Association between neighborhood food environment and bariatric surgery outcomes

Abstract

Background:

Individual characteristics associated with weight loss after bariatric surgery are well established, but the neighborhood characteristics that influence outcomes are unknown.

Objective:

To determine if neighborhood characteristics, including social determinants and lifestyle characteristics, were associated with weight loss after bariatric surgery.

Setting:

Single university healthcare system, United States

Methods:

In this retrospective cohort study, all patients who underwent primary bariatric surgery from 2008–2017 and had at least one year of follow-up data were included. Patient-level demographics and neighborhood-level social determinants (area deprivation index, urbanicity, and walkability) and lifestyle factors (organic food use, fresh fruit/vegetable consumption, diet to maintain weight, soda consumption, and exercise) were analyzed. Median regression with percent total body weight (%TBW) loss as the outcome was applied to examine factors associated with weight loss after surgery.

Results:

Of the 647 patients who met inclusion criteria, the average follow-up period was 3.1 years, and the mean %TBW loss at follow-up was 22%. In adjusted median regression analyses, Roux-en-Y gastric bypass was associated with greater %TBW loss (11.22%, 95% CI [8.96, 13.48]) compared to sleeve, while longer follow-up time (−2.42% TBW loss per year, 95% CI [−4.63, −0.20]) and a pre-operative diagnosis of diabetes (−1.00% TBW loss, 95% CI [−1.55, −0.44]) were associated with less. None of the 8 neighborhood level characteristics was associated with weight loss.

Conclusions:

Patient characteristics rather than neighborhood-level social determinants and lifestyle factors were associated with weight loss after bariatric surgery in our cohort of bariatric surgery patients. Patients from socioeconomically deprived neighborhoods can achieve excellent weight loss after bariatric surgery.

INTRODUCTION

Approximately 260,000 bariatric surgery procedures are performed annually in the U.S.⁽¹⁾ Compared to behavioral weight management alone, bariatric surgery is more effective at achieving sustained weight loss, comorbidity resolution, and improved quality of life.^(2,3) Over 80% of patients will achieve optimal weight loss, or ≥ 50% excess weight loss (EWL) one year post-operatively.⁽⁴⁾

Despite excellent weight loss outcomes for bariatric patients overall, the literature is mixed regarding the impact of socioeconomic status on weight loss after bariatric surgery. One systematic review of 21 studies found that Medicaid patients experienced higher mortality rates and short-term healthcare utilization, but similar comorbidity resolution.⁽⁵⁾ Of the seven studies that examined weight loss in the systematic review, four reported similar amounts of weight loss at follow-up while three reported that Medicaid patients experienced less weight loss compared to non-Medicaid patients.⁽⁵⁾ It is unknown if neighborhood characteristics, such as the average amount of produce consumption or exercise, are associated with differences in bariatric surgery outcomes.

The objective of this study was to assess whether neighborhood characteristics, including social determinants and lifestyle characteristics, were associated with weight loss after bariatric surgery. At the neighborhood level, we analyzed three social determinant variables (area deprivation index [ADI], urbanicity, and walkability) and five lifestyle factors (organic food consumption, fresh fruit/vegetable consumption, soda consumption, dieting to maintain weight, and routine exercise). The lifestyle factors were obtained from the Esri database, which uses proprietary Geographic Information System (GIS) mapping tools to track and map customer and environmental data obtained from a variety of sources including census data, GIS data, business distribution databases, and large-scale consumer surveys.⁽⁶⁾ We hypothesized that neighborhood level factors, such as produce consumption and routine exercise, would be associated with weight loss outcomes. Factors that facilitate routine exercise, such as suburban neighborhoods and higher walkability, were hypothesized to be associated with greater weight loss.

METHODS

Study population

We included patients ≥ 18 years old who underwent primary laparoscopic Roux-en-Y gastric bypass (RYGB) or laparoscopic sleeve gastrectomy (SG) between 2008 and 2017 at University of Wisconsin Hospital and Clinics (UW). Patients were required to have one year follow-up, defined as ≥ 240 days of follow-up data at the time of analysis since some “one year follow up visits” are scheduled earlier than 365 days after surgery. The study cohort was identified using the UW institutional bariatric surgery registry (BSR), which includes all patients who have undergone bariatric surgery at UW. The BSR is populated using inpatient and outpatient electronic health record (EHR) data by a trained data extractor who is not involved with patient care.⁽⁷⁾

Patient characteristics

The BSR was used to identify patient age at the time of surgery, sex (male or female), body mass index (BMI) on the day of surgery, race and ethnicity (White non-Hispanic or non-White [Hispanic, Native American, Asian, Black, and other/unspecified]), smoking status (never smoker, quit < 6 months or ≥ 6 months before surgery, and current smoker) and insurance type (Medicare, Medicaid, or commercial). The insurance type for Medicare and commercial patients was defined as insurance used at the time of surgery, whereas insurance type for Medicaid patients was defined as having Medicaid insurance within three years prior to surgery. We have previously used this definition for Medicaid insurance⁽⁷⁾ since Medicaid patients typically have limited socioeconomic mobility.⁽⁸⁾ Preoperative diagnoses of coronary artery disease, gastroesophageal reflux disease (GERD), hyperlipidemia (HLD), hypertension (HTN), obstructive sleep apnea (OSA), and type 2 diabetes mellitus (T2D) were obtained from the BSR. Clinic notes, imaging studies, and laboratory values were used to identify these patient characteristics and diagnoses as we have previously described.⁽⁹⁾ Anxiety and depression are not captured in the BSR, so they were obtained from the EHR using ICD-9 and -10 codes (see Appendix A). Follow-up time length was defined as time from surgery to the most recent clinical encounter in the EHR.

Neighborhood characteristics

Social determinants of health:

The latitude/longitude of patient household location at the time of the last follow-up visit was obtained from the EHR and used to analyze three social determinants at the census block group level using geocoded databases: (1) neighborhood socioeconomic deprivation using ADI;⁽¹⁰⁾ (2) neighborhood urbanicity using the rural-urban commuting area (RUCA) database;⁽¹¹⁾ and (3) neighborhood walkability using the walkability index.⁽¹²⁾ A census block group is a subdivision of a census tract containing 600 to 3,000 people allowing for more granular demographic analysis compared to zip codes, for example, that can include up to 114,000 individuals.^(13,14) ADI is a continuous variable from 1 to 100, with 1 being the least deprived and 100 being the most deprived neighborhood. It represents a composite of 17 variables from 2015 U.S. Census data that describe the domains of housing quality, employment, education, and income.⁽¹⁰⁾ RUCA uses working-commuting data from the 2000 U.S. Census to describe neighborhood urbanization with a scale of 1–3 indicating urban/metropolitan; 4–6 suburban/micropolitan; and 7–10 indicating rural.⁽¹¹⁾ The walkability index, developed by U.S. Environmental Protection Agency, is a continuous, composite variable that uses features of the built environment such as intersection density and diversity of land use.⁽¹²⁾ One represents the least walkable neighborhood, while 20 represents the most walkable.

Lifestyle factors -

Patient household location was used to analyze five lifestyle factors from the Esri Community Analyst database, which uses proprietary GIS mapping tools to track and map customer and environmental data.⁽⁶⁾ The data were obtained from the 2018 Survey of the American Consumer,⁽¹⁵⁾ which was conducted by MRI-Simmons and includes an annual sample size of over 24,000 U.S. households and over 2,400 variables. Eisenberg et al. have shown that the survey has high response rates, low levels of missing data, and is similar to U.S. Census data in terms of age, gender, race and ethnicity, income, and health insurance coverage.^(16,17) The questions were directed at the “head of household,” defined as the individual responsible for the majority of budgeting and buying, and were either individual level (e.g., behavioral questions such as diet or exercise) or household level (e.g., type of groceries purchased). The survey responses were geocoded based on census block groups and labeled within existing tapestry segmentation to allow application of the variables across the U.S. for similar tapestries. Tapestry segmentation is the process of classifying neighborhoods into one of 67 distinct consumer-segments in the U.S. that have shared sociodemographic and economic “clusters,” or categories, derived from U.S. Census variables. The potential demand for a good or lifestyle preference was then multiplied by the number of households within that specific tapestry level to get the final demand level for that census block group.⁽⁶⁾

Three members of the study team (NL, LPH, LMF) independently reviewed the Esri data dictionary and subsequently met as a group to establish consensus regarding which variables to include. Five lifestyle factors that were hypothesized a priori to be associated with weight changes following bariatric surgery were selected. Appendix B contains the survey questions used for the construction of each variable.

1. Organic food consumption was a single variable defined by the household consumption level within the last six months (higher values indicated higher product demand).

2. Fresh fruit/vegetable consumption was a single variable defined by the household consumption level within the last six months (higher values indicated more consumption).

3. Soda consumption was a Z-score standardized composite of five variables assessing individual consumption of soda in the last six months (higher values indicated more consumption).

4. Dieting to maintain weight was a Z-score standardized composite of 9 variables assessing whether an individual was presently altering their diet to maintain or lose weight (higher values indicated more alteration of diet).

5. Exercise was a Z-score standardized composite of four variables defined as an individual following a regular exercise routine (higher values indicated a higher ratio of local exercise rate compared to the US rate).^(6,15)

All Esri variables used were standardized around a national average of 100 to quantify how low or high the product or activity demand was in each census block group. A value of < 100 represented less consumption/activity for a neighborhood compared to the national average. For example, a value of 85 implies a demand that is 15% lower than the national average. We note that these variables only provide an ordinal comparison.⁽¹⁸⁾

Spearman correlation coefficients were calculated to identify co-linearity between lifestyle characteristics prior to inclusion in regression models. Differences in neighborhood characteristics between responders (defined as EWL ≥ 50%) and non-responders (defined as EWL < 50%) were examined. Paired T-test was used to determine statistically significant differences between responders and non-responders.

Outcomes

Post-operative outcomes within 90 days and one year of surgery were obtained from the BSR. The 90-day outcomes included all bariatric surgery-related emergency department (ED) visits, readmissions, reoperations, anastomotic/staple line leaks, wound complications, hemorrhage requiring blood transfusion, and an “other complication” variable. One-year postoperative outcomes included anastomotic strictures/sleeve stenosis, revisional operations, and comorbidity resolution (GERD, HLD, HTN, OSA, and T2D).

The EHR was used to identify postoperative BMI, absolute BMI change, %EWL, and percent total body weight (%TBW) loss at the time of the most recent BMI measurement in the EHR. Height and weight data were cleaned using a methodology adapted from Cheng et al.⁽¹⁹⁾ that has been previously described.⁽⁷⁾ No patients were excluded during the data cleaning process.

Statistical analysis

Median regression with three different models was used to evaluate associations between patient and neighborhood characteristics and %TBW loss. Median regression analysis was selected to mitigate the influence of outliers. Model 1 included patient characteristics (age, sex, race and ethnicity, insurance type, preoperative BMI, preoperative T2D, type of surgery [RYGB vs SG], and length of follow-up time) with %TBW loss as the outcome. Model 2 included patient characteristics and the three social determinant variables. As has been described by Kind et al., ADI was modified to be a categorical variable (< 70, ≥ 70) which corresponds to a cut-off of the most socioeconomically disadvantaged 15% of our patient population.⁽¹⁰⁾ Urbanicity and walkability were included as continuous variables in the model. Model 3 included patient characteristics, social determinant variables, and the five lifestyle variables.

SAS Version 9.4 was used for all analyses (SAS Institute Inc.; Cary, NC, USA). The University of Wisconsin-Madison Minimal Risk Institutional Review Board approved this study (#2017-0443). In the reporting and methodology of this study, we followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (Appendix C).⁽²⁰⁾

RESULTS

Patient Characteristics

We identified 768 patients who underwent bariatric surgery from 2008 to 2017. A final cohort of 647 patients had one-year follow-up data available in the EHR and were included in the study (151 patients were excluded due to having undergone adjustable gastric banding or having inadequate follow-up time). Patients had a mean age and pre-operative BMI of 49 years and 46.6 kg/m², respectively. They were more commonly female (77.4%) and White (89.2%). Nearly two-thirds had undergone RYGB (65.5%). OSA, HTN, and GERD were the most common preoperative comorbidities affecting 70.9%, 70.0%, and 55.2% of patients, respectively (Table 1).

Table 1.

Patient characteristics

	n = 647
Age (years), mean (SD)	49.00 (12.2)
Sex (n, %)
Male	146 (22.6)
Female	501 (77.4)
Race and ethnicity (n, %)
White, non-Hispanic	577 (89.2)
Other (non-White)	70 (10.8)
Insurance type (n, %)
Commercial	320 (49.5)
Medicare	172 (26.6)
Medicaid	155 (24.0)
Preoperative BMI (kg/m 2 ), mean (SD)	46.6 (8.0)
Preoperative comorbidities (n, %)
Anxiety	109 (16.9)
Coronary artery disease	63 (9.7)
Depression	178 (27.5)
Gastroesophageal reflux disease	357 (55.2)
Hyperlipidemia	308 (47.6)
Hypertension	453 (70.0)
Obstructive sleep apnea	459 (70.9)
Type 2 diabetes mellitus	400 (46.4)
Preoperative smoking status (n, %)
Never smoked	334 (51.6)
Smoked within past 6 months	21 (3.3)
Former smoker, quit ≥ 6 months prior to surgery	292 (45.1)
Surgery type (n, %)
Sleeve gastrectomy	223 (34.5)
Roux-en-Y gastric bypass	424 (65.5)

Patients were from 366 unique census block groups in the U.S., with 87% of patients residing in Wisconsin. Most lived in urban areas (69.8%) with 64.6% of the cohort residing in three metropolitan areas (Madison, WI; Janesville, WI; Rockford, IL). Table 2 displays the mean values for ADI, urbanicity, walkability index, and the five Esri variables. Relative to the national baseline of 100, patients in our cohort lived in neighborhoods that had lower mean rates of organic food use (88.7±25.6 [mean ±SD]) and exercise (90.7±18.9). Rates of fresh fruit/vegetable consumption (100.1±3.9), dieting to maintain weight (100.6±18.3), soda consumption (98.8±13.3), and fast-food consumption (100.8±12.0) were similar to the national averages. Appendix D and E provide median, interquartile range, range and histograms to further characterize the distribution of the neighborhood-level variables.

Table 2.

Social determinants of health at the neighborhood level

	n = 647
Area Deprivation Index (ADI), mean (SD) a,b	46.3 (20.2)
Urbanicity (RUCA) (n, %) a
Urban/Metropolitan	451 (69.8)
Suburban/Micropolitan	79 (12.2)
Rural	116 (18.0)
Walkability Index, mean (SD) a,c	8.1 (3.7)
Esri variables, mean (SD) d,e
Organic food use	88.7 (25.6)
Fresh fruit/vegetable consumption	100.1 (3.9)
Diet to maintain weight	100.6 (18.3)
Soda consumption	98.8 (13.3)
Exercise	90.7 (18.9)

^aData missing from n=1 due to no address

^bADI ranges from 1 to 100, with 1 being the least deprived and 100 being the most deprived.

^cWalkability ranges from 1 to 20 with 1 representing the least walkable neighborhood and 20 representing the most walkable.

^dData missing from n=4 due to no FIPS code, and patients from Alabama or Wyoming

^eEsri variables have a baseline national average set at 100. Values less than 100 correspond to less product use, diet, or exercise.

Organic food use, fresh fruit and vegetable consumption, diet to maintain weight, and exercise were positively associated with each other, while soda consumption was negatively associated with the other four lifestyle factor variables (Appendix F). There were no significant differences for neighborhood-level characteristics between patients who lost ≥ 50% versus those who did not (Appendix G).

Outcomes

Patients had a median follow-up time of 3.1 years. Within 90 days of surgery, 25.0% of patients had visited the ED, 14.6% had been re-admitted, and 2.8% underwent a reoperation. At 1-year post-operatively, co-morbidity resolution was highest for T2D (51.0%) and GERD (38.1%). At the time of the most recent clinical encounter, the mean post-operative BMI was 36.1 kg/m² with a mean absolute BMI change of 10.4 kg/m², mean %TBW loss of 22.2%, and mean %EWL of 44.3% (Table 3).

Table 3.

Postoperative outcomes

	n = 647
90-day outcomes (n, %)
Anastomotic/staple line leak	6 (0.9)
Hemorrhage	11 (1.7)
Wound complications	23 (3.6)
Other complicationsc	40 (6.2)
Reoperations	18 (2.8)
Emergency department visits	161 (25.0)
Readmissions	94 (14.6)
1-year comorbidity resolution (n, %) a
Gastroesophageal reflux disease	136 (38.1)
Hyperlipidemia	80 (26.8)
Hypertension	145 (32.0)
Obstructive sleep apnea	155 (33.8)
Type 2 diabetes mellitus	153 (51.0)
1-year outcomes (n, %)
Revisions	19 (2.9)
Anastomotic strictures	28 (4.3)
Sleeve stenosis	5 (0.8)
Weight loss outcomes
Length of follow-up (years), median (IQR)	3.1 (3.4)
Postoperative BMI (kg/m2), mean (SD)b	36.1 (8.5)
Absolute BMI change, mean (SD)b	10.4 (6.6)
% Total body weight loss, mean (SD)b	22.2 (12.9)
% Excess weight loss, mean (SD)b	44.3 (26.9)

^aComorbidity resolution percentage is calculated from individuals with comorbidity pre-operatively

^bRepresent data from most recent clinical encounter

^c“Other complications” includes acute renal failure, cerebral vascular accident, deep vein thrombosis, myocardial ischemia, pneumonia, pulmonary embolism, and urinary tract infection

Predictors of Weight Loss after Bariatric Surgery

In median regression analysis with %TBW loss at the time of most recent follow-up as the outcome and patient characteristics as the independent variables (Model 1), RYGB was associated with 10.82%, 95% CI [8.65, 13.00] more TBW loss compared to SG, whereas longer follow-up time (−1.06% TBW loss per year of follow-up time, 95% CI [−1.61, −0.51]) and a pre-operative diagnosis of T2D (−2.77% TBW loss, 95% CI [−4.77, −0.77]) were associated with less %TBW loss. Age, sex, race and ethnicity, insurance type, and preoperative BMI were not associated with postoperative weight loss. In the fully adjusted median regression model (Model 3), longer follow-up time (−2.42% TBW loss per year of follow-up time, 95% CI [−4.63, −0.20]) and a pre-operative diagnosis of T2D (−1.00% TBW loss, 95% CI [−1.55, −0.44]) were associated with less %TBW, whereas RYGB was associated with 11.22%, 95% CI [8.96, 13.48] more TBW loss compared to SG. Neighborhood socioeconomic deprivation, urbanicity, walkability, and the five lifestyle factors were not associated with postoperative weight loss (Table 4).

Table 4.

Median regression with %TBW loss as the outcome

	Model 1	Model 2	Model 3
Variable	Median % TBW loss (95% CI)	Median % TBW loss (95% CI)	Median % TBW loss (95% CI)
Patient characteristics
Age (+5 yrs.)	−0.08 (−0.60, 0.45)	−0.17 (−0.71, 0.37)	−0.24 (−0.85, 0.37)
Female sex	0.32 (−1.97, 2.61)	0.62 (−2.08, 3.32)	0.60 (−2.25, 3.44)
Race and ethnicity
Non-White	Ref	Ref	Ref
White, non-Hispanic	−0.09 (−3.52, 3.35)	0.61 (−3.17, 4.39)	−0.15 (−4.49, 4.19)
Insurance type
Private/self-pay	Ref	Ref	Ref
Medicare	−0.54 (−3.48, 2.40)	−0.71 (−3.75, 2.34)	−0.31 (−3.19, 2.57)
Medicaid	1.24 (−0.83, 3.30)	1.22 (−0.94, 3.38)	0.96 (−1.47, 3.40)
Preoperative BMI	0.11 (−0.03, 0.26)	0.13 (0.00, 0.27)	0.07 (−0.10, 0.23)
Preoperative type 2 diabetes	−2.77 (−4.77, −0.77)	−2.70 (−4.80, −0.60)	−2.42 (−4.63, −0.20)
Roux-en-Y gastric bypass (vs. sleeve gastrectomy)	10.82 (8.65, 13.00)	10.81 (8.35, 13.27)	11.22 (8.96, 13.48)
Follow-up time (+ 1 yr.)	−1.06 (−1.61, −0.51)	−1.07 (−1.58, −0.56)	−1.00 (−1.55, −0.44)
Neighborhood characteristics
Area deprivation indexa
Low (< 70)	--	Ref	Ref
High (≥ 70)		1.17 (−1.80, 4.13)	−0.02 (−3.62, 3.57)
Urbanicity	--	−0.20 (−0.58, 0.18)	−0.36 (−0.91, 0.18)
Walkability	--	−0.02 (−0.40, 0.35)	−0.09 (−0.47, 0.29)
Organic food	--	--	−0.02 (−0.13, 0.09)
Fresh fruit/vegetable consumption	--	--	−0.21 (−0.66, 0.23)
Diet to maintain weight	--	--	0.04 (−0.06, 0.14)
Soda consumption	--	--	−0.01 (−0.17, 0.14)
Exercise	--	--	−0.03 (−0.14, 0.08)

Abbreviations: % TBW = percent total body weight loss; BMI = body mass index

^aArea deprivation index (ADI) is modified to be a categorical variable (< 70, ≥ 70) which corresponds to a cut-off of the most socioeconomically disadvantaged 15% of our patient population

DISCUSSION

Our findings suggest that patient characteristics, such as preoperative diagnosis of T2D and length of follow up time, along with type of surgery were associated with weight loss, while neighborhood socioeconomic deprivation, urbanicity, walkability, and lifestyle factors were not.

Neighborhood-level socioeconomic deprivation, assessed by ADI, was not associated with weight loss after bariatric surgery. Higher socioeconomic deprivation has been linked to increased risk of hospital readmission,⁽¹⁰⁾ increased prevalence of obesity,⁽²¹⁾ worse surgical outcomes following colon and rectal surgery,⁽²²⁾ and higher “no show” rates for pre- and post-operative appointments.⁽²³⁾ Although we did not find a statistically significant association of ADI with weight loss after bariatric surgery, it has been shown that less affluent individuals have lower rates of bariatric surgery^(24–26) and that lower socioeconomic status patients must navigate more barriers to access care, transportation, and healthy food options.^(27,28) If outcomes can be comparable regardless of ADI, this study and others like it may serve to expand surgery access in underserved areas (e.g., Esri mapping to identify discrepancies in patient geographical distribution versus estimated bariatric surgery need) and to identify public health concerns at a neighborhood-level for intervention to help mitigate barriers patients are encountering.

We did not identify associations between the five Esri lifestyle variables that were examined and weight loss outcomes. To our knowledge, no studies have examined relationships between Esri lifestyle variables and medical or surgical outcomes. Lifestyle variables from the Survey of the American Consumer have been used to examine the influence of advertising on vitamin demand⁽¹⁶⁾ and soda consumption.⁽¹⁷⁾ Furthermore, Kantor et al. published a case report examining the use of Esri Community Analyst, a platform that combines GIS mapping with census, lifestyle, and business variables, to identify vulnerable populations. The study showed that lifestyle variables and census data combined with Esri tapestry segmentation can be used to identify county areas at increased health risk due to dietary limits.⁽²⁹⁾ Although our project did not identify associations between these neighborhood level variables and weight loss, the use of Esri lifestyle variables provides the opportunity to investigate neighborhoods at a more granular level than typical zip code analyses. Furthermore, they facilitate assessment of the built environment’s relationship with surgical outcomes. For example, studying the walkability of a neighborhood is feasible, while examining walkability of a zip code is not relevant. Studies incorporating Esri consumer variables at the individual-level are needed to identify variables that are correlated with obesity risk. The identification of variables at the census block group level correlated with obesity and/or worse health outcomes also has public health and resource allocation implications. Associating poor access or low utilization (e.g., low produce consumption that is linked with low grocery store density) to an impact on the health of individuals provides additional data that is actionable.

In our study, neighborhood walkability and urban versus rural status were not associated with weight loss. We had hypothesized that neighborhood walkability would be associated with greater weight loss because neighborhoods that are easier to walk should be easier to exercise in. Additionally, neighborhoods that are more walkable are associated with lower rates of obesity.⁽³⁰⁾ Similarly, we had hypothesized that suburban neighborhoods would be associated with greater weight loss compared to rural or highly urban areas. In a study of 50 bariatric surgery patients in West Virginia, living is a rural versus suburban/urban environment has previously been shown to not be associated in differences in weight loss outcomes at a year.⁽³¹⁾ We suspect that these variables were not associated with weight loss in our study because they may be inaccurate, or too broad, when applied at the neighborhood-level. For example, a patient may live in a suburban area with no sidewalks.

Several patient-level characteristics were associated with weight loss. Our finding of RYGB being strongly associated with more weight loss is consistent with well-established literature.^(32,33) A preoperative diagnosis of diabetes was associated with less weight loss. The literature regarding the association of T2D with weight loss is mixed with some finding less weight loss⁽³⁴⁾ and others greater weight loss.⁽³⁵⁾ Regarding the follow-up interval, longer intervals were associated with less weight loss. This is consistent with other studies that have described weight regain after the 1–2 year postoperative period.^(32,33) However, the degree and timing of weight regain is difficult to fully assess given the heterogeneity of weight regain measures, timing of assessment, study design, and surgical procedures.⁽³⁶⁾ Our findings support the incorporation of individual-level variables, especially the type of bariatric surgery, into predictive outcome algorithms for bariatric surgery. In addition to examining weight loss outcomes, future studies should determine if differences in co-morbidity resolution, complication rates, follow-up care attendance, and type of bariatric surgery are associated with neighborhood-level variables.

Our study has several limitations. First, the Esri survey question constructions were designed to capture consumer habits rather than associations with individual health. For instance, the fresh fruit and vegetable consumption variable is at the household level, so it may not reflect an individual’s proportion of consumption accurately. Second, Esri proprietary restrictions may limit the interpretability of composite variables as the weighting is not disclosed. Third, our findings may not be generalizable to bariatric populations due to differences in race and ethnicity, insurance and surgery type, neighborhood characteristics. Although we did not find an association between neighborhood-level factors, repeating this analysis in neighborhoods different regions of the U.S. or greater socioeconomic diversity may find an association. Fourth, patients may have moved during the study period, and this would not have been captured in our dataset. Finally, there may be observational bias due to unmeasured confounding since this is a retrospective study.

In conclusion, neighborhood characteristics including Esri lifestyle variables, socioeconomic deprivation, urbanicity, and walkability were not associated with differences in weight loss in the neighborhoods evaluated. These findings suggest that patients living in disadvantaged neighborhoods similar to those in our study can achieve comparable weight loss outcomes if they are appropriately selected for surgery. Additional research is needed to further explore the use of consumer databases such as Esri in predicting patient outcomes, which would be ideally suited to machine learning method-based studies that can incorporate large numbers of predictive variables into their algorithms.

HIGHLIGHTS.

• Neighborhood characteristics that influence bariatric surgery outcomes are unknown

• 647 patients had an average of 22% total body weight (TBW) loss & 3 years follow-up

• The 8 neighborhood-level variables were not associated with %TBW loss

• Surgery type, follow-up length, & diabetes were associated with %TBW loss