Article Text
PDF
Abstract
Introduction As the rate of obesity increases, so does the incidence of obesity-related comorbidities. Metabolic and bariatric surgery (MBS) is the most effective treatment for obesity, yet this treatment is severely underused. MBS can improve, resolve, and prevent the development of obesity-related comorbidities; this improvement in health also results in lower healthcare costs. The studies that have examined these outcomes are often limited by small sample sizes, reliance on outdated data, inconsistent definitions of outcomes, and the use of simulated data. Using recent real-world data, we will identify characteristics of individuals who qualify for MBS but have not had MBS and address the gaps in knowledge around the impact of MBS on health outcomes and healthcare costs.
Methods and analysis Using a large US employer-based retrospective claims database (Merative), we will identify all obese adults (21+) who have had a primary MBS from 2016 to 2021 and compare their characteristics and outcomes with obese adults who did not have an MBS from 2016 to 2021. Baseline demographics, health outcomes, and costs will be examined in the year before the index date, remission and new-onset comorbidities, and healthcare costs will be examined at 1 and 3 years after the index date.
Ethics and dissemination As this was an observational study of deidentified patients in the Merative database, Institutional Review Board approval and consent were exempt (in accordance with the Health Insurance Portability and Accountability Act Privacy Rule). An IRB exemption was approved by the wcg IRB (#13931684). Knowledge dissemination will include presenting results at national and international conferences, sharing findings with specialty societies, and publishing results in peer-reviewed journals. All data management and analytic code will be made available publicly to enable others to leverage our methods to verify and extend our findings.
- Bariatric Surgery
- HEALTH ECONOMICS
- EPIDEMIOLOGIC STUDIES
- Health Equity
- Obesity
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
The protocol aims to use health information encompassing millions of patients with obesity to determine real-world impact of metabolic and bariatric surgery on obesity-related comorbidities and healthcare costs.
The studies will focus on a broad set of obesity-related comorbidities.
Healthcare costs in the 3 years postindex date will be derived from adjudicated healthcare costs including all costs associated with inpatient admissions, outpatient services, and outpatient pharmaceutical claims.
Due to the lack of linked laboratory data, this study will rely on cessation of treatment to determine remission of disease. This approach has been validated in previous studies using similar data.
The use of commercial claims data excludes individuals whose healthcare is delivered through other plans, limiting generalisability.
Introduction
The rate of obesity is increasing around the world,1 resulting in a higher incidence of obesity-related comorbidities, which can substantially impact an individual’s quality of life.2 The most effective and durable treatment for obesity is metabolic and bariatric surgery (MBS),3 yet only about 1% of individuals who could benefit from MBS receive this treatment (based on 2016 NIS estimates of bariatric surgeries and 2015/2016 NHANES estimates of eligible individuals).4 5 MBS both impacts obesity, and also improves—and in some cases resolves—obesity-related comorbidities,6 and reduces risk of developing obesity-related comorbidities among those who do not have these comorbidities.7 The most common obesity-related comorbidities are type 2 diabetes (T2D), hypertension, and dyslipidaemia, obstructive sleep apnea (OSA), and osteoarthritis (OA).8 The resolution or avoidance of obesity-related comorbidities that results from MBS can reduce healthcare costs and improve quality of life.
Obesity is associated with a range of comorbidities, and the remission of these comorbidities is possible through sustained weight loss.9–14 However, rates of obesity-related comorbidity improvement and remission vary substantially across and within studies examining different types of MBS.15–20 These variations are in part due to differences in baseline obesity and comorbidity severity across studies, and variations in definitions of improvement and remission, making it difficult to compare study results. Substantial weight loss is also associated with preventing the onset of obesity-related comorbidities. Previous research has found that individuals who had MBS are substantially less likely to develop T2D, hypertension, and dyslipidaemia, and reduced the likelihood of initiating treatment for these conditions;7 21–24 however, since these studies are primarily from outside of the USA, there is a gap in knowledge regarding this association in the USA. Recent data also suggest that individuals with knee OA who were planning to have a total knee arthroplasty (TKA) and have an MBS to reduce their weight before the surgery end up having lower rates of TKA because the weight loss alleviated their symptoms sufficiently.25 Both the improvement, resolution, and prevention of obesity-related comorbidities impacts healthcare use and costs. Research on the impact of MBS to healthcare costs also vary across studies, which are often limited by small sample sizes, use simulated data, or rely on data from several decades ago.26–30 These variations make it difficult to identify subpopulations that would benefit most (in terms of health outcomes and healthcare costs) from MBS.
There are several barriers to access, including lack of coverage by insurers, and patient and provider perceptions and attitudes.31 A better understanding of the characteristics and conditions of individuals who could benefit from MBS but have not received this treatment can identify target populations to work with to reduce barriers to access.
As the number of MBS performed in the USA increases every year,32 the quantity of real-world data available to assess the safety and efficacy increases. The evidence from this real-world data can address gaps in knowledge and reduce barriers to access for MBS. To address the identified knowledge gaps, we will conduct studies on the following six objectives:
Examine demographic and health characteristics of obese adults who have and have not had an MBS, and determine the rate of MBS by demographic and health characteristics
Compare the rate of remission of obesity-related comorbidities at 1 and 3 years postindex date for individuals who did and did not have bariatric surgery
Compare the rates of new onset obesity-related comorbidity at 1 and 3 years postindex date for individuals who did and did not have bariatric surgery.
Investigate the association of bariatric surgery and the risk of TKA among adults with obesity
Compare rates of total postindex date healthcare costs between those who did and did not have bariatric surgery.
Compare rates of out-of-pocket postindex date healthcare costs between those who did and did not have bariatric surgery.
Methods and analysis
Data
This is a retrospective claims data analysis that will use the Merative MarketScan Research Databases (Merative) (formerly known as IBM MarketScan Databases and Truven Health MarketScan), an aggregated database of employer-based health plans that contains all paid claims and encounter data generated by more than 273 million unique patients.33 The database includes enrolment, inpatient, outpatient, and prescription drug service use, representing the medical experience of insured employees and their dependents.33 The Merative MarketScan Research Databases consists of deidentified healthcare records and claims.
Enrolment and demographic information will be obtained from the Annual Enrolment Table. Diagnoses are identified in the inpatient (up to 15 diagnoses per record) and outpatient (up to four diagnoses per record) databases through International Classification of Diseases (ICD), 10th revision diagnosis codes. Up to 15 procedures are listed per record in the inpatient database and defined using Current Procedural Terminology (CPT) fourth edition codes. One procedure is listed per record in the outpatient database, and is defined using International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) procedure codes, Current Procedural Terminology fourth edition (CPT-4) codes, and Healthcare Common Procedure Coding System (HCPCS) codes. Medication use is identified using National Drug Codes (NDC) in the Outpatient Pharmaceutical Claims Table and through Generic Drug Names identified through RED BOOK.
Intervention: metabolic and bariatric surgery
Primary metabolic procedures from 2016 to 2021 will be identified; in this time, the most common MBS procedures were roux-en-Y gastric bypass (RYGB), sleeve gastrectomy (SG), and biliopancreatic diversion with duodenal switch (BPD-DS).32 These procedures will be identified in the inpatient admissions and outpatient services claims using CPT-4 and ICD-10 procedure codes (see table 1). For individuals who had more than one bariatric procedure identified in the study period, the first one will be selected to reduce potential bias associated with having multiple surgeries.
Codes use to identify bariatric procedures
Body mass index (BMI)
All inpatient and outpatient claims from 2016 to 2021 will be examined to identify ICD-10 codes for BMI (Z68.x). For individuals who had an MBS, claims will be examined in the year before the bariatric surgery; for those who had more than one BMI claim in this period, the one that was recorded closest to the surgery date will be selected. For individuals who did not have an MBS, one BMI claim will be randomly selected, and the date of this BMI claim will be used as their index date. Obesity is divided into three classes based on BMI: BMI 30–34.9 kg/m2 (class I; ICD-10 codes Z68.30–Z68.34), BMI 35–39.9 kg/m2 (class II; ICD-10 codes Z68.35–Z68.39), and BMI ≥ 40 kg/m2 (class III; ICD-10 codes Z68.41–Z68.45). Since those with class I obesity (BMI 30–34.9) are not eligible for insurance coverage for MBS, they are not the primary population of interest in these studies.
Study population
The research objectives will be examined using data from obese individuals aged 21–65 identified in the Merative data from 1 January 2016 to 31 December 2021. For cases, the index date will be the date of bariatric surgery, and for controls, the index date will be defined as the date they had a BMI diagnosis; for both groups, the index date will occur on or after 1 January 2017 (to obtain at least 1 year of baseline data). To determine whether there will be a sufficient sample size to conduct our planned analysis, a preliminary analysis was conducted to identify the number of cases and controls we can identify in the data. Cases are defined as individuals who had a bariatric surgery (RYGB, SG, and BPD-DS), had a BMI diagnosis in the year before surgery, had an insurance plan with pharmaceutical coverage, and had at least 1 year of continuous enrolment in the year before the surgery (n = 50 347). Controls are defined as individuals who did not have a bariatric surgery, an adjustable gastric band, a bariatric revision or indication of a previous bariatric surgery, had an insurance plan with pharmaceutical coverage, and had at least 1 year of continuous enrolment in the year before the BMI diagnosis (n = 1 814 810) (figure 1). Research questions examining outcomes in the year after the index date require individuals to have at least 1 year of continuous enrolment after the index date; there were 28 683 cases and 1 063 432 controls that had at least 1 year of continuous enrolment after their index date. Research questions examining outcomes in the 3 years after the index date require individuals needed to have at least 3 years of continuous enrolment; there were 8354 cases and 354 704 controls that met this criterion.
Cohort creation flow diagram. BMI, Body Mass Index.
Demographic information
Demographic information will be obtained from the enrolment file, and includes sex, age, region, and type of benefit plan. Sex was defined as sex at birth (male/female), and age, region, and type of benefit plan were defined at the index date. In these data, the USA is divided into five regions: Northeast, North Central, South, West, and Unknown, and there are nine types of benefit plans: basic/major medical, comprehensive, exclusive provider organisation, health maintenance organisation, point-of-service (POS), Preferred Provider Organisation, POS with capitation, consumer-driven health plan, and high-deductible health plan.
Baseline health and outcomes
Six obesity-related comorbidities will be examined in the year before the index date: T2D, hypertension, dyslipidaemia, OSA, and knee OA. We will also identify whether the person had an inpatient admission or a diagnosis for gastro-oesophageal reflux disease (GERD) and non-alcoholic fatty liver disease/non-alcoholic steatohepatitis (NAFLD/NASH).
Baseline healthcare costs will be examined from 6 to 12 months before the index date (figure 2). Baseline comorbidities will be defined using diagnoses and prescriptions filled in the year before the index date. Remission of T2D, hypertension, dyslipidaemia, and OSA will be examined in the 6–12 months and in the 6–36 months postindex date, new-onset obesity-related comorbidities and TKA will be examined in the 1 and 3 years postindex date, and healthcare costs are examined in the 6–12 and 6–36 months postindex date.
Baseline health and outcomes.
Baseline type 2 diabetes (T2D)
Individuals will be identified as having T2D if they have at least one diagnosis claim for T2D and at least one diabetes-related pharmacy claim from year before the index date. T2D diagnoses will be identified in the inpatient and outpatient claims using ICD-9 codes and ICD-10 diagnosis code (see table 2). Diabetes-related pharmacy claims will be identified through NDC, and will be separated into three groups: Metformin, Antidiabetic Medications (ADM), and Insulin.34 35 ADMs include alpha-glucosidase inhibitors, amylin analogues, antidiabetic combinations, dipeptidyl peptidase-4 inhibitors, glucagon-like peptide-1 inhibitors, meglitinides, sodium glucose cotransporter-2 inhibitors, sulfonylureas, and thiazolidinediones.
Diagnosis codes used to define health outcomes
Individuals identified as having T2D will be categorised based on their treatment for T2D, which was used as a proxy for the complexity of their T2D. Those who are using insulin (with or without Metformin/ADM) will be identified as having high-complexity T2D, those who are using at least one ADM (with or without Metformin) but not using insulin will be identified as having mid-complexity T2D, and those who only use metformin will be identified as having low-complexity T2D.
Baseline hypertension
Individuals will be identified as having hypertension if they have at least one diagnosis claim for hypertension and at least one antihypertensive medication pharmacy claim in the year before their index date. Hypertension diagnoses will be defined in the inpatient and outpatient diagnosis claims (see table 2). Antihypertensive medications will be identified through generic name from RED BOOK. First-line antihypertensive medications are thiazide or thiazide-type diuretics, angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), calcium chanel blocker (CCB)-dihydropyridines, and CCB-nondihydropyridines.36 Second-line antiypertensive treatments are diuretics–loop, diuretics–potassium sparing, diuretics, aldosterone antagonists, beta blockers–cardioselective, beta blocker–cardioselective and vasodilatory, beta blockers–non-cardioselective, beta blockers–intrinsic sympathomimetic activity, beta blockers–combined alpha-receptor and beta-receptor, direct renin inhibitor, alpha-1 blockers, central alpha-agonist and other centrally acting drugs, and direct vasodilators.36
Baseline dyslipidaemia
Individuals will be identified as having dyslipidaemia if they have at least one diagnosis claim for dyslipidaemia and at least one lipid lowering medication pharmacy claim in the year before their index date. Dyslipidaemia diagnoses will be identified in the inpatient and outpatient diagnosis claims (see table 2). Lipid lowering medications—both statins and non-statins—will be identified in the outpatient pharamceutical claims through generic name from RED BOOK. Statins included are atorvastatin, rosuvastatin, simvastatin, luvastatin, lovastatin, pitavastatin, and pravastatin.37 Non-statins included are ezetimibe, bile acid sequestrants, PCSK9 inhibitors, niacin, and fibrates.37
Baseline knee osteoarthritis (OA)
Knee OA will be defined as a knee OA diagnosis and a non-operative procedure in the year before the index date. A knee OA diagnosis will be identified from diagnosis claims in the inpatient and outpatient records (see table 2). Non-operative procedure for knee OA is physical therapy, bracing, intra-articular injections (hyaluronic and corticosteroids), and medications (opioids, NSAIDS, and acetaminophen), and knee-specific imaging (see table 2 for specific codes used to identify these procedures).38 Medications will be identified from the outpatient pharmaceutical database, with non-steroidal anti-inflammatory drugs (NSAIDS) and opioids being identified through NDCs and acetaminophen identified based on the generic name in RED BOOK.39 40
Baseline obstructive sleep apnea (OSA)
OSA will be defined as an OSA diagnosis in the year before the index date, and having an outpatient claim for CPAP/BiPAP machine supplies in the 6–12 months before the index date and in the 0–6 months before the index date. OSA diagnosis will identified in the inpatient and outpatient diagnosis claims (see table 2). CPAP/BiPAP machine and supplies will be identified in the outpatient claims using HCPCS codes (see table 2).
Baseline gastroesophageal reflux disease (GERD)
Individuals will be identified as having GERD if they have at least one diagnosis claim for GERD in the year before the index; GERD diagnoses will be identified in the inpatient and outpatient diagnosis claims (see table 2).
Baseline non-alcoholic fatty liver disease/non-alcoholic steatohepatitis (NAFLD/NASH)
Individuals will be identified as having NAFLD/NASH if they have at least one diagnosis claim for NAFLD/NASH in the year before the index; NAFLD/NASH diagnoses will be identified in the inpatient and outpatient diagnosis claims (see table 2).
Baseline healthcare costs
Baseline healthcare costs will be defined as all costs identified in the inpatient admissions (payments total case), outpatient services (payments net), and outpatient pharmaceutical claims (payment) from 6 to 12 months before the index date. The 6 months before the index date will be excluded as there were likely costs associated with preparing for the bariatric surgery in that period that would not reflect regular healthcare use.
All healthcare costs will be adjusted to 2021 constant dollars using the Medical Care component of the Bureau of Labour Statistics Consumer Price Index (http://www.bls.gov/cpi/). To remove the bias associated with extreme outliers, costs in each component (inpatient admissions, outpatient services, and outpatient pharmaceutical claims) will be truncated at the 1st and 99th percentile.
T2D, hypertension, dyslipidaemia remission
We will assess T2D, hypertension, and dyslipidaemia remission in the 6–12 months and in the 6–36 after bariatric surgery. We will exclude information from the first 6 months after the bariatric surgery to allow for a ‘wash-out’ period in which prescriptions may have be used that were filled before the surgery.19 Remission will be defined as not filling a disease-related pharmacy claim in the defined follow-up period.
OSA remission
OSA remission will be defined as an individual who had OSA at baseline having no outpatient claims for CPAP/BiPAP machine or supplies (as defined above) in the 6–12 months or in the 6–36 months after the index date.
Total knee arthroplasty (TKA)
TKA will be examined in the 3 years after the index date among individuals who had knee OA but did not have a TKA in the year before the index date. TKA will be identified in inpatient admissions and outpatient services by CPT-4 codes (see table 2)
Total healthcare costs
Total healthcare costs (inpatient admissions, outpatient services, and outpatient pharmaceutical claims) will be examined in the 6–12 and 6–36 months after the index date. The first 6 months after the index date were excluded as individuals who had a bariatric surgery likely had healthcare expenses in this period related to the surgery, which do not reflect regular healthcare use. As with healthcare costs at baseline, all expenses will be adjusted to 2021 constant dollars, and truncated at the 1st and 99th percentile.
Out-of-pocket healthcare costs
Out-of-pocket healthcare costs include the copays, coinsurance costs, and deductibles that individuals pay for inpatient admissions, outpatient services, and pharmaceutical claims. These costs will be examined in the 6–12 and 6–36 months after the index date; overall and in 6 month time periods. Again, costs will be adjusted to 2021 constant dollars, and truncated at the 1st and 99th percentile.
Data analysis plan
To address the first objective, we will examine demographic information (BMI, sex, age, region, and benefit plan type) and health (obesity-related comorbidities, healthcare costs, inpatient admission, and medical counselling for obesity) in the year before the index date among obese individuals who did and did not have bariatric surgery. We will then examine whether the rate of bariatric surgery varied based on baseline demographic and health. For example, whether the rate of bariatric surgery differs by sex, age, region, T2D diagnosis, etc.
Next, we will compare the rate of remission of obesity-related comorbidities (T2D, hypertension, dyslipidaemia, and OSA) among obese individuals who did and did not have bariatric surgery and had an obesity-related comorbidity at baseline. For each analysis, we limited the cohort to individuals who had the specific obesity-related comorbidity at baseline and had at least 1 year of continuous enrolment after their index date. For individuals with T2D at baseline, we examined rates of remission based on baseline T2D complexity. Individuals who had a bariatric surgery will be matched 1:1 using greedy nearest neighbour propensity scores with individuals who did not have a bariatric surgery based on index year, region, healthcare plan, sex, age, BMI, baseline healthcare costs, and presence and number of obesity-related comorbidities (T2D, hypertension, dyslipidaemia, OSA, GERD, and knee OA), and whether the individual had an inpatient admission. Balance diagnostics were examined to ensure that the matched groups were balanced on all baseline characteristic. A second set of analyses will use a similar approach to examine remission of obesity-related comorbidities in the 3 years after the index date.
Third, we will compare the rate of new-onset obesity-related comorbidities (T2D, hypertension, dyslipidaemia, and OSA) among obese individuals who did and did not have bariatric surgery and did not have an obesity-related comorbidity at baseline. For each analysis, we limited the cohort to individuals who did not have the specific obesity-related comorbidity at baseline and had at least 1 year of continuous enrolment after their index date. For this analysis, 1:1 matching will be done in a similar fashion as for objective 2. New-onset obesity-related comorbidities will be examined in the 1 and 3 years after the index date.
For individuals with knee OA, we wanted to determine whether those who had bariatric surgery had different rates of TKA in the 3 years postindex date than obese individuals with knee OA and similar baseline health who did not have bariatric surgery. For this analysis, we will limit the cohort to individuals who had a knee OA diagnosis at baseline, did not have TKA in the year before the index date, and had at least 3 years of continuous enrolment after the index date. Individuals who had a bariatric surgery will be matched 1:1 with individuals who did not have a bariatric surgery based on the same demographic and health characteristics outlined in the objective 2 analysis plan. Rates of TKA between the two groups will be compared using paired t-tests.
Next, we will compare total healthcare costs in the postindex period among those who did and did not have bariatric surgery but had similar health and demographic characteristics. For this analysis, we will limit the cohort to individuals who had at least one and then at least 3 years of continuous enrolment after the index date. Individuals who had a bariatric surgery will be matched 1:1 with individuals who did not have a bariatric surgery based on the same demographic and health characteristics outlined above. Overall healthcare costs will be examined, as well as the source of healthcare costs (pharmaceutical, inpatient admissions, and outpatient services). For the final analysis, the previous analysis will be replicated, but only out-of-pocket costs will be considered. A subanalysis will be done for the healthcare costs objective comparing bariatric surgery and antiobesity medication use.
Given the recent increased focus on bariatric surgery among individuals with class I obesity,41 we repeated each matched analysis for a subcohort of individuals with BMI 30–34.9.
Patient and public involvement
Patients and the public were not involved in the design of this study in any way.
Ethics and dissemination
In the USA, retrospective analyses of the Merative data are considered exempt from informed consent and Institutional Review Board (IRB) approval pursuant to 45 CFR 46.101(b)(4). An IRB exemption was approved by the wcg IRB (#13931684).
Knowledge dissemination will include presenting results at national and international conferences, sharing findings with specialty societies, and publishing results in peer-reviewed journals.
The data are available through Merative, and the data management and analysis code will be available in a GitHub library, and on request.
Ethics statements
Patient consent for publication
References
Footnotes
Contributors All authors contributed to the conception and design of the study. EW-W will perform data analysis, and with VG, YL, FZ, and MSA, will interpret the data. All authors participated in the preparation of this manuscript, revised it critically for important intellectual content and approved the version submitted for publication.
Funding Intuitive Surgical Inc funded access to the IBM MarketScan Research Database.
Competing interests Yes, there are competing interests for one or more authors and I have provided a Competing Interests statement in my manuscript and in the box below.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.