Objective To determine whether bariatric surgery is associated with reduced health care expenditures in a multisite cohort of predominantly older male patients with a substantial disease burden.
Design Retrospective cohort study of bariatric surgery. Outpatient, inpatient, and overall health care expenditures within Department of Veterans Affairs (VA) medical centers were examined via generalized estimating equations in the propensity-matched cohorts.
Setting Bariatric surgery programs in VA medical centers.
Participants Eight hundred forty-seven veterans who were propensity matched to 847 nonsurgical control subjects from the same 12 VA medical centers.
Intervention Bariatric surgical procedures.
Main Outcome Measure Health expenditures through December 2006.
Results Outpatient, inpatient, and total expenditures trended higher for bariatric surgical cases in the 3 years leading up to the procedure and then converged back to the lower expenditure levels of nonsurgical controls in the 3 years after the procedure.
Conclusions Based on analyses of a cohort of predominantly older men, bariatric surgery does not appear to be associated with reduced health care expenditures 3 years after the procedure.
Obesity is difficult to treat, and bariatric surgery is the most effective means to induce weight loss for the severely obese (body mass index [BMI; calculated as weight in kilograms divided by height in meters squared], >40), whose prevalence increased 50% from 2000 to 2005.1,2 Few long-term studies of the health and economic benefits of bariatric surgery exist,3 and most prior studies examined outcomes in younger, primarily white and female populations. As the demand for bariatric surgery has increased,4 numbers of nonwhite, older, and male patients with a greater prevalence of obesity-related comorbidities have increased, and their expenditure trends have not been evaluated extensively.5,6
Obesity-associated health expenditures were estimated at $147 billion in 2008,7 are disproportionately higher for morbidly obese patients,8 and increase with age.9 Prior studies of predominantly younger female cohorts have found that bariatric surgery results in cost savings 2 to 5 years after the procedure.10-13 Several simulation studies have also found bariatric surgery to be cost-effective, particularly for patients with diabetes mellitus.14-18 Despite this evidence, health insurers have been slow to provide coverage for bariatric surgery and are increasingly requiring risk-adjusted outcomes to assess return on investment.
The purpose of this study is to compare expenditures 3 years before and after operations in a cohort of veterans who underwent bariatric surgical procedures at 1 of 12 Veterans Affairs medical centers (VAMCs) in 2000 through 2006 with expenditures in a matched cohort of veterans who did not undergo bariatric surgery. We hypothesized that this cohort of surgical patients with a substantial obesity-related comorbid disease burden would have lower health expenditures in the years after bariatric surgery than nonsurgical control subjects. This study is the first, to our knowledge, to examine health expenditures associated with bariatric surgery in a multisite cohort of predominantly male patients who were older than patients in prior bariatric surgery evaluations.
Study design and study population
As described previously,5 892 bariatric surgical patients from this retrospective cohort study were identified from a database of major surgical procedures performed in VAMCs maintained by the VA Surgical Quality Improvement Program (formerly known as the VA National Surgical Quality Improvement Program). Veterans in the program's database were identified as undergoing bariatric surgical procedures in fiscal years 2000 through 2006 based on an algorithm of codes from the Current Procedural Terminology 4 (43842, 43843, 43846, 43847, and 43848) and International Classification of Diseases, Ninth Revision (ICD-9) (278.01 associated with Current Procedural Terminology 4 codes 43659, 43621, or 43633), with a sensitivity of 99.2% and specificity of 99.9%.19 Time of study entry began with each veteran's first instance of observed BMI of at least 35 after January 1, 2000. Patients were excluded on the basis of death dates before the bariatric procedure (n = 4), presurgical BMIs of less than 35 (n = 9), or missing presurgical comorbidity scores (n = 25). The final surgical cohort included 850 high-risk veterans who underwent bariatric surgical procedures (see eFigure 1). Nearly 97% of these patients underwent Roux-en-Y gastric bypass; the remainder, vertical banded gastroplasty.20
Nonsurgical controls were identified from a VA registry of 1.8 million veterans whose height and weight data from fiscal year 2000 were extracted from VA electronic medical records of outpatient visits from 136 VAMCs.21 We identified 98 545 veterans who had recorded BMIs between 29 and 80 and had never undergone bariatric surgery in the VA system (see eFigure 2). Controls were excluded owing to missing patient identifiers needed to link to other VA databases (n = 563), death before fiscal year 2000 (n = 373), or incomplete data to construct patient covariates (n = 157). We also excluded nonsurgical controls without BMI values of at least 35 (n = 4853), with missing baseline comorbidity scores (n = 5836), 80 years or older during the study period (n = 8509), or with clinical exclusions related to gastrointestinal tract cancer requiring surgery or peritoneal effusion/ascites (n = 299). To more closely match the availability of medical care resources between surgical and control groups, we excluded severely obese veterans who lived in Veterans Integrated Service Networks lacking a bariatric surgery center in 2000 through 2006 (n = 36 711). The nonsurgical control cohort included 41 244 veterans with an initial eligible BMI in 2000 through 2006 and complete data on patient characteristics and survival status.
Health care expenditure outcomes
We examined outpatient, inpatient, and total VA expenditure outcomes. Outpatient and inpatient expenditures were aggregated from the Health Economics Resource Center average cost data, and medication expenditures were subsumed within outpatient expenditures.22 Details on these data have been described previously.23 Total expenditures were constructed by summing VA outpatient and inpatient expenditures. The unit of analysis was the person–half year, and we constructed 6 half-year outcomes for the 3 years before the bariatric procedure and 6 half-year outcomes for the 3 years after it. For patients in the surgical cohort, the surgical and inpatient costs associated with the hospitalization for the bariatric surgical procedure were attributed to the half year immediately before the procedure.
Differences between patients undergoing or not undergoing bariatric surgical procedures were compared using χ2 tests for categorical variables and 2-tailed unpaired t tests for continuous variables in the unmatched cohorts, McNemar tests and paired t tests in the matched cohorts, and standardized differences to enable comparison of covariate imbalance between the matched and unmatched cohorts.24
Given the numerous differences between the surgical patients and nonsurgical controls, we conducted expenditure analyses on propensity-matched cohorts.5 Despite the large sample of controls (n = 41 244), we conducted one-to-one matching to avoid the possible bias of many-to-one matching,25 and cases were matched to controls using a greedy algorithm.26 Each surgical case was matched to a single nonsurgical control if their predicted propensity scores were identical to 8 digits. If such a match was not found, the case was matched to a control on the basis of a 7-, 6-, 5-, 4-, 3-, 2-, or 1-digit match.
As described previously, the propensity score model included presurgical age, age squared, baseline comorbidity via the Diagnostic Cost Group score, baseline BMI, BMI squared, BMI cubed, sex, race, marital status, Veterans Integrated Service Networks, and numerous 2-way interactions27 and had a concordance index of 0.85.5 Diagnostic Cost Group scores were as predictive of veterans' 1-year mortality as other comorbidity scores28-30 and were highly predictive of mortality,20 use of medical resources, and expenditures31 in bariatric surgical procedures. This process matched 847 surgical cases (of 850 possible [99.6%]) to 847 nonsurgical controls.5
After examining the expenditure distributions and Pregibon, Hosmer-Lemeshow, modified Park, and Pearson specification tests,32 outpatient and total expenditures were estimated using 1-part generalized estimating equations with a log-link function and distribution in which the mean and variance are proportional, which allowed for the possibility of overdispersion. We modeled inpatient expenditures using a 2-part generalized estimating equation in which a logistic regression was estimated for the probability of inpatient expenditures and a separate generalized estimating equation was estimated for the level of inpatient expenditures among the subset who were hospitalized. Because these analyses were performed on well-matched cohorts, we adjusted for fixed effects for each 6-month period and interactions of treatment with each 6-month period. We then generated predicted expenditures for the surgical cases and nonsurgical controls in each of the 6 half years to determine whether health expenditures were different between surgical cases and nonsurgical controls. For surgical patients, the presurgical period ended on the day of hospital discharge for the bariatric operation. For nonsurgical controls, the presurgical and postsurgical periods were defined by the day of discharge for the surgical patient to whom he or she was matched.
Expenditures were inflation adjusted to 2006 dollars using the Consumer Price Index because the medical Consumer Price Index does not adequately account for technological improvement, quality change, and improved health outcomes.33 All regression models were adjusted for clustering on the patient level using commercially available software (STATA, version 10.1; StataCorp). The a priori level of statistical significance was set at P < .05 for all analyses. This study was approved by the VA Surgical Quality Data Use Group and by the institutional review boards of the Durham, North Texas, and Denver VAMCs and the Group Health Research Institute.
Patient characteristics in unmatched and matched cohorts
As discussed previously,5 the unmatched cohort of 850 veterans in the surgical cohort and 41 244 veterans in the nonsurgical control cohort were different in nearly all respects (Table 1). Propensity score matching resulted in well-balanced cohorts of surgical cases (n = 847) and nonsurgical controls (n = 847), which were similar in all observed characteristics except year of study entry (Table 2), improving covariate balance from the unmatched cohorts. As reported in a prior analysis, 6.8% of the surgical patients died within 6 years of surgery and 12.8% of the matched nonsurgical controls died within 6 years of the index date.5
Expenditure results on the matched cohorts
In the 3 years before the bariatric procedure, the difference in adjusted outpatient expenditures between surgical patients and nonsurgical controls diverged from $123 higher for surgical patients (P = .31) in the presurgical 36 to 31 months to $2758 higher (P < .001) in the presurgical 6 months (Figure 1). In the 3 years after the procedure, outpatient expenditures converged again. Adjusted outpatient expenditures were $1223 higher (P < .001) for surgical patients in the first postsurgical 6 months but only $40 higher (P = .82) in the 31 to 36 postsurgical months.
Differences in adjusted inpatient expenditures followed a similar diverging trend before the bariatric procedure, followed by a similar converging trend after the procedure (Figure 2). In the presurgical 36 to 31 months, adjusted inpatient expenditures were $694 lower (P = .18) for surgical patients but became $25 645 higher (P < .001) in the 3 presurgical months owing to the cost of the admission associated with bariatric surgery. Adjusted inpatient expenditures converged in the 3 years after the procedure, from $3033 higher (P < .001) in the first 6 postsurgical months among surgical patients but only $45 higher (P = .90) in the 31 to 36 postsurgical months.
The trends in adjusted total expenditures mirrored the trends of outpatient and inpatient expenditures (Figure 3). In the presurgical 36 to 31 months, adjusted total expenditures were $595 lower (P = .26) for surgical patients but became $28 400 higher (P < .001) in the 6 months leading up to and including the procedure. Adjusted total expenditures converged in the 3 years after the operation, from $4397 higher (P < .001) in the first 6 postsurgical months to similar expenditures (P = .75) in the 31 to 36 postsurgical months.
In a propensity-matched cohort of obese, high-risk, primarily male patients, bariatric surgery was not significantly associated with lower health expenditures 3 years after the procedure. As expected, health expenditures were similar 2 and 3 years before the surgical procedure because surgical patients and nonsurgical controls had similar weight and health care use trajectories several years before giving serious consideration to bariatric surgery. In addition, most patients who eventually undergo bariatric surgical procedures have not yet begun intense use of health care resources associated with evaluation for the operation and/or nonsurgical weight loss treatment at this point.
In the year leading up to the bariatric surgical procedure, we expected health expenditures to be higher for patients who underwent the operation for 2 reasons. First, it is common to see weight steadily increasing in bariatric surgical patients in the year before the procedure, which may result in more use of health care resources and expenditures than in prior years when weight was more stable. Second, this increase in weight likely prompts patients and their health care providers to determine eligibility for surgical or nonsurgical weight loss treatment. Once eligibility is confirmed, expenditures of surgical patients increase owing to presurgical workup and evaluation (more providers and tests involved in their care), intense medical weight loss treatment that is requested of some patients to shed weight before the procedure to reduce risk and make the operation technically easier, and the one-time significant expenditure of the operation itself. Several prior studies have demonstrated this short-term increase in health expenditures,11,15,34,35 and our results found a similar increase.
In the year after bariatric surgical intervention, we expected health expenditures to be higher for patients who underwent the procedure owing to care associated with postsurgical follow-up, treatment of postsurgical complications, and frequent adjustment of long-term medication therapy required during periods of rapid weight loss as is typically seen after bariatric surgical procedures. Two and three years after the operation, we expected expenditures for surgical cases to continue to decline from the immediate postsurgical period and to become lower than expenditures of nonsurgical controls (based on results from prior literature). Our results did not support this hypothesis but are consistent with a prior study of patients with diabetes that observed a similar trend 2 years after the surgical procedure.12 The finding that expenditures were unchanged within 2 years after bariatric surgical procedures is also consistent with a study of hospitalizations that found no decrease in use of hospital services attributable to the operations.36
These results are notable because they contrast with results from several prior observational studies that found expenditures among postsurgical cases to be lower than those of nonsurgical controls 2 to 4 years after the procedures, which can be explained by important differences in the populations examined and the methods of analysis. The proportion of women was much lower in this study (26.1%) than in a study by Christou et al10 (66%) and 2 studies of American patients with employer-based insurance by Cremieux et al11 (86%) and Finkelstein et al15 (80%). The average age of surgical patients was also higher in our study than in these prior studies (49.5 vs 44-45 years). In addition, the study by Finkelstein et al15 evaluated cost trends for patients undergoing laparoscopic adjustable gastric banding, whereas we examined patients undergoing gastric bypass.5
The results from our study also differ from these earlier studies owing to differences in how nonsurgical controls were identified and matched to surgical patients. In the prior studies, nonsurgical controls were identified as obese by having a diagnosis of obesity (ICD-9 code 278.00) or morbid obesity (ICD-9 code 278.01) in their medical record, which might have identified a sicker and more costly subset of surgery-eligible patients. The study by Finkelstein et al15 was more comparable to our study because self-reported BMI data were available on a subset of patients who completed a health risk assessment.
All 3 studies attempted to reduce confounding bias by matching surgical cases and nonsurgical controls. Controls in the study by Christou et al10 were matched 6:1 to cases based on patient age, sex, and duration of follow-up, which resulted in a sample of 1035 surgical patients and 5746 control patients. No clinical information was available for comorbidity or BMI adjustment or matching. During a mean follow-up of 2.5 years, total direct expenditures were found to be lower for surgical patients than nonsurgical controls ($8813 vs $11 854; P < .001) owing to fewer inpatient admissions and physician visits.10
Cases and controls in the study by Cremieux et al11 were matched on age, sex, state of residence, 10 comorbidities, and health care costs in the 5 months before surgery (−6 to −2 months). During a mean of 18 months of follow-up and under an assumption of constant cost differences beyond 19 months, health expenditures were found to decrease after surgical procedures until 18 months and remained $545 lower every month after month 18. In an editorial that accompanied this study, Finkelstein and Brown34 questioned the assumption of consistently lower ($545) expenditures for surgical patients every month after 18 months. The 3-year expenditure results from our high-risk cohort suggest that this assumption is not likely to hold in all surgical populations. A recent study by Finkelstein et al15 matched cases and controls based on age, sex, metropolitan statistical area status, health plan type, 24 comorbidities, and health care costs in the quarters (−5 to −2 months) before surgical procedures. The investigators found that health expenditures of surgical patients decreased sufficiently to fully offset the cost of the procedure at 4 years.
Although the inclusion of preperiod outcomes is typically reasonable for improving the effectiveness of propensity score matching,37 matching on presurgical expenditures in the studies by Cremieux et al11 and Finkelstein et al15 likely induces the selection of matched controls with rapidly accelerating expenditures. If surgery-related expenditures increased only on the day of the operation, then matching on presurgical expenditures would not be problematic. However, surgery-related expenditures begin in the 6 to 12 months before the day of the operation as a result of the presurgical workup and intensive nonsurgical weight loss treatment. Thus, expenditure comparisons with matched controls who have increasing expenditures in the immediate presurgical period likely overstate the cost reductions associated with bariatric surgery. In their editorial accompanying the article by Cremieux et al,11 Finkelstein and Brown note that the “return on investment estimates seem[s] to result from substantially higher cost increases in the control group relative to the surgery group in the postsurgery period.”34(p561) We speculate that the increasing costs of the nonsurgical controls are likely induced by matching on presurgical expenditures in both studies and an important reason why the expenditure results from these studies differ from the results we found in our high-risk cohort.
However, this study is subject to several limitations. First, we focused on a cohort of older, predominantly male, sicker patients, so results may not generalize to nonveteran, younger, female, or healthier populations. Bariatric surgical procedures may reduce expenditures for younger patients and not for older patients, but we did not have a sufficient sample size to examine whether the association varied across subgroups. Future studies with larger samples should also consider stratification across clinically relevant patient factors to fully understand changes in expenditures across various subgroups. Second, our study does not include patients who underwent laparoscopic adjustable gastric banding procedures; thus, we cannot generalize our results to such patients. Third, these results do not account for unobserved confounding that may persist even after propensity score matching because this analysis was based on a quasi-experimental design from administrative data, not a randomized trial.38 Our results represent an association and not necessarily the causal effect of bariatric surgery on health care expenditures. We were able to reduce confounding of this association via one-to-one propensity score matching without a significant loss in sample size in the surgical group.
Although bariatric surgery was not associated with reduced expenditures in this cohort of older, predominantly male patients, many patients may still choose to undergo bariatric surgery given the strong evidence of significant reductions in body weight and comorbidities and improved quality of life. Expenditures may decline further for surgical cases in the longer term, but there were no differences in health expenditures between the surgical and nonsurgical cases during 3 years of follow-up.
Correspondence: Matthew L. Maciejewski, PhD, Center for Health Services Research in Primary Care (Mail Stop 152), Durham VA Medical Center, 411 W Chapel Hill St, Ste 600, Durham, NC 27705 (matthew.maciejewski@va.gov).
Accepted for Publication: February 6, 2012.
Author Contributions:Study concept and design: Maciejewski, Kahwati, Henderson, and Arterburn. Acquisition of data: Maciejewski and Henderson. Analysis and interpretation of data: Maciejewski, Livingston, Smith, Kahwati, Henderson, and Arterburn. Drafting of the manuscript: Maciejewski and Arterburn. Critical revision of the manuscript for important intellectual content: Maciejewski, Livingston, Smith, Kahwati, Henderson, and Arterburn. Statistical analysis: Maciejewski, Smith, and Henderson. Obtained funding: Maciejewski, Henderson, and Arterburn. Administrative, technical, and material support: Maciejewski, Kahwati, Henderson, and Arterburn. Study supervision: Maciejewski and Arterburn.
Financial Disclosure: Dr Maciejewski has received consultation funds from Takeda Pharmaceuticals, Novartis, and the Surgical Review Corporation and owns stock in Amgen. Dr Livingston has received consulting funds from Texas Instruments and serves as a contributing editor to JAMA. Dr Arterburn receives research funding and has received salary support as a medical editor for the not-for-profit (501[c][3]) Foundation for Informed Medical Decision Making (http://www.fimdm.org), which develops content for patient education programs. The Foundation has an arrangement with a for-profit company, Health Dialog, to coproduce and market these programs to health care organizations. Dr Arterburn was formerly with the Cincinnati VA, and Dr Livingston was formerly with the Dallas VA.
Funding/Support: This study was supported by grants IIR 05-201 and SHP 08-137 from the Office of Research and Development, Health Services Research and Development Service, Department of Veterans Affairs, and by Research Career Scientist award RCS 10-391 from the Department of Veterans Affairs (Dr Maciejewski).
Role of the Sponsor: The Health Services Research and Development Service, Department of Veterans Affairs, had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Disclaimer: The opinions expressed herein are those of the authors and not necessarily those of the Department of Veterans Affairs, the US government, Duke University, The University of Texas Southwestern Medical Center, The University of Texas, the University of Colorado, the Group Health Research Institute, or the University of Washington.
Additional Contributions: The VA Surgical Quality Data Use Group served as scientific advisors and provided a critical review of data use and analysis presented herein.
2.Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ, Flegal KM. Prevalence of overweight and obesity in the United States, 1999-2004.
JAMA. 2006;295(13):1549-155516595758
PubMedGoogle ScholarCrossref 5.Maciejewski ML, Livingston EH, Smith VA,
et al. Survival among high-risk patients after bariatric surgery.
JAMA. 2011;305(23):2419-242621666276
PubMedGoogle ScholarCrossref 6.Fontaine KR, Redden DT, Wang C, Westfall AO, Allison DB. Years of life lost due to obesity.
JAMA. 2003;289(2):187-19312517229
PubMedGoogle ScholarCrossref 7.Finkelstein EA, Trogdon JG, Cohen JW, Dietz W. Annual medical spending attributable to obesity: payer- and service-specific estimates.
Health Aff (Millwood). 2009;28(5):w822-w83119635784
PubMedGoogle ScholarCrossref 8.Arterburn DE, Maciejewski ML, Tsevat J. Impact of morbid obesity on medical expenditures in adults.
Int J Obes (Lond). 2005;29(3):334-33915685247
PubMedGoogle ScholarCrossref 9.Bell JF, Zimmerman FJ, Arterburn DE, Maciejewski ML. Health-care expenditures of overweight and obese males and females in the Medical Expenditures Panel survey by age cohort.
Obesity (Silver Spring). 2011;19(1):228-23220467420
PubMedGoogle ScholarCrossref 10.Christou NV, Sampalis JS, Liberman M,
et al. Surgery decreases long-term mortality, morbidity, and health care use in morbidly obese patients.
Ann Surg. 2004;240(3):416-42415319713
PubMedGoogle ScholarCrossref 11.Cremieux PY, Buchwald H, Shikora SA, Ghosh A, Yang HE, Buessing M. A study on the economic impact of bariatric surgery.
Am J Manag Care. 2008;14(9):589-59618778174
PubMedGoogle Scholar 12.Keating CL, Dixon JB, Moodie ML, Peeters A, Playfair J, O’Brien PE. Cost-efficacy of surgically induced weight loss for the management of type 2 diabetes: a randomized controlled trial.
Diabetes Care. 2009;32(4):580-58419171726
PubMedGoogle ScholarCrossref 13.Sampalis JS, Liberman M, Auger S, Christou NV. The impact of weight reduction surgery on health-care costs in morbidly obese patients.
Obes Surg. 2004;14(7):939-94715329183
PubMedGoogle ScholarCrossref 14.Hoerger TJ, Zhang P, Segel JE, Kahn HS, Barker LE, Couper S. Cost-effectiveness of bariatric surgery for severely obese adults with diabetes.
Diabetes Care. 2010;33(9):1933-193920805271
PubMedGoogle ScholarCrossref 15.Finkelstein EA, Allaire BT, Burgess SM, Hale BC. Financial implications of coverage for laparoscopic adjustable gastric banding.
Surg Obes Relat Dis. 2011;7(3):295-30321195677
PubMedGoogle ScholarCrossref 16.Ikramuddin S, Klingman D, Swan T, Minshall ME. Cost-effectiveness of Roux-en-Y gastric bypass in type 2 diabetes patients.
Am J Manag Care. 2009;15(9):607-61519747025
PubMedGoogle Scholar 17.Campbell J, McGarry LA, Shikora SA, Hale BC, Lee JT, Weinstein MC. Cost-effectiveness of laparoscopic gastric banding and bypass for morbid obesity.
Am J Manag Care. 2010;16(7):e174-e18720645663
PubMedGoogle Scholar 18.Salem L, Devlin A, Sullivan SD, Flum DR. Cost-effectiveness analysis of laparoscopic gastric bypass, adjustable gastric banding, and nonoperative weight loss interventions.
Surg Obes Relat Dis. 2008;4(1):26-3218069075
PubMedGoogle ScholarCrossref 19.Livingston EH, Arterburn D, Schifftner TL, Henderson WG, DePalma RG. National Surgical Quality Improvement Program analysis of bariatric operations: modifiable risk factors contribute to bariatric surgical adverse outcomes.
J Am Coll Surg. 2006;203(5):625-63317084323
PubMedGoogle ScholarCrossref 20.Arterburn D, Livingston EH, Schifftner T, Kahwati LC, Henderson WG, Maciejewski ML. Predictors of long-term mortality after bariatric surgery performed in Veterans Affairs medical centers.
Arch Surg. 2009;144(10):914-92019841358
PubMedGoogle ScholarCrossref 21.Das SR, Kinsinger LS, Yancy WS Jr,
et al. Obesity prevalence among veterans at Veterans Affairs medical facilities.
Am J Prev Med. 2005;28(3):291-29415766618
PubMedGoogle ScholarCrossref 22.Wagner TH, Chen S, Barnett PG. Using average cost methods to estimate encounter-level costs for medical-surgical stays in the VA.
Med Care Res Rev. 2003;60(3):(suppl)
15S-36S15095543
PubMedGoogle ScholarCrossref 23.Chapko MK, Liu CF, Perkins M, Li YF, Fortney JC, Maciejewski ML. Equivalence of two healthcare costing methods: bottom-up and top-down.
Health Econ. 2009;18(10):1188-120119097041
PubMedGoogle ScholarCrossref 24.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples.
Stat Med. 2009;28(25):3083-310719757444
PubMedGoogle ScholarCrossref 25.Austin PC. Statistical criteria for selecting the optimal number of untreated subjects matched to each treated subject when using many-to-one matching on the propensity score.
Am J Epidemiol. 2010;172(9):1092-109720802241
PubMedGoogle ScholarCrossref 26.Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques [paper 214]. In: SUGI 26 Proceedings. Cary, NC: SAS Institute Inc; 2001
27.Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods.
JAMA. 2007;297(3):278-28517227979
PubMedGoogle ScholarCrossref 28.Maciejewski ML, Liu CF, Derleth A, McDonell M, Anderson S, Fihn SD. The performance of administrative and self-reported measures for risk adjustment of Veterans Affairs expenditures.
Health Serv Res. 2005;40(3):887-90415960696
PubMedGoogle ScholarCrossref 29.Maciejewski ML, Liu CF, Fihn SD. Performance of comorbidity, risk adjustment, and functional status measures in expenditure prediction for patients with diabetes.
Diabetes Care. 2009;32(1):75-8018945927
PubMedGoogle ScholarCrossref 30.Fan VS, Maciejewski ML, Liu CF, McDonell M, Fihn SD. Comparison of risk adjustment measures based on self-report, administrative data and pharmacy records to predict clinical outcomes.
Health Serv Outcomes Res Methodol. 2006;6(1-2):21-36
Google ScholarCrossref 31.Maciejewski ML, Smith VA, Livingston EH,
et al. Health care utilization and expenditure changes associated with bariatric surgery.
Med Care. 2010;48(11):989-99820940651
PubMedGoogle ScholarCrossref 32.Manning WG, Basu A, Mullahy J. Generalized modeling approaches to risk adjustment of skewed outcomes data.
J Health Econ. 2005;24(3):465-48815811539
PubMedGoogle ScholarCrossref 33.Berndt ER, Bir A, Busch SH, Frank RG, Normand SL. The medical treatment of depression, 1991-1996: productive inefficiency, expected outcome variations, and price indexes.
J Health Econ. 2002;21(3):373-39612022264
PubMedGoogle ScholarCrossref 34.Finkelstein EA, Brown DS. Return on investment for bariatric surgery.
Am J Manag Care. 2008;14(9):561-56218778170
PubMedGoogle Scholar 35.Makary MA, Clarke JM, Shore AD,
et al. Medication utilization and annual health care costs in patients with type 2 diabetes mellitus before and after bariatric surgery.
Arch Surg. 2010;145(8):726-73120713923
PubMedGoogle ScholarCrossref 36.Zingmond DS, McGory ML, Ko CY. Hospitalization before and after gastric bypass surgery.
JAMA. 2005;294(15):1918-192416234498
PubMedGoogle ScholarCrossref 37.Cook TD, Steiner PM. Case matching and the reduction of selection bias in quasi-experiments: the relative importance of pretest measures of outcome, of unreliable measurement, and of mode of data analysis.
Psychol Methods. 2010;15(1):56-6820230103
PubMedGoogle ScholarCrossref 38.Flum DR. Administrative data analyses in bariatric surgery: limits of the technique.
Surg Obes Relat Dis. 2006;2(2):78-8116925326
PubMedGoogle ScholarCrossref