Regional Variations in Gastric Bypass Surgery: Results from the 2005 Nationwide Inpatient Sample
- First Online:
- Cite this article as:
- Weller, W.E. & Rosati, C. OBES SURG (2008) 18: 1225. doi:10.1007/s11695-008-9524-9
- 60 Views
The purpose of this study is to use nationally representative data to examine regional variations in the use and outcomes of gastric bypass surgery.
Using data from the Nationwide Inpatient Sample (NIS), we identified adults undergoing gastric bypass surgery (n = 92,910) in 2005. Following descriptive analyses, multiple logistic regression models were constructed to examine regional variations in the likelihood of laparoscopic vs. open approaches and in the likelihood of complications while controlling for patient and hospital characteristics.
After indirectly adjusting for age and sex, the gastric bypass rates per 100,000 were as follows: Northeast, 70; Midwest, 39; South, 37; and West, 61. After adjusting for both patient and hospital characteristics, the odds of receiving laparoscopic surgery for patients living in the West were 1.79 times the mean [95% confidence interval (CI): 1.67–1.92], while the odds of receiving laparoscopic surgery for patients in the Midwest were 0.66 of the mean (95% CI: 0.62–0.70) and those of the Northeast were 0.88 of the mean (95% CI: 0.83–0.94). When adjusting for both patient and hospital characteristics, the odds of one or more postoperative complications among patients living in the South were greater than the mean (OR: 1.14, 95% CI: 1.02–1.26).
Findings from this study suggest that gastric bypass surgery is more common in the Northeast and West. There is a greater likelihood of gastric bypass being performed laparoscopically in the West; it is less likely to be performed in the Northeast and Midwest. Postoperative complications are more likely to occur in the South.
KeywordsGastric bypassRegional variationsNationwide inpatient sample
Bariatric surgery has been shown to be an effective treatment for persons with a body mass index (BMI) ≥40 kg/m2 or persons with a BMI ≥35 kg/m2 who have a significant comorbidity, such as hypertension or diabetes . The number of bariatric surgeries has grown rapidly in recent years, increasing from 13,365 in 1998 to a projected 102,794 in 2003 . Likewise, membership in the American Society for Metabolic and Bariatric Surgery increased from 258 to 631 between 1998 and 2002 .
As the number of bariatric surgeries has grown, there has been a concomitant growth in the number of bariatric surgery studies based on administrative data. These studies have largely examined trends in the use of bariatric surgery [2, 3] and outcomes of bariatric surgery, including mortality [4–7]. Despite a wide body of literature demonstrating geographic variations in the utilization and outcomes of healthcare services, including a variety of surgical techniques [8–14], few studies have focused on bariatric surgery. A review of the literature reveals only one study that examined regional variations in morbid obesity and bariatric surgery use .
The purpose of this study is to use nationally representative data to examine regional variations in the use and outcomes of gastric bypass surgery. Specifically, this study examined regional differences in (1) the tendency to perform gastric bypass surgery, (2) the likelihood to perform laparoscopic versus open gastric bypass surgery, and (3) the likelihood of complications.
Data Source and Study Population
This is a cross-sectional study based on data from the 2005 Nationwide Inpatient Sample (NIS) from the Healthcare Cost and Utilization Project (HCUP) . The NIS is the largest all-payer inpatient database in the USA, containing information from approximately 8 million hospital stays per year from a sample of approximately 1,000 hospitals in 37 US states. The NIS represents a 20% stratified sample of inpatient admissions to acute care hospitals nationwide. Stratifications are based on hospital characteristics including the following: ownership/control, bed size, teaching status, urban/rural location, and US census region. Sampling weights are provided to account for the complex survey design and to provide national estimates. The NIS contains data such as demographics (age, gender, race, median income for zip code), principal and secondary diagnoses, expected payment source, and hospital characteristics.
The study population consisted of adults (18 years or older) who had a principal diagnosis code for obesity (ICD-9-CM codes 278.0, 278.00, 278.01, 278.1) and a primary procedure code for open gastric bypass (ICD-9-CM codes 44.31, 44.39) or laparoscopic gastric bypass (ICD-9-CM code 44.38). The final study population consisted of 19,052 discharges (weighted n = 92,910).
Gastric Bypass Rates
The age/sex adjusted rate of gastric bypass per 100,000 persons was determined for each hospital region using the indirect method. Population data based on age and sex were drawn from 2005 census estimates based on the 2000 US Census. Regions in the NIS are based on US Census Bureau definitions and included the following states in 2005: Northeast (CT, MA, NH, RI, NJ, NY); Midwest (IN, IL, MI, OH, WI, IA, KS, MN, MO, NE, SD); South (FL, GA, MD, NC, SC, WV, KY, TN, AK); and West (CO, UT, NV, CA, HI, OR, WA).
Laparoscopic Procedure by Region
Preliminary analyses using chi-square tests were used to test for differences in the percentage of gastric bypass patients undergoing laparoscopic surgery by region. Two logistic regression models were then constructed. The first model examined differences in the use of laparoscopic versus open gastric bypass surgery by region after controlling for patient characteristics. The second model included both patient and hospital characteristics in order to determine the impact of differences in hospital characteristics on regional differences. The major dependent variable in the models was laparoscopic surgery status, with open procedure as the reference group. The major independent variable in the model was hospital region, with effects coding used to compare each region to the mean. Patient characteristics included age, gender, race, household income, comorbid conditions, and primary payer. After examining the age distribution of the sample, age was categorized as 18 to 29 year, 30 to 39 year, 40 to 49 years, and 50 years and older. Race/ethnicity was categorized as white, black, other, and missing. A separate category for “missing” was created for race due to the large percentage of missing race data. Household income represented the median household income of the patient’s ZIP code and was categorized into quartiles ($1–$35,999; $36,000–$44,999; $45,000–$58,999; ≥$59,000). Payer represented the expected primary payer and was categorized as Medicare, Medicaid, private, and other. The “other” category included self-paying patients, patients with no charge, and those categorized as “other.” Comorbidity was measured using the approach developed by Elixhauser et al. . We excluded several of the 30 comorbidities because they were the focus of the study (obesity, weight loss) or could be the result of surgery rather than a condition existing prior to surgery (renal failure, anemia from blood loss, deficiency anemias, fluid and electrolyte disorders) or had total unweighted cell sizes ≤20 (paralysis, peptic ulcer disease, AIDS, lymphoma, metastatic cancer, coagulopathy, alcohol abuse, drug abuse). Hospital characteristics included hospital gastric bypass volume (categorized into quartiles; ≤92, 93–179, 180–348, ≥348); teaching status (teaching vs. non-teaching); location (urban vs. rural); and bed size (small, medium, large). Hospital volumes were determined by summing the number of times a gastric bypass was performed in a given hospital.
Postoperative Complications by Region
Preliminary analyses using chi-square tests were used to test for differences in the percentage of gastric bypass patients with one or more in-hospital postoperative complication by region. Two logistic regression models were then constructed. The first model examined differences in the likelihood of complications by region after controlling for patient characteristics. The second model included both patient and hospital characteristics in order to determine the impact of differences in hospital characteristics on regional differences. The major dependent variable in the models was one or more complication. The major independent variable in the model was hospital region, with effects coding used to compare each region to the mean. Patient and hospital characteristics used in the previous model to assess laparoscopy were also used in the models as well as a variable indicating if the surgery was laparoscopic or open. Postoperative complications occurring in the hospital included those based on earlier studies [18, 19]. Complications were based on ICD-9 codes and included infections, gastrointestinal, pulmonary, cardiovascular, and complications during procedure. Complications and their associated diagnosis coded are listed in Appendix. Deaths were not examined separately due to the small number of patients who died in hospital (n = 29). Rather, they were included as a postoperative complication.
All analyses were conducted using SAS Version 9.1 using the PROC SURVEY method to account for the complex sampling design of the NIS. All analyses were performed on the raw survey data and then reanalyzed using a weighting variable to reflect national population estimate. Unless otherwise specified, results reflect weighted data.
There was a tendency to do more gastric bypass surgery in the Northeast and the West compared to the South and the Midwest. After indirectly adjusting for age and sex, the gastric bypass rates per 100,000 in each region were as follows: Northeast, 70; Midwest, 39; South, 37; and West, 61.
Regional variation in gastric bypass surgery by patient and hospital characteristics
Northeast (n = 22,739)
Midwest (n = 16,894)
South (n = 27,010)
West (n = 26,267)
Congestive heart failure
Pulmonary circulation disorders
Other neurological disorders
Chronic obstructive pulmonary disease
Solid tumor, no metastasis
25th quartile (≤92)
50th quartile (93–179)
75th quartile (180–348)
100th quartile (≥348)
Hospital bed size
Comorbidity also varied by region. For six of the 14 comorbidities, the percentage of patients with those comorbidities was higher in the South than the other regions (congestive heart failure, cardiac arrhythmia, valvular disorders, hypertension, hypertension complicated, and diabetes). Patients in the Midwest and South also were also noticeably more likely to have depression than patient in the Northeast or the West (27.47% and 24.08% vs. 17.00% and 16.54%, respectively, p <0 .001). Other noticeable differences among comorbidities included chronic obstructive pulmonary disease (COPD) and liver disease. Patients in the Northeast were more likely to have COPD than patients from other regions (20.00%, p < 0.001) while patients from the Northeast and West were more likely to have liver disease (10.78% and 10.16%, respectively) than patients in the Midwest and South (5.66% and 5.47%, respectively; p < 0.001). Gastric bypass patients in the West were more likely to be operated on in high volume hospitals than patients from other regions (37.51%, p < 0.001) but were less likely to have their surgery be performed in a teaching hospital (36.65%, p < .001). Patients in the West were also less likely than patients from other regions to have their surgery performed in a small hospital (p < 0.001).
Regional variations in laparoscopic procedures performed
Number of laparoscopic gastric bypass patients
Percentage of all gastric bypass patients*
Multivariable odds of undergoing laparoscopic gastric bypass after adjusting for patient characteristics†
Multivariable odds of undergoing laparoscopic gastric bypass after adjusting for patient and hospital characteristics†
Several of the hospital and patient characteristics were significant predictors of the use of laparoscopic surgery in the fully specified model. Each of the hospital characteristics was significant except for bed size. The odds of laparoscopy were higher for patients in higher volume hospitals compared to low volume hospitals. Surgeons at urban hospitals were significantly more likely to perform laparoscopic gastric bypass than surgeons from rural hospitals, whereas those in teaching hospitals were less likely to perform laparoscopic procedures. Among patient characteristics, the odds of receiving laparoscopic surgery were greater for females than males and for individuals with higher incomes. Patients without private health insurance were less likely to undergo laparoscopic surgery. Likewise, those with CHF or liver disease were less likely to undergo a laparoscopic procedure, whereas those with depression were more likely than those without depression.
Regional variations in postoperative complications
Number with ≥ 1 complication
Percentage with ≥ 1 complication*
Multivariable odds of ≥ 1 complication after adjusting for patient characteristics†
Multivariable odds of ≥ 1 complication after adjusting for patient and hospital characteristics†
A number of hospital and patient characteristics were also significant in the fully specified model for complications. Significant hospital factors included hospital location (urban/rural), hospital volume, and hospital bed size. Patients in urban hospitals had lower odds of postoperative complications than patients in rural hospitals. The odds of complications decreased with increasing hospital volume. Both medium and large hospitals had lower odds of complications relative to small hospitals. Among patient characteristics, the likelihood of one or more complication increased with age, while the odds of one or more complication were lower for females relative to males. The odds of complications were higher for Medicare patients compared to private pay patients. Five of the comorbidities remained significant in the multivariable models. The odds of one or more complications were higher for individuals who had CHF, cardiac arrhythmia, or “other neurological conditions” compared to those who did not. Individuals with depression or uncomplicated hypertension, however, were less likely to experience one or more complication.
The number of bariatric procedures performed in the USA is increasing rapidly as the prevalence of obesity and morbid obesity continues to grow. The purpose of this study was to use nationally representative hospital discharge data to examine regional differences in the use and outcomes of gastric bypass surgery in 2005.
The results showed differences in the age/sex adjusted rates of gastric bypass surgery by region, with much higher rates in the Northeast and West (70 and 61 per 100,000, respectively) than the Midwest and South (39 and 37 per 100,000, respectively). It is possible that these rate differences reflect differences in the prevalence of morbid obesity among the four regions. However, at least one previous study suggests this is probably not the case. Poulose et al.  used NIS data linked to data from the Behavioral Risk Factor Surveillance System (BRFSS) to determine the rates of bariatric surgery use and the burden of morbid obesity by age, gender, and census region. Overall, they found that bariatric surgery rates do not parallel morbid obesity rates by region. Similar to this study, Poulose et al.  found that the Northeast and the West had the highest rates of bariatric surgery. The South, however, had the highest burden of morbid obesity. The largest regional disparities found were in the West, which had a high rate of bariatric surgery and low prevalence of morbid obesity, and the South, which had a low rate of bariatric surgery and high prevalence of morbid obesity . Other possible explanations for differences in bariatric surgery rates include differences in insurance coverage, patient/provider attitudes, and surgeon supply.
The results showed significant differences in who received gastric bypass by region. Patients from the West were older and were more likely to be men than other regions. Patients from the South were more likely to be socioeconomically disadvantaged, as measured by income and coverage by private insurance, compared to other regions. The burden of comorbidity was also greater for those living in the South compared to other regions. For six of the 14 comorbidities, the percentage with those comorbidities was significantly higher for those living in the South than those living in other regions. While patient characteristics tended to differ for those living in the South, hospital characteristics tended to differ for those living in the West. For example, gastric bypass patients in the West were more likely to be operated on in high volume hospitals, were less likely to have their surgery performed at a teaching hospital, and were more likely to have their surgery be performed in a small hospital than patients from other regions.
Increasingly, gastric bypass surgery is being performed laparoscopically. Across the USA in 2005, 74.06% of all gastric bypass surgeries were performed laparoscopically. However, the percentage of gastric bypasses performed laparoscopically varied significantly by region. Specifically, 85.46% of patients in the West underwent a laparoscopic procedure, 70.47% in the South, 69.72% in the Northeast, and 67.92% in the Midwest. Surgeons in the West remained much more likely to perform gastric bypass laparoscopically than surgeons in the other regions in multivariable models. After adjusting for a wide variety of patient and hospital characteristics, the odds of surgeons performing gastric bypass surgery laparoscopically were 1.79 times as high in the West compared to the mean. Multivariable analyses adjusting for patient and hospital characteristics also showed that the odds of surgeons performing gastric bypass laparoscopically were significantly lower in the Midwest and the Northeast compared to the mean. The higher use of laparoscopic surgery in the West may reflect early work conducted in the West in the 1990s and early twenty-first century which translated to more experience and preference for the laparoscopic approach [20, 21]. Differences may also reflect availability of surgical resources. It was not possible to take surgeon level factors into account using the NIS data. However, future studies should account for such factors, including surgeon volume, training, and the number of surgeons per capita. Hospital characteristics may also account for some of the differences between regions. When patient factors only were in the model, patients from the South were also significantly less likely to undergo laparoscopic surgery than the mean (in addition to the Northeast and Midwest). However, once hospital factors were added to multivariable models, the South was no longer significantly different from the mean because hospital characteristics associated with a lower use of laparoscopic surgery were more prevalent in the South.
Finally, we examined differences in postoperative complications by region. Across the USA in 2005, 6.40% of gastric bypass patients had one or more of the in hospital postoperative complications included in the study. Again, there were variations by region, with the Midwest and South (6.54% and 7.51%, respectively) demonstrating higher percentages of patients with one or more complications compared to patients in the Northeast and West (5.73% and 5.75%, respectively). After adjustment for patient and hospital characteristics, patients from the South were significantly more likely to experience one or more postoperative complication compared to the mean. Again, future studies should examine possible factors that may explain a higher likelihood of postoperative complications among patients in the South. It is possible that these patients are, on average, sicker than patients from other regions undergoing gastric bypass surgery. Although we adjusted for comorbidity in this study (and there was a higher rate of comorbidity in the South), it is possible that multivariable adjustment did not completely adjust for health status. For example, it is possible that patients in the South had higher BMIs than patients in other regions, and those patients with higher BMIs are at greater risk for postoperative complications. Unfortunately, BMI is not available in the NIS, so it was not possible to control for BMI in multivariate models. As with laparoscopic surgery, differences in hospital characteristics are likely to explain some of the findings. When patient characteristics only were included in the multivariable model, none of the regions significantly differed from the mean. When hospital characteristics were added to the model, patients from the South were significantly more likely to have one or more complications compared to the mean. One of the major hospital level factors included in the model was hospital volume. Consistent with previous studies, the likelihood of complication decreased as hospital volume increased. There was a somewhat higher percentage of patients in the South undergoing their surgery in higher volume hospitals. It is possible that once this difference was taken into account, significant differences between the South and the mean emerged.
There are several limitations to this study. Although population-based administrative data can be useful due to large sample sizes and, in this case, generalizability, they do have some limitations. As mentioned, although the analyses were adjusted for patient risk factors, some clinical information that has previously been shown to be associated with postoperative complications, such as BMI, is not available in the database. Likewise, surgeon level factors were not included in the database, so it was not possible to adjust for certain surgeon level factors, such as surgeon volume, that may be associated with outcomes such as postoperative complications. Accuracy and completeness of the coding are also potential concerns when using population-based data. Hospitals in different states could record different numbers of secondary diagnosis codes, which could affect the prevalence of comorbidities across the regions. However, analyses of the number of secondary diagnosis codes used per region showed that, on average, the Midwest and South reported only one more secondary diagnosis code than in the Northeast and West. The data also do not allow us to examine longer-term outcomes, such as complications occurring after hospital discharge and the ability to keep weight off over time.
Despite these limitations, this study adds to the bariatric surgery literature by comparing regional differences in the use and outcomes of gastric bypass surgery. Findings from this study suggest that gastric bypass surgery is more common in the Northeast and West. There is a greater likelihood of gastric bypass being performed laparoscopically in the West; it is less likely to be performed in the Northeast and Midwest. Postoperative complications are more likely to occur in the South. While hospital characteristics, such as hospital volume, may explain some of these differences, additional studies should be undertaken to further address regional differences.