more often than other approaches. LRYGB patients with chronic pulmonary disease experienced a PPR more often than other surgical approaches. SG patients diagnosed with diabetes, with or without chronic complications, experienced a PPR more often than other patients. Compared to LAGB, the odds of a PPR in patients with RYGB and LRYGB increased by 2.4 fold and 1.8 fold respectively (Table 2). The odds of a PPR for Black, nonHispanic patients was almost 2-fold greater than White, non-Hispanic patients (odds ratio 2.0, P o .0001). Patients with a complication were at greater risk of a PPR than those without a complication (odds ratio 1.9, P o .0001). Patients with chronic pulmonary disease (odds ratio 1.5, P ¼ .0001), diabetes with (odds ratio 1.9, P ¼ .0290) and without (odds ratio 1.3, P ¼ .0118) chronic complications or rheumatoid arthritis/collagen disease (odds ratio 1.8, P ¼ .0304) had statistically higher odds of a PPR, compared to those patients without the condition at the time of bariatric surgery. Discussion Hospital readmissions are one potential indicator of the quality of inpatient care. Those readmissions that may have been prevented are of particular importance to healthcare organizations, clinicians, and payors given their link to reimbursement and comparative quality assessments. The major finding of this study was that the potentially preventable readmission rates varied significantly
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873
869
Table 1 Potentially preventable readmissions (PPR) by patient demographic characteristics and co-morbidities, by surgical approach Characteristic
Surgical approach RYGB
LRYGB
Total PPR % Surgical Approach Gender Female Male Age (Years) 18–29 30–39 40–49 50 and older Race / Ethnicity White, non-Hispanic Black, non-Hispanic Hispanic Other Payor Medicaid Medicare Commercial Other payor BMI BMI 19–39.9 kg/m2 BMI 40–44.9 kg/m2 BMI 45–49.9 kg/m2 BMI 50þ kg/m2 Other Complication Yes No Elixhauser Co-morbidities Valvular disease Yes No Hypertension Yes No Other neurologic disorders Yes No Chronic pulmonary disease Yes No Diabetes without chronic complications Yes No Diabetes with chronic complications Yes No Hypothyroidism Yes No Liver disease Yes No Rheumatoid arthritis/collagen disease Yes No
476
42
LAGB
P Value Total PPR %
8.82
4807 295
362 114
31 11
8.56 9.65
44 108 144 180
4 8 18 12
9.09 7.41 12.50 6.67
331 68 44 33
19 10 6 7
5.74 14.71 13.64 21.21
57 92 315 12
3 6 30 3
5.26 6.52 9.52 25.00
65 123 93 178 17
6 13 12 9 2
9.23 10.57 12.90 5.06 11.76
55 421
7 35
12.73 8.31
4 472
0 42
.00 8.90
286 190
26 16
9.09 8.42
7 469
1 41
14.29 8.74
111 365
13 29
11.71 7.95
3852 250 955 45
6.49 4.71
710 1260 1335 1502
6.06 6.11 7.12 5.33
8.04 7.27 5.79 8.91
.22 44 5.98 92 5.65 52 4.74 94 7.65 13 11.02
288 36 12.50 4519 259 5.73 .53
5.95 6.36
6.25 9.22
42 434
4 38
9.52 8.76
12 464
1 41
8.33 8.84
.84 2.36 3.98 4.46
8.85 5.35
475 133 115 123
20 6 2 0
4.21 4.51 1.74 .00
6.25 6.09
91 7 4716 288
7.69 6.11
583 35 4224 260
6.00 6.16
857 51 3950 244
5.95 6.18
.71
14 136 692 4
0 8 20 0
.00 5.88 2.89 .00
183 306 186 135 36
5 12 6 5 0
2.73 3.92 3.23 3.70 .00
26 820
1 27
3.85 3.29
18 828
2 26
11.11 3.14
4.84 3.79 4.51 4.36
1936 791 986 606
73 54 37 23
3.77 6.83 3.75 3.80
246 15 182 9 3860 162 31 1
6.10 4.95 4.20 3.23
854 1494 860 925 186
34 68 28 51 6
3.98 4.55 3.26 5.51 3.23
194 15 4125 172
7.73 4.17
28 0 4291 187
.00 4.36
406 440
13 15
3.20 3.41
2152 103 2167 84
4.79 3.88
10 836
0 28
.00 3.35
62 1 4257 186
1.61 4.37
172 674
10 18
5.81 2.67
937 50 3382 137
5.34 4.05
.71
o.01
.52
.16
.02
.06
.26
.87
.14
.56
.29
.04
.09
.24 221 625
10 18
4.52 2.88
5 841
0 28
.00 3.33
115 731
3 25
2.61 3.42
50 796
1 27
2.00 3.39
16 830
1 27
6.25 3.25
.01 1111 63 3208 124
5.67 3.87 o.01
.68 46 7 15.22 4273 180 4.21 .65
.80 438 20 3881 167
4.57 4.30
366 16 3953 171
4.37 4.33
68 6 4251 181
8.82 4.26
.59
.05 57 7 12.28 4750 288 6.06
35 44 55 53
.88
.80
.95
723 1162 1219 1215
P Value .75
.76
.89
.87
4.28 4.52
.28
.53
.44
3435 147 884 40
.08
.83 1472 92 3335 203
4.33
.22
o.01 1085 96 3722 199
.20
4 38
1 5 8 14
.05 91 10 10.99 4716 285 6.04
.22
64 412
119 212 201 314
.55 2573 153 2234 142
.61
6.25 8.91
3.51 2.73
.01 26 5 19.23 4781 290 6.07
.80
1 41
22 6
o.01
.28
16 460
626 220
.01 736 1629 1096 1228 118
4319 187 .57
.16 311 25 440 32 3955 229 101 9
P Value Total PPR %
3.31
o.01 2833 133 4.69 571 58 10.16 839 68 8.10 564 36 6.38
.13
10.82 7.45
28
.27 43 77 95 80
o.01
21 21
846 .04
.29
194 282
P Value Total PPR %
6.14
.72
SG
.97
.51
.07
870
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
Table 1 Continued. Characteristic
Surgical approach RYGB
LRYGB
Total PPR % Psychoses Yes No Depression Yes No
LAGB
P Value Total PPR % .19
17 459
0 42
.00 9.15
97 379
9 33
9.28 8.71
SG
P Value Total PPR % .16
107 10 4700 285
9.35 6.06
1115 72 3692 223
6.46 6.04
.86
P Value Total PPR % .14
8 838
1 27
151 695
6 22
12.50 3.22
P Value .41
81 5 4238 182
6.17 4.29
668 36 3651 151
5.39 4.14
.61
.14 3.97 .61 3.17
PPR ¼ Potentially preventable readmissions; RYGB ¼ Open Roux-en-Y gastric bypass; LRYGB ¼ Laparoscopic Roux-en-Y gastric bypass; LAGB ¼ Laparoscopic adjustable gastric band; SG ¼ Sleeve gastrectomy; BMI ¼ Body mass index.
depending on the bariatric surgical approach employed, even after accounting for patient differences. This was determined using a patient risk-based algorithm to identify only those 30 day hospital readmissions that were clinically related to the patient’s bariatric surgery. We found that 5.3% of the bariatric surgeries included in this study had a potentially preventable readmission within 30 days that could be related to the patient’s inpatient and/ or postdischarge outpatient care. In comparison, using the same data, 5.9% of the bariatric surgeries had any type of readmission within 30 days. Weller et al. [13], using 2003 New York hospital discharge data, found a bariatric all cause readmission rate of 7.6 per hundred. Since not all readmissions are potentially preventable, we would expect that the PPR rate would be less than these overall readmission rates. Conversely, a comparison of these rates suggests that many readmissions after bariatric surgery may be potentially preventable. To the best of our knowledge, there are no published studies describing PPR rates for bariatric surgical procedures, so there is no direct comparison available. We found significant variation in the crude PPR rates among the 4 surgical procedures. Saunders et al. found a 30-day all-cause bariatric surgery readmission rate of 7.3 per hundred for LRYGB and 3.1 per hundred for LAGB [10]. Similarly, Hutter found the highest 30-day readmission rate for LRYGB (6.5%), and the lowest for LAGB (1.7%), with SG in the middle at 5.4% [29]. Our method, which uses potentially preventable readmissions, found the same pattern in rates by surgical approach. It would be expected that LAGB, a minimally invasive procedure would result in a lower PPR rate than RYGB, a more invasive approach. Due to its complexity, this could explain the small number of RYGB procedures in New York State. It may not always be appropriate to select a procedure that will be less effective for a patient based solely on the procedures readmission rate. Many factors play a role in determining the type of surgical approach best for the patient (patient and surgeon preference, amount of long
term weight loss, intensity of postsurgical care, complications). Readmission rates should be considered as an additional factor when determining the potential risks and benefits of each surgical approach and deciding which surgical approach is best for a patient. Our results indicating that gender did not impact the odds of a PPR were similar to those of Zingmond et al. [11], but opposite of Weller et al.’s [12] analysis that found males at a greater risk for readmission. Baker et al. [16] found that patients with a BMI o40 kg/m2 were more likely to experience a hospital readmission compared to patients with a higher BMI, while our results indicated that BMI at any level was not significantly associated with the risk of PPRs. Weller et al. [12] found hypothyroidism and peptic ulcer disease increased the odds of subsequent hospital readmission. In our study, chronic pulmonary disease, diabetes with and without chronic complications and rheumatoid arthritis/collagen disease were significantly related to readmission. This study shares many of the limitations inherent in administrative data based studies, including incomplete or inaccurate diagnostic coding. In addition, previous work [12,14] has illustrated a significant relationship between the hospital and/or surgical volume of bariatric surgery and subsequent hospital readmissions. Unfortunately, a single year of data did not provide enough cases to perform a volume analysis by surgical approach. Despite these limitations, the findings of this study indicate that administrative data can be used to identify potentially preventable hospital readmissions after bariatric surgery, and that a significant number of potentially preventable readmissions occurred after bariatric surgery in New York State. In addition, these findings show that the amount of potentially preventable readmissions varies by the bariatric surgical approach employed, and that when all patient risk factors are taken into account, surgical approach, the presence of a complication during the initial admission and specific co-morbidities are significant predictors of a potentially preventable hospital readmission.
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873 Table 2 Logistic regression results of predicting a potentially preventable readmission Risk factor Surgical approach RYGB LRYGB LAGB SG Gender Male Female Age (Years) 18–29 30–39 40–49 50 and older Race/Ethnicity White, non-Hispanic Black, non-Hispanic Hispanic Other Payor Medicaid Medicare Commercial Other payor BMI BMI 19–39.9 kg/m2 BMI 40–44.9 kg/m2 BMI 45–49.9 kg/m2 BMI 50þ kg/m2 Other Complication Yes No Co-morbidities Valvular disease Hypertension Other neurologic disorders Chronic pulmonary disease Diabetes without chronic complications Diabetes with chronic complications Hypothyroidism Liver disease Rheumatoid arthritis/collagen disease Psychoses Depression
Adjusted odds ratio (95% CI)
P Value
2.39 (1.45–3.96) 1.79 (1.20–2.68) 1.00 1.21 (.80–1.83)
.00 .00
871
illustrated that hospital readmissions vary by surgical approach as well.
Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article.
.36
References .93 (.74–1.18) 1.00
.56
1.10 (.81–1.51) 1.00 (.77–1.29) 1.25 (.99–1.57) 1.00
.53 .99 .06
1.00 2.03 (1.61–2.57) 1.39 (1.09–1.77) 1.15 (.86–1.54) .69 (.36–1.33) .61 (.32–1.16) .59 (.33–1.07) 1.00 1.00 1.07 .87 1.20 1.20
(.83–1.40) (.64–1.17) (.91–1.58) (.73–1.98)
1.91 (1.43–2.55) 1.00
o.01 .01 .34 .26 .13 .08
.59 .34 .21 .47 o.01
2.01 .94 1.25 1.45 1.29
(.90–4.49) (.78–1.14) (.68–2.30) (1.20–1.76) (1.06–1.58)
.09 .55 .47 .00 .01
1.88 .98 .93 1.85
(1.07–3.31) (.74–1.30) (.72–1.21) (1.06–3.21)
.03 .90 .60 .03
1.35 (.79–2.28) 1.22 (.98–1.51)
.27 .08
RYGB ¼ Open Roux-en-Y gastric bypass; LRYGB ¼ Laparoscopic Roux-en-Y gastric bypass; LAGB ¼ Laparoscopic adjustable gastric band; SG ¼ Sleeve gastrectomy; BMI ¼ Body mass index. C-statistic ¼ .64
Conclusion Two measurable adverse outcomes of bariatric surgery are postoperative complications and hospital readmissions. We established in our previous research that postoperative complications vary by surgical approach [8]. This study has
[1] Liu JH, Zingmond D, Etzioni DA, et al. Characterizing the performance and outcomes of obesity surgery in California. Am Surg 2003;69:823–8. [2] Livingston EH. Procedure incidence and in-hospital complication rates of bariatric surgery in the United States. Am J Surg 2004;188: 105–10. [3] Trus TL, Pope GD, Finlayson RG. National trends in utilization and outcomes of bariatric surgery. Surg Endosc 2005;19:616–20. [4] Pope GD, Birkmeyer JD, Finlayson RG. National trends in utilization and in-hospital outcomes of bariatric surgery. J Gastrointest Surg 2002;6:855–61. [5] Mason EE, Renquist KE, Jiang D. Perioperative risks and safety of surgery for severe obesity. Am J Clin Nutr 1992;55:5735–65. [6] Livingston EH, Langert J. The impact of age and Medicare status on Bariatric Surgical outcomes. Arch Surg 2006;141:1115–20. [7] Weller WE, Rosati C, Hannan EL. Predictors of in-hospital postoperative complications among adults undergoing bariatric procedures in New York State, 2003. Obesity Surgery 2006;16:702–8. [8] Lindsey ML, Patterson WL, Gesten FC, Roohan PJ. Bariatric Surgery for Obesity: Surgical Approach and Variation in In-Hospital Complications in New York State. Obesity Surgery 2009;19:688–700. [9] Hong B, Stanley E, Reinhardt S, et al. Factors Associated with Readmission after Laparoscopic gastric bypass surgery. Surg Obes Relat Dis 2012;8:691–5. [10] Saunders J, Ballantyne G, Belsley S, et al. 30 day readmission rates at a high volume bariatric surgery center: Laparoscopic adjustable gastric banding, laparoscopic gastric bypass and vertical banded gastroplasty Roux-en-Y gastric bypass. Obesity Surgery 2007;17: 1171–7. [11] Zingmond DS, McGory ML, Ko CY. Hospitalization before and after gastric bypass surgery. JAMA 2005;294:1918–24. [12] Weller WE, Rosati C, Hannan EL. Relationship between surgeon and hospital volume and readmission after bariatric operation. J Am Coll Surg 2007;204:383–91. [13] Finkelstein EA, Brown DS. A Cost-Benefit Simulation Model of Coverage for Bariatric Surgery among Full-Time Employees. Am J Manag Care 2005;11:641–6. [14] Nguyen NT, Paya M, Stevens CM, et al. The Relationship between Hospital Volume and Outcome in Bariatric Surgery at Academic Medical Centers. Ann Surg 2004;240:586–94. [15] McCarty TM, Arnold DT, Lamont JP, et al. Optimizing Outcomes in Bariatric Surgery: Outpatient Laparoscopic Gastric Bypass. Ann Surg 2005;242:494–8. [16] Baker MT, Lara MD, Larson CJ, et al. Length of Stay and Impact on Readmission Rates After Laparoscopic Gastric Bypass. Surg Obes Relat Dis 2006;2:435–9. [17] Bradley DW, Sharma BK. Centers of Excellence in Bariatric Surgery: Design, Implementation, and One-Year Outcomes. Surg Obes Relat Dis 2006;2:513–7. [18] Escinosa WE, Bernard DM, Chen C, et al. Healthcare Utilization and Outcomes After Bariatric Surgery. Medical Care 2006;44:706–12.
872
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
[19] Ashton CM, DelJunco DJ, Souchek J, et al. The Association Between the Quality of Inpatient Care and Early Readmission: A MetaAnalysis of the Evidence. Medical Care 1997;35:1044–5. [20] Carlin AM, Zeni TM, English WJ, et al. The Comparative Effectiveness of Sleeve Gastrectomy, Gastric Bypass, and Adjustable Gastric Banding Procedures for the Treatment of Morbid Obesity. Ann Surg 2013;257:791–7. [21] Kwon S, Wang B, Wong E, et al. The Impact of Accreditation on Safety and Cost of Bariatric Surgery. Surg Obes Relat Dis 2013;9:617–22. [22] Nguyen NT, Nguyen B, Nguyen VQ, et al. Outcomes of Bariatric Surgery Performed at Accredited vs. Nonaccredited Centers. J Am Coll Surg 2012;215:467–74. [23] Friedman B, Basu J. The Rate and Cost of Hospital Readmissions for Preventable Conditions. Med Care Res and Rev 2004;61:225–40. [24] Dorman RB, Miller CJ, Leslie DB, et al. Risk for Hospital Readmission following Bariatric Surgery. PLoS ONE 2012;7:e32506.
[25] Goldfield NI, McCullough EC, Hughes JS, et al. Identifying Potentially Preventable Readmissions. Health Care Financing Review 2008;30:75–91. [26] 3 MTM Health Information Systems. Potentially Preventable Readmissions Classification System: Methodology Overview. Document number GRP-139 05/08. May 2008. [27] Saunders J, Ballantyne GH, Belsley B, et al. One year readmission rates at a high volume bariatric surgery center: Laparoscopic adjustable gastric banding, laparoscopic gastric bypass and vertical banded gastroplasty Roux-en-Y gastric bypass. Obesity Surgery 2008;18: 1233–40. [28] Elixhauser A, Syeiner C, Harris D. Co-morbidity measures for use with Administrative Data. Med Care 1998;36:8–27. [29] Hutter MM, Schirmer BD, Jones DB, et al. First Report from the American College of Surgeons Bariatric Surgery Center Network: Laparoscopic Sleeve Gastrectomy has Morbidity and Effectiveness Positioned Between the Band and the Bypass. Ann Surg 2011;254: 410–22.
Editorial
Comment on: Predicting potentially preventable hospital readmissions following bariatric surgery Decreasing the rate of readmissions in bariatric surgery has the potential to improve the delivery of patient care, increase patient satisfaction, and allow for a lower overall cost of care. There are numerous hurdles to truly impact readmission rates. Perhaps the first is to understand the true number of readmissions, and more importantly, those that can be prevented. The authors should be commended in their analysis presented in “Predicting Potentially Preventable Hospital Readmissions Following Bariatric Surgery,” which helps us get closer to understanding that true rate [1]. Additionally, they are carefully responsible with their consistent focus on potentially preventable readmissions as opposed to examining only the total readmission rate. In our program, when we implemented a process to more aggressively capture and understand bariatric readmissions, we witnessed an observer effect with an increase in the number of readmissions in the early months of our analysis. The plethora of definitions of what constitutes a readmission unfortunately confounds the study of this outcome variable. Standardizing the definition of readmissions in the broadest sense will help to identify the true baseline rate. Capturing and classifying all return visits to the hospital—whether an ER visit, 23 hour observation, or actual multiple-midnight stay readmission, as well as capturing patients that present to another hospital is critical to establishing this baseline. In the past, we observed programs that claimed extremely low readmission rates when in fact the readmissions were being admitted to another local hospital or a larger regional tertiary care facility and not being captured as part of the original surgery program data.
No matter what approach we use, there must be a reasonable attempt by CMS and other payors to determine what are acceptable preventable and total readmission rates [2]. At this point in time, an “acceptable” rate might in fact increase as the entire bariatric community improves its capture of readmission data. We must also be conscious of rewarding a reduction of readmissions. Physician compensation methodologies will all drive certain behaviors, potentially resulting in some unintentional consequences. Could either penalizing readmissions or rewarding a reduction of readmissions actually encourage a healthcare system to prevent a recent postoperative bariatric patient from coming to the hospital when that is indeed the best treatment? I would like to think not, but this is indeed within the realm of possibility. An alternative to penalizing a readmission could be to reward the behaviors that, when employed, can reduce preventable readmissions. One possibility is to incentivize physician, bariatric program, and healthcare system behaviors that allow for the investigation of the root cause of each readmission with a goal of reviewing 100% of readmissions. In this manner, the program can classify the readmission as potentially preventable or nonpreventable. If preventable, the program can then determine if there were any missed opportunities to do better. This multidisciplinary approach should be relatively easy to undertake within the constructs of an MBSAQIP-accredited center of excellence program in bariatric surgery. Readmissions will only be able to be affected through the development of standardized criteria for the definition of
Original article
Predicting potentially preventable hospital readmissions following bariatric surgery Wendy L. Patterson, M.P.H.*, Brittany D. Peoples, M.S., Foster C. Gesten, M.D. Office of Quality and Patient Safety, New York State Department of Health, Albany, NY Received September 23, 2014; accepted December 17, 2014
Abstract
Background: Using hospital readmissions as a quality of care measure predicates that some readmissions were preventable. Objectives: This study identifies predictors of potentially preventable readmissions (PPR) within 30 days of bariatric surgery discharge. Setting: New York State acute care hospitals. Methods: Adult inpatient surgical discharges, during 2012, with a principal diagnosis of overweight or obesity and a principal procedure for bariatric surgery were identified. Logistic regression was used to evaluate surgical approach, sex, age, race/ethnicity, payor, body mass index, complications and co-morbidities recorded during the surgical admission. Results: There were 10,448 surgeries studied for readmission of which 552 were followed by a PPR, for a statewide rate of 5.3 per 100 surgeries. Laparoscopic Roux-en-Y Gastric Bypass (LRYGB) was the most common surgical approach (46.0%), then Sleeve Gastrectomy (SG) (41.3%), Laparoscopic Adjustable Gastric Band (LAGB) (8.1%), and Open Roux-en-Y Gastric Bypass (RYGB) (4.6%). RYGB had the highest PPR rate (8.8), followed by LRYGB (6.1), SG (4.3) and LAGB (3.3). Compared to LAGB, the odds of a PPR in patients with RYGB, LRYGB, and SG increased by 2.4 fold, 1.8 fold and 1.2 fold respectively. Black, non-Hispanic patients were at a greater risk of PPR (odds-ratio 2.0, P o .0001) compared to White, non-Hispanic patients while the risk of a PPR increased by 2-fold in patients with a surgical complication. Conclusions: Taking all patient risk factors into account, the most significant predictors of a PPR were surgical approach, race and the presence of a surgical complication. (Surg Obes Relat Dis 2015;11:866–873.) r 2015 American Society for Metabolic and Bariatric Surgery. All rights reserved.
Keywords:
Bariatric surgery; Gastric bypass; Laparoscopy; Sleeve resection; Co-morbidity; Risk factors; Logistic regression; Readmission; Preventable; PPR
While there are benefits of bariatric surgery, there are potential consequences as well. These include postoperative complications before discharge [1–8] and hospital readmissions [9–22]. As the total volume of bariatric surgery continues to increase, and the link between quality of inpatient care, inpatient costs and readmissions receives continued attention [19,23], it becomes increasingly * Correspondence: Wendy L. Patterson, New York State Department of Health, 878 Corning Tower, Empire State Plaza, Albany, NY 12237. Phone: þ1-518-486-9012. E-mail: [email protected]
important to better understand outcomes associated with this intervention. Previous studies have used any type of hospital readmission within a specified time frame [24]. However, all readmissions are not related to inadequate care and using them in the calculation of readmission rates may be misleading. The usefulness of readmissions as a measure of quality of care is predicated on the notion that the readmissions might have been prevented. It is essential that credible and consistent clinical criteria be used to define that subset of readmissions that were potentially preventable [25].
http://dx.doi.org/10.1016/j.soard.2014.12.019 1550-7289/r 2015 American Society for Metabolic and Bariatric Surgery. All rights reserved.
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873
The purpose of this study is to identify predictors of a potentially preventable readmission after bariatric surgery. This study is the first to evaluate readmissions after bariatric surgery using the Potentially Preventable Readmission (PPR) Classification System Software developed by 3 M Health Information Systems [26]. This algorithm uses precisely defined clinical criteria to identify those readmissions that are clinically related to previous hospitalizations and thus may have been preventable. In previous research, Saunders et al. [10,27] distinguished among the various bariatric surgical approaches and compared readmission rates between approaches. Lindsey et al. [8] found that complication rates varied dramatically according to the bariatric surgical approach. In this paper, we will explore potentially preventable readmission rates by surgical approach and develop a multivariate logistic model to calculate the odds of a potentially preventable readmission based on surgical approach and patient risk factors, including patient demographic characteristics, postoperative complications and co-morbidities. Methods Hospital inpatient discharge data from the New York State Statewide Planning and Research Cooperative System (SPARCS) were used in this study. SPARCS collects patient level detail regarding patient demographic characteristics, diagnoses, procedures, discharge status, and charges for every inpatient discharge from New York State hospitals. The study population was limited to adult (age 18 or older) inpatient discharges between January 1, 2012 and December 31, 2012 from all hospitals in New York State with an International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) principal diagnosis code for overweight or obesity (278.00, 278.01, or 278.02) and an ICD-9-CM principal procedure code for bariatric surgery. Bariatric surgery was defined as Open Roux-en-Y Gastric Bypass (RYGB) (44.31, 44.39, 44.69), Laparoscopic Roux-en-Y Gastric Bypass (LRYGB) (44.38), Laparoscopic Adjustable Gastric Band (LAGB) (44.68, 44.95), and Sleeve Gastrectomy (SG) (43.82, 43.89). SPARCS data for all of 2012 were analyzed for bariatric surgery, with inpatient readmissions extending through January 31, 2013, to allow 30 days for readmission. The age and diagnostic restrictions ensured the bariatric surgery was for weight loss. The dependent variable was whether or not the bariatric surgical inpatient admission was followed by a potentially preventable inpatient admission within 30 days of discharge using the PPR software (Version 30.0 [3M, Salt Lake City, UT]) [25,26]. We used the 30 day timeframe because readmissions during that time are more likely to be related to the initial hospitalization and the transition of care to the outpatient setting. In addition, this timeframe is used in other
867
publicly reported measures from the National Quality Forum (NQF) and the Centers for Medicare and Medicaid Services (CMS). There were no restrictions that the readmission had to occur at the same hospital as the bariatric surgery. The PPR software first identifies globally excluded inpatient admissions and removes them from analysis. These include; most types of major metastatic malignancies, trauma, burns, many types of obstetric admissions and newborns, as well as patients whose treatment abruptly ended (the patient left against medical advice or the patient was transferred to another hospital). If suitable for analysis, the admission is classified as an initial admission, readmission, or only admission. The PPR software determines if an admission is clinically related to a prior admission and occurred in the readmission time interval. Clinically related means the underlying reason for admission is plausibly related to the care rendered during or immediately following the prior hospital admission and may have resulted from the process of care and treatment during the prior admission or from a lack of postadmission follow up. An initial admission is an inpatient admission followed by at least one clinically related inpatient admission within the readmission timeframe. A readmission is a clinically related inpatient admission that followed the initial inpatient admission within the readmission timeframe. An initial admission and all the readmissions associated with that initial admission form a PPR chain. If the admission was suitable to be followed by a readmission (not excluded) but was not, it is classified an only admission. The patient characteristics evaluated included gender, age, race/ethnicity, payor, and body mass index (BMI). Age was calculated at the time of admission and categorized into 4 groups: 18–29 years of age, 30–39 years of age, 40–49 years of age, and 50 years and older. Race/ethnicity was categorized into the following 4 groups: white, non-Hispanic; black, non-Hispanic; Hispanic; and Other. The Other group included Asian, Native Hawaiian/Pacific Islander, Native American, and those for whom race was recorded as other or unknown. Payor was identified by the source of payment listed on the surgical admission. Payors were split into 4 major groups: Medicaid, Medicare, Commercial (workers compensation, insurance company, self-pay, Blue Cross) and Other (Other federal program, CHAMPUS, other nonfederal program). BMI was identified by the diagnosis code recorded on the surgical admission and was grouped into 5 groups: BMI of 19.0–39.9 kg/m2, 40.0–44.9 kg/m2, 45.0– 49.9 kg/m2, 50.0 kg/m2 and over, and other (where BMI was either missing or miscoded as a pediatric code). In addition, we included whether or not the patient experienced a postoperative complication during the bariatric surgery admission [8]. Patient co-morbidities were defined using the approach developed by Elixhauser et al. [8,28]. We excluded a
868
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
number of these co-morbidities for the following reasons: those which defined the condition we were investigating (obesity and weight loss), those which had 6 or fewer patients that had a PPR (congestive heart failure, pulmonary circulation disease, peripheral vascular disease, paralysis, peptic ulcer disease, AIDS, lymphoma, metastatic cancer, solid tumor w/out metastasis, coagulopathy, chronic blood loss anemia, alcohol abuse, and drug abuse), and those that could be considered postoperative bariatric surgical complications (renal failure, fluid and electrolyte disorders, and deficiency anemias). Bivariate analyses were performed to investigate the relationship between each patient risk factor present at the time of surgery and whether or not the surgery was followed by a PPR for each of the surgical approaches. We used χ2 tests to examine the relationship between nonordered categorical variables (race/ethnicity, gender, the presence or absence of each of the risk factors) and the presence or absence of a PPR. For age, the CochraneArmitage test for trend was used to assess the existence of any linear trend in PPR frequency across the 4 age categories. Data for all 4 surgical approaches were pooled and a multivariate logistic regression model was constructed that contained indicators for all patient risk factors. While surgical approach may be influenced by the patient’s overall health status and surgical preference of the physician, creating some preselection of risk, we tried to control for much of this with our statistical model. This model permitted us to calculate the odds of a PPR following bariatric surgery associated with each of these patient risk factors. All analyses were conducted using SAS version 9.4 (Cary, NC). Results A total of 10,489 inpatient bariatric surgeries for obesity were performed in New York hospitals during 2012. After globally excluded admissions were removed, 10,448 discharges remained for analysis. The most common bariatric surgical approach was LRYGB, 4,807 (46.0%), while 4,319 surgeries (41.3%) were SG, 846 (8.1%) were LAGB, and 476 (4.6%) were RYGB. Overall, there were 552 bariatric surgeries that were followed by at least one PPR, for a statewide PPR rate of 5.3 per 100 discharges. The PPR rate was calculated by dividing the number of surgeries that were followed by at least one PPR by the total number of bariatric surgeries (552/10,448). These 552 surgeries were followed by 711 potentially preventable readmissions, an average of 1.3 readmissions per surgery among patients with a readmission. Of those patients experiencing a readmission, the vast majority, 69%, had only a single readmission, 20% had 2, and 11% had 3 or more readmissions. The most common primary diagnosis codes
for readmissions were ICD-9 539.89 (other complication after bariatric surgery, 16%), and ICD-9 276.51 (dehydration, 12%). Demographic characteristics and co-morbidities for the bariatric surgical population are displayed in Table 1. The population was predominantly female (79.2%); over 40 years of age (58.5%); white, non-Hispanic (53.4%); and covered by commercial insurance (84.4%). Nearly half (45.0%) of the patients had a BMI of 45 kg/m2 or greater and 5.4% of the patients had a complication during the surgical admission. The most common co-morbidities were hypertension (51.9%), diabetes without chronic complications (28.7%), chronic pulmonary disease (22.1%), and depression (19.4%). Table 1 also demonstrates substantial variation in PPR rates among the 4 surgical approaches. RYGB discharges had the highest PPR rate, 8.8 per 100 discharges, followed by LRYGB (6.1 per 100), SG (4.3 per 100), and LAGB (3.3 per 100). The table indicates there are differences among the 4 surgical approaches in terms of the patient characteristics and co-morbidities associated with PPRs. Black, non-Hispanic patients experienced a PPR more often when they had RYGB compared to any other surgical approach. LRYGB patients experiencing a complication had PPRs more often than other approaches. LRYGB patients with chronic pulmonary disease experienced a PPR more often than other surgical approaches. SG patients diagnosed with diabetes, with or without chronic complications, experienced a PPR more often than other patients. Compared to LAGB, the odds of a PPR in patients with RYGB and LRYGB increased by 2.4 fold and 1.8 fold respectively (Table 2). The odds of a PPR for Black, nonHispanic patients was almost 2-fold greater than White, non-Hispanic patients (odds ratio 2.0, P o .0001). Patients with a complication were at greater risk of a PPR than those without a complication (odds ratio 1.9, P o .0001). Patients with chronic pulmonary disease (odds ratio 1.5, P ¼ .0001), diabetes with (odds ratio 1.9, P ¼ .0290) and without (odds ratio 1.3, P ¼ .0118) chronic complications or rheumatoid arthritis/collagen disease (odds ratio 1.8, P ¼ .0304) had statistically higher odds of a PPR, compared to those patients without the condition at the time of bariatric surgery. Discussion Hospital readmissions are one potential indicator of the quality of inpatient care. Those readmissions that may have been prevented are of particular importance to healthcare organizations, clinicians, and payors given their link to reimbursement and comparative quality assessments. The major finding of this study was that the potentially preventable readmission rates varied significantly
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873
869
Table 1 Potentially preventable readmissions (PPR) by patient demographic characteristics and co-morbidities, by surgical approach Characteristic
Surgical approach RYGB
LRYGB
Total PPR % Surgical Approach Gender Female Male Age (Years) 18–29 30–39 40–49 50 and older Race / Ethnicity White, non-Hispanic Black, non-Hispanic Hispanic Other Payor Medicaid Medicare Commercial Other payor BMI BMI 19–39.9 kg/m2 BMI 40–44.9 kg/m2 BMI 45–49.9 kg/m2 BMI 50þ kg/m2 Other Complication Yes No Elixhauser Co-morbidities Valvular disease Yes No Hypertension Yes No Other neurologic disorders Yes No Chronic pulmonary disease Yes No Diabetes without chronic complications Yes No Diabetes with chronic complications Yes No Hypothyroidism Yes No Liver disease Yes No Rheumatoid arthritis/collagen disease Yes No
476
42
LAGB
P Value Total PPR %
8.82
4807 295
362 114
31 11
8.56 9.65
44 108 144 180
4 8 18 12
9.09 7.41 12.50 6.67
331 68 44 33
19 10 6 7
5.74 14.71 13.64 21.21
57 92 315 12
3 6 30 3
5.26 6.52 9.52 25.00
65 123 93 178 17
6 13 12 9 2
9.23 10.57 12.90 5.06 11.76
55 421
7 35
12.73 8.31
4 472
0 42
.00 8.90
286 190
26 16
9.09 8.42
7 469
1 41
14.29 8.74
111 365
13 29
11.71 7.95
3852 250 955 45
6.49 4.71
710 1260 1335 1502
6.06 6.11 7.12 5.33
8.04 7.27 5.79 8.91
.22 44 5.98 92 5.65 52 4.74 94 7.65 13 11.02
288 36 12.50 4519 259 5.73 .53
5.95 6.36
6.25 9.22
42 434
4 38
9.52 8.76
12 464
1 41
8.33 8.84
.84 2.36 3.98 4.46
8.85 5.35
475 133 115 123
20 6 2 0
4.21 4.51 1.74 .00
6.25 6.09
91 7 4716 288
7.69 6.11
583 35 4224 260
6.00 6.16
857 51 3950 244
5.95 6.18
.71
14 136 692 4
0 8 20 0
.00 5.88 2.89 .00
183 306 186 135 36
5 12 6 5 0
2.73 3.92 3.23 3.70 .00
26 820
1 27
3.85 3.29
18 828
2 26
11.11 3.14
4.84 3.79 4.51 4.36
1936 791 986 606
73 54 37 23
3.77 6.83 3.75 3.80
246 15 182 9 3860 162 31 1
6.10 4.95 4.20 3.23
854 1494 860 925 186
34 68 28 51 6
3.98 4.55 3.26 5.51 3.23
194 15 4125 172
7.73 4.17
28 0 4291 187
.00 4.36
406 440
13 15
3.20 3.41
2152 103 2167 84
4.79 3.88
10 836
0 28
.00 3.35
62 1 4257 186
1.61 4.37
172 674
10 18
5.81 2.67
937 50 3382 137
5.34 4.05
.71
o.01
.52
.16
.02
.06
.26
.87
.14
.56
.29
.04
.09
.24 221 625
10 18
4.52 2.88
5 841
0 28
.00 3.33
115 731
3 25
2.61 3.42
50 796
1 27
2.00 3.39
16 830
1 27
6.25 3.25
.01 1111 63 3208 124
5.67 3.87 o.01
.68 46 7 15.22 4273 180 4.21 .65
.80 438 20 3881 167
4.57 4.30
366 16 3953 171
4.37 4.33
68 6 4251 181
8.82 4.26
.59
.05 57 7 12.28 4750 288 6.06
35 44 55 53
.88
.80
.95
723 1162 1219 1215
P Value .75
.76
.89
.87
4.28 4.52
.28
.53
.44
3435 147 884 40
.08
.83 1472 92 3335 203
4.33
.22
o.01 1085 96 3722 199
.20
4 38
1 5 8 14
.05 91 10 10.99 4716 285 6.04
.22
64 412
119 212 201 314
.55 2573 153 2234 142
.61
6.25 8.91
3.51 2.73
.01 26 5 19.23 4781 290 6.07
.80
1 41
22 6
o.01
.28
16 460
626 220
.01 736 1629 1096 1228 118
4319 187 .57
.16 311 25 440 32 3955 229 101 9
P Value Total PPR %
3.31
o.01 2833 133 4.69 571 58 10.16 839 68 8.10 564 36 6.38
.13
10.82 7.45
28
.27 43 77 95 80
o.01
21 21
846 .04
.29
194 282
P Value Total PPR %
6.14
.72
SG
.97
.51
.07
870
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
Table 1 Continued. Characteristic
Surgical approach RYGB
LRYGB
Total PPR % Psychoses Yes No Depression Yes No
LAGB
P Value Total PPR % .19
17 459
0 42
.00 9.15
97 379
9 33
9.28 8.71
SG
P Value Total PPR % .16
107 10 4700 285
9.35 6.06
1115 72 3692 223
6.46 6.04
.86
P Value Total PPR % .14
8 838
1 27
151 695
6 22
12.50 3.22
P Value .41
81 5 4238 182
6.17 4.29
668 36 3651 151
5.39 4.14
.61
.14 3.97 .61 3.17
PPR ¼ Potentially preventable readmissions; RYGB ¼ Open Roux-en-Y gastric bypass; LRYGB ¼ Laparoscopic Roux-en-Y gastric bypass; LAGB ¼ Laparoscopic adjustable gastric band; SG ¼ Sleeve gastrectomy; BMI ¼ Body mass index.
depending on the bariatric surgical approach employed, even after accounting for patient differences. This was determined using a patient risk-based algorithm to identify only those 30 day hospital readmissions that were clinically related to the patient’s bariatric surgery. We found that 5.3% of the bariatric surgeries included in this study had a potentially preventable readmission within 30 days that could be related to the patient’s inpatient and/ or postdischarge outpatient care. In comparison, using the same data, 5.9% of the bariatric surgeries had any type of readmission within 30 days. Weller et al. [13], using 2003 New York hospital discharge data, found a bariatric all cause readmission rate of 7.6 per hundred. Since not all readmissions are potentially preventable, we would expect that the PPR rate would be less than these overall readmission rates. Conversely, a comparison of these rates suggests that many readmissions after bariatric surgery may be potentially preventable. To the best of our knowledge, there are no published studies describing PPR rates for bariatric surgical procedures, so there is no direct comparison available. We found significant variation in the crude PPR rates among the 4 surgical procedures. Saunders et al. found a 30-day all-cause bariatric surgery readmission rate of 7.3 per hundred for LRYGB and 3.1 per hundred for LAGB [10]. Similarly, Hutter found the highest 30-day readmission rate for LRYGB (6.5%), and the lowest for LAGB (1.7%), with SG in the middle at 5.4% [29]. Our method, which uses potentially preventable readmissions, found the same pattern in rates by surgical approach. It would be expected that LAGB, a minimally invasive procedure would result in a lower PPR rate than RYGB, a more invasive approach. Due to its complexity, this could explain the small number of RYGB procedures in New York State. It may not always be appropriate to select a procedure that will be less effective for a patient based solely on the procedures readmission rate. Many factors play a role in determining the type of surgical approach best for the patient (patient and surgeon preference, amount of long
term weight loss, intensity of postsurgical care, complications). Readmission rates should be considered as an additional factor when determining the potential risks and benefits of each surgical approach and deciding which surgical approach is best for a patient. Our results indicating that gender did not impact the odds of a PPR were similar to those of Zingmond et al. [11], but opposite of Weller et al.’s [12] analysis that found males at a greater risk for readmission. Baker et al. [16] found that patients with a BMI o40 kg/m2 were more likely to experience a hospital readmission compared to patients with a higher BMI, while our results indicated that BMI at any level was not significantly associated with the risk of PPRs. Weller et al. [12] found hypothyroidism and peptic ulcer disease increased the odds of subsequent hospital readmission. In our study, chronic pulmonary disease, diabetes with and without chronic complications and rheumatoid arthritis/collagen disease were significantly related to readmission. This study shares many of the limitations inherent in administrative data based studies, including incomplete or inaccurate diagnostic coding. In addition, previous work [12,14] has illustrated a significant relationship between the hospital and/or surgical volume of bariatric surgery and subsequent hospital readmissions. Unfortunately, a single year of data did not provide enough cases to perform a volume analysis by surgical approach. Despite these limitations, the findings of this study indicate that administrative data can be used to identify potentially preventable hospital readmissions after bariatric surgery, and that a significant number of potentially preventable readmissions occurred after bariatric surgery in New York State. In addition, these findings show that the amount of potentially preventable readmissions varies by the bariatric surgical approach employed, and that when all patient risk factors are taken into account, surgical approach, the presence of a complication during the initial admission and specific co-morbidities are significant predictors of a potentially preventable hospital readmission.
Predicting Preventable Hospital Readmissions / Surgery for Obesity and Related Diseases 11 (2015) 866–873 Table 2 Logistic regression results of predicting a potentially preventable readmission Risk factor Surgical approach RYGB LRYGB LAGB SG Gender Male Female Age (Years) 18–29 30–39 40–49 50 and older Race/Ethnicity White, non-Hispanic Black, non-Hispanic Hispanic Other Payor Medicaid Medicare Commercial Other payor BMI BMI 19–39.9 kg/m2 BMI 40–44.9 kg/m2 BMI 45–49.9 kg/m2 BMI 50þ kg/m2 Other Complication Yes No Co-morbidities Valvular disease Hypertension Other neurologic disorders Chronic pulmonary disease Diabetes without chronic complications Diabetes with chronic complications Hypothyroidism Liver disease Rheumatoid arthritis/collagen disease Psychoses Depression
Adjusted odds ratio (95% CI)
P Value
2.39 (1.45–3.96) 1.79 (1.20–2.68) 1.00 1.21 (.80–1.83)
.00 .00
871
illustrated that hospital readmissions vary by surgical approach as well.
Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article.
.36
References .93 (.74–1.18) 1.00
.56
1.10 (.81–1.51) 1.00 (.77–1.29) 1.25 (.99–1.57) 1.00
.53 .99 .06
1.00 2.03 (1.61–2.57) 1.39 (1.09–1.77) 1.15 (.86–1.54) .69 (.36–1.33) .61 (.32–1.16) .59 (.33–1.07) 1.00 1.00 1.07 .87 1.20 1.20
(.83–1.40) (.64–1.17) (.91–1.58) (.73–1.98)
1.91 (1.43–2.55) 1.00
o.01 .01 .34 .26 .13 .08
.59 .34 .21 .47 o.01
2.01 .94 1.25 1.45 1.29
(.90–4.49) (.78–1.14) (.68–2.30) (1.20–1.76) (1.06–1.58)
.09 .55 .47 .00 .01
1.88 .98 .93 1.85
(1.07–3.31) (.74–1.30) (.72–1.21) (1.06–3.21)
.03 .90 .60 .03
1.35 (.79–2.28) 1.22 (.98–1.51)
.27 .08
RYGB ¼ Open Roux-en-Y gastric bypass; LRYGB ¼ Laparoscopic Roux-en-Y gastric bypass; LAGB ¼ Laparoscopic adjustable gastric band; SG ¼ Sleeve gastrectomy; BMI ¼ Body mass index. C-statistic ¼ .64
Conclusion Two measurable adverse outcomes of bariatric surgery are postoperative complications and hospital readmissions. We established in our previous research that postoperative complications vary by surgical approach [8]. This study has
[1] Liu JH, Zingmond D, Etzioni DA, et al. Characterizing the performance and outcomes of obesity surgery in California. Am Surg 2003;69:823–8. [2] Livingston EH. Procedure incidence and in-hospital complication rates of bariatric surgery in the United States. Am J Surg 2004;188: 105–10. [3] Trus TL, Pope GD, Finlayson RG. National trends in utilization and outcomes of bariatric surgery. Surg Endosc 2005;19:616–20. [4] Pope GD, Birkmeyer JD, Finlayson RG. National trends in utilization and in-hospital outcomes of bariatric surgery. J Gastrointest Surg 2002;6:855–61. [5] Mason EE, Renquist KE, Jiang D. Perioperative risks and safety of surgery for severe obesity. Am J Clin Nutr 1992;55:5735–65. [6] Livingston EH, Langert J. The impact of age and Medicare status on Bariatric Surgical outcomes. Arch Surg 2006;141:1115–20. [7] Weller WE, Rosati C, Hannan EL. Predictors of in-hospital postoperative complications among adults undergoing bariatric procedures in New York State, 2003. Obesity Surgery 2006;16:702–8. [8] Lindsey ML, Patterson WL, Gesten FC, Roohan PJ. Bariatric Surgery for Obesity: Surgical Approach and Variation in In-Hospital Complications in New York State. Obesity Surgery 2009;19:688–700. [9] Hong B, Stanley E, Reinhardt S, et al. Factors Associated with Readmission after Laparoscopic gastric bypass surgery. Surg Obes Relat Dis 2012;8:691–5. [10] Saunders J, Ballantyne G, Belsley S, et al. 30 day readmission rates at a high volume bariatric surgery center: Laparoscopic adjustable gastric banding, laparoscopic gastric bypass and vertical banded gastroplasty Roux-en-Y gastric bypass. Obesity Surgery 2007;17: 1171–7. [11] Zingmond DS, McGory ML, Ko CY. Hospitalization before and after gastric bypass surgery. JAMA 2005;294:1918–24. [12] Weller WE, Rosati C, Hannan EL. Relationship between surgeon and hospital volume and readmission after bariatric operation. J Am Coll Surg 2007;204:383–91. [13] Finkelstein EA, Brown DS. A Cost-Benefit Simulation Model of Coverage for Bariatric Surgery among Full-Time Employees. Am J Manag Care 2005;11:641–6. [14] Nguyen NT, Paya M, Stevens CM, et al. The Relationship between Hospital Volume and Outcome in Bariatric Surgery at Academic Medical Centers. Ann Surg 2004;240:586–94. [15] McCarty TM, Arnold DT, Lamont JP, et al. Optimizing Outcomes in Bariatric Surgery: Outpatient Laparoscopic Gastric Bypass. Ann Surg 2005;242:494–8. [16] Baker MT, Lara MD, Larson CJ, et al. Length of Stay and Impact on Readmission Rates After Laparoscopic Gastric Bypass. Surg Obes Relat Dis 2006;2:435–9. [17] Bradley DW, Sharma BK. Centers of Excellence in Bariatric Surgery: Design, Implementation, and One-Year Outcomes. Surg Obes Relat Dis 2006;2:513–7. [18] Escinosa WE, Bernard DM, Chen C, et al. Healthcare Utilization and Outcomes After Bariatric Surgery. Medical Care 2006;44:706–12.
872
W. L. Patterson et al. / Surgery for Obesity and Related Diseases 11 (2015) 866–873
[19] Ashton CM, DelJunco DJ, Souchek J, et al. The Association Between the Quality of Inpatient Care and Early Readmission: A MetaAnalysis of the Evidence. Medical Care 1997;35:1044–5. [20] Carlin AM, Zeni TM, English WJ, et al. The Comparative Effectiveness of Sleeve Gastrectomy, Gastric Bypass, and Adjustable Gastric Banding Procedures for the Treatment of Morbid Obesity. Ann Surg 2013;257:791–7. [21] Kwon S, Wang B, Wong E, et al. The Impact of Accreditation on Safety and Cost of Bariatric Surgery. Surg Obes Relat Dis 2013;9:617–22. [22] Nguyen NT, Nguyen B, Nguyen VQ, et al. Outcomes of Bariatric Surgery Performed at Accredited vs. Nonaccredited Centers. J Am Coll Surg 2012;215:467–74. [23] Friedman B, Basu J. The Rate and Cost of Hospital Readmissions for Preventable Conditions. Med Care Res and Rev 2004;61:225–40. [24] Dorman RB, Miller CJ, Leslie DB, et al. Risk for Hospital Readmission following Bariatric Surgery. PLoS ONE 2012;7:e32506.
[25] Goldfield NI, McCullough EC, Hughes JS, et al. Identifying Potentially Preventable Readmissions. Health Care Financing Review 2008;30:75–91. [26] 3 MTM Health Information Systems. Potentially Preventable Readmissions Classification System: Methodology Overview. Document number GRP-139 05/08. May 2008. [27] Saunders J, Ballantyne GH, Belsley B, et al. One year readmission rates at a high volume bariatric surgery center: Laparoscopic adjustable gastric banding, laparoscopic gastric bypass and vertical banded gastroplasty Roux-en-Y gastric bypass. Obesity Surgery 2008;18: 1233–40. [28] Elixhauser A, Syeiner C, Harris D. Co-morbidity measures for use with Administrative Data. Med Care 1998;36:8–27. [29] Hutter MM, Schirmer BD, Jones DB, et al. First Report from the American College of Surgeons Bariatric Surgery Center Network: Laparoscopic Sleeve Gastrectomy has Morbidity and Effectiveness Positioned Between the Band and the Bypass. Ann Surg 2011;254: 410–22.
Editorial
Comment on: Predicting potentially preventable hospital readmissions following bariatric surgery Decreasing the rate of readmissions in bariatric surgery has the potential to improve the delivery of patient care, increase patient satisfaction, and allow for a lower overall cost of care. There are numerous hurdles to truly impact readmission rates. Perhaps the first is to understand the true number of readmissions, and more importantly, those that can be prevented. The authors should be commended in their analysis presented in “Predicting Potentially Preventable Hospital Readmissions Following Bariatric Surgery,” which helps us get closer to understanding that true rate [1]. Additionally, they are carefully responsible with their consistent focus on potentially preventable readmissions as opposed to examining only the total readmission rate. In our program, when we implemented a process to more aggressively capture and understand bariatric readmissions, we witnessed an observer effect with an increase in the number of readmissions in the early months of our analysis. The plethora of definitions of what constitutes a readmission unfortunately confounds the study of this outcome variable. Standardizing the definition of readmissions in the broadest sense will help to identify the true baseline rate. Capturing and classifying all return visits to the hospital—whether an ER visit, 23 hour observation, or actual multiple-midnight stay readmission, as well as capturing patients that present to another hospital is critical to establishing this baseline. In the past, we observed programs that claimed extremely low readmission rates when in fact the readmissions were being admitted to another local hospital or a larger regional tertiary care facility and not being captured as part of the original surgery program data.
No matter what approach we use, there must be a reasonable attempt by CMS and other payors to determine what are acceptable preventable and total readmission rates [2]. At this point in time, an “acceptable” rate might in fact increase as the entire bariatric community improves its capture of readmission data. We must also be conscious of rewarding a reduction of readmissions. Physician compensation methodologies will all drive certain behaviors, potentially resulting in some unintentional consequences. Could either penalizing readmissions or rewarding a reduction of readmissions actually encourage a healthcare system to prevent a recent postoperative bariatric patient from coming to the hospital when that is indeed the best treatment? I would like to think not, but this is indeed within the realm of possibility. An alternative to penalizing a readmission could be to reward the behaviors that, when employed, can reduce preventable readmissions. One possibility is to incentivize physician, bariatric program, and healthcare system behaviors that allow for the investigation of the root cause of each readmission with a goal of reviewing 100% of readmissions. In this manner, the program can classify the readmission as potentially preventable or nonpreventable. If preventable, the program can then determine if there were any missed opportunities to do better. This multidisciplinary approach should be relatively easy to undertake within the constructs of an MBSAQIP-accredited center of excellence program in bariatric surgery. Readmissions will only be able to be affected through the development of standardized criteria for the definition of
