Changes in Utilization of Bariatric Surgery in the United States From 1993 to 2016

Guilherme M. Campos, MD, PhD; Jad Khoraki, MD; Matthew G. Browning, PhD; Bernardo M. Pessoa, MD; Guilherme S. Mazzini, MD, PhD; Luke Wolfe, MS


Annals of Surgery. 2020;271(2):201-209. 

In This Article


Estimates of Annual Bariatric Surgeries

For the analyses and interpretation of the National Inpatient Sample (NIS) dataset, we followed the most updated available recommendations.[18] We identified hospital admissions during which patients underwent primary bariatric surgery from the NIS from 1993 to 2016. The Healthcare Cost and Utilization Project (HCUP) through the NIS provides information from a stratified sample of approximately 20% of discharges from a variety of community hospital types and sizes, ranging from large academic institutions to smaller privately owned facilities across all geographical regions. The HCUP-NIS records include information related to patient demographics, insurance provider, diagnoses, and procedure(s) during that admission. Diagnoses and procedures are coded according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) from 1993 through the third quarter of 2015 and according to the Tenth Revision (ICD-10-CM) codes thereafter. Participation in HCUP-NIS expanded from 913 hospitals across 17 states in 1993 to 4575 hospitals across 46 States and the District of Columbia in 2016 (encompassing 97% of the population), making it the largest nationally representative all-payer inpatient care database in the United States.[19]

Sampling probabilities were calculated to select all discharges from 20% of participating hospitals within each stratum from 1993 through 2011. Starting in 2012, the NIS includes a 20% sample of all database discharges. Total discharges were divided by the number of sampled discharges within each HCUP-NIS stratum to derive discharge weights for nationwide estimates.

Beginning in 1998, discharges were calculated as the sum of admissions and births for each community hospital, excluding rehabilitation and long-term acute care hospitals that were included before this redesign. To account for these changes, the HCUP-NIS recalculated the discharge weights to create national estimates for trends analyses that are consistent with 2012 data and onward.[19] Approval for the use of the NIS data was obtained from the HCUP and confidentiality instructions were followed as directed by the HCUP.

Admission Selection and Surgery Codes

ICD-9-CM and ICD-10-CM codes used are presented in Table 1. Admission records were extracted for patients aged 18 years and older, with a principal diagnosis code for obesity who underwent primary bariatric surgery from 1993 through 2016. We excluded revisional procedures to maximize the likelihood that individual admissions represent distinct patients. Patients with diagnosis codes for noninfective enteritis, colitis, or abdominal neoplasms were excluded. Primary procedures included open or laparoscopic Roux-en-Y gastric bypass (RYGB), vertical banded gastroplasty, adjustable gastric band (AGB), biliopancreatic diversion/duodenal switch (BPD-DS), and sleeve gastrectomy (SG). The specific ICD-9 procedure code for laparoscopic RYGB (44.38) and laparoscopic SG (43.82) only became available in 2004 and 2011, respectively. Therefore, open procedure codes were initially used for laparoscopic procedures. Consequently, exact numbers of open or laparoscopic techniques cannot be determined during years when a specific code was unavailable.

Patient and Hospital Characteristics

Age, sex, race/ethnicity (white, black, Hispanic, or other), obesity-related comorbidities, and insurance (commercial, Medicare, Medicaid, self-pay, or other) were extracted. Hypertension and T2D diagnoses were captured using the HCUP Elixhauser Comorbidity Software Version 3.7. In addition, we created a variable for obstructive sleep apnea (absent from Elixhauser), identified by ICD-9-CM codes 327.23, 780.51, 780.53, 780.57, and 327.2 and ICD-10-CM codes G47.30, G47.31, G47.33, G47.37, and G47.39. Hospital descriptors included location and teaching status (rural, urban nonteaching, urban teaching), size (small, medium, large), and geographic region (Northeast, Midwest, South, West).

Safety Measures

Safety outcomes of bariatric surgery admissions included occurrence and type of complications, hospital length of stay (LOS), and mortality rates. Complications were identified based on specific ICD-9-CM and ICD-10-CM diagnosis codes across a broad range of categories (Table 1). Similar to other publications evaluating complications of bariatric surgery using the NIS dataset,[20,21] we chose to exclude complications that might be considered minor (such as atelectasis), which have less of an impact on patients than major complications. The codes chosen also differentiate in-hospital complications (eg, pulmonary embolism/sepsis) from preexisting comorbidities that would contraindicate primary bariatric surgery.


Total cost of admissions in US dollars for the years 2001 to 2016 was calculated from each hospitalization total charge multiplied by the distinct cost-to-charge ratio for each hospital that the NIS has provided beginning in 2001. This was adjusted to 2016 US dollar value to account for inflation using the medical component of the annual Consumer Price Index.[22]

Estimates of Eligible Population and Utilization of Bariatric Surgery

Data from the National Health and Nutrition Examination Survey (NHANES) were used to estimate numbers of adults possibly eligible for bariatric surgery. Information regarding the NHANES participants, study design, and outcome measures is available elsewhere.[23,24] Bariatric surgery eligibility criteria were in accordance with the 1991 National Institutes of Health (NIH) guidelines.[16] Specifically, we considered NHANES participants aged 18 years or older with either class III obesity (BMI ≥40 kg/m2) or class II obesity (BMI 35–39.9 kg/m2) with hypertension, hypercholesterolemia, and/or T2D as eligible for surgery. Hypertension was defined as mean systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, and/or self-reported antihypertensive medication requirement.[25] Hypercholesterolemia was defined as total cholesterol ≥200 mg/dL and/or self-reported lipid-lowering medication requirement.[26] Diabetes was defined as a glycated hemoglobin (HbA1c) ≥6.5%, fasting plasma glucose (PG) ≥126 mg/dL, 2-hour PG ≥200 mg/dL during an oral glucose tolerance test, and/or self-reported antihyperglycemic medication requirement (including insulin).[27]

Utilization of bariatric surgery was calculated as the number of bariatric surgeries from the HCUP-NIS database divided by the number of eligible adults from NHANES. These values were then multiplied by 100 and 100,000 to yield utilization estimates as a percentage of the total eligible population and per 100,000 adults, respectively. We used the NHANES III datasets to estimate the average number of eligible adults from 1988 through 1994 and calculated a single utilization value for the year 1993. When NHANES resumed in 1999, a new strategy was adopted using continuous bi-yearly survey cycles that allowed us to estimate the number of eligible adults in 2000 (eg, using 1999–2000 cycle data) and then every other year thereafter through 2016.

Statistical Analysis

Statistical analyses were performed using the survey procedures in SAS V9.4 (SAS Institute, Cary, NC) to account for stratification and clustering in sampling designs of the HCUP-NIS and NHANES surveys. Select measures were evaluated for possible significant changes over time. Categorical variables are presented as the number (N) and percentage (%) and were compared using the Rao-Scott chi-square test in the SURVEYFREQ procedure. Continuous variables are presented as the mean and standard errors of the mean (SEM), except for cost data which are presented as medians (interquartile [IQ] ranges), and were compared using the SURVEYREG procedure. All reported P values are two-sided, and those below 0.05 were considered statistically significant.