Inequalities in bariatric surgery in Australia: findings from 49 364 obese participants in a prospective cohort study

Obesity is a major public health challenge for Australia. In the 2007–08 National Health Survey, 24% of Australian adults were reported to be obese and a further 37% overweight.1 Obesity rates are growing and the continuing increase in severe obesity is of particular concern.2,3 It is a major risk factor for type 2 diabetes and a range of other chronic diseases, including cardiovascular, digestive and musculoskeletal disorders,4 as well as overall mortality.5

Clinical guidelines recommend bariatric surgery for those with a body mass index (BMI) over 40 kg/m2, or BMI over 35 kg/m2 and comorbid conditions, after non-surgical options have failed.6,7 This surgery is more effective than non-surgical interventions for the treatment of severe obesity, and it is cost-effective. In addition to substantial weight loss, bariatric surgery can lead to improvements in comorbid conditions including lipid abnormalities, obstructive sleep apnoea and joint disease.8-11 Of particular note is its effectiveness in treating type 2 diabetes,12,13 with one recent trial showing remission rates of 75%–95% within 2 years after surgery.13

Bariatric surgery procedures have been listed on the Medicare Benefits Schedule (MBS) since 1992. Admissions for this surgery rose from 535 to around 17 000 between 1998–1999 and 2007–2008.14 Notably, most of this surgery is carried out in private hospitals and incurs substantial out-of-pocket costs,14 while obesity is concentrated among those of lower socioeconomic status (SES).15 This suggests that groups that are most likely to need surgery are the least likely to have it. However, despite equity concerns,16 there are no published data to date in Australia on the extent of variation in bariatric surgery by health status, SES and other key factors among those potentially eligible for the procedure. In this study, we investigate variation in primary bariatric surgery rates in an obese population, quantify socioeconomic inequalities in rates of surgery and examine the extent to which holding private health insurance (PHI) explains these inequalities.

We used data from the 45 and Up Study, a cohort study involving 266 848 men and women aged 45 years and over from New South Wales. Study participants were randomly sampled from the Medicare enrolment database. More than 10% of the NSW population aged 45 years and over is included in the cohort (response rate of about 18%).17 Participants received a baseline questionnaire (between 1 January 2006 and 31 December 2008) and gave signed consent for follow-up, including linkage to routine health databases. The study is described in detail elsewhere,17 and questionnaires can be viewed at http://www.45andup.org.au.

Questionnaire data were linked to death data from the NSW Registry of Births, Deaths and Marriages (to 30 June 2010) and to hospital data from the NSW Admitted Patient Data Collection (APDC) from 1 July 2000 to 30 June 2010. The NSW APDC includes records of all hospitalisations in NSW, including reasons for admission (coded using ICD-10-AM) and procedures performed (coded using the Australian Classification of Health Interventions).18 Data were linked probabilistically by the Centre for Health Record Linkage (http://www.cherel.org.au).

The current study included only participants who were obese (BMI ≥ 30 kg/m2), with BMI calculated from weight and height as self-reported on the questionnaire. We excluded anyone who had had previous bariatric surgery recorded in the APDC (ie, between July 2000 and recruitment). The outcome was incident primary bariatric surgery for obesity, defined as the first bariatric surgery procedure recorded after recruitment, identified from the procedure fields in the APDC. Procedures included adjustable gastric banding or gastroplasty (procedure code 30511) or gastric bypass (30512). Partial gastrectomy (30518) may also be used for the treatment of obesity, but we did not include this as it is mostly used for other indications, and our sample included only two such procedures. Participants were followed from the date of recruitment to either the date of admission for bariatric surgery, death, or 30 June 2010, whichever occurred first.

Data on participant characteristics were based on self-reported data from the questionnaire. Variables were categorised as shown in Box 1 and Box 2. Socioeconomic variables included annual pre-tax household income, education and area-level disadvantage. Area-level disadvantage was based on the Socio-Economic Indexes for Areas Index of Relative Socio-Economic Disadvantage (IRSD),19 derived from postcode of residence and categorised into quintiles using cut-off scores from the 2006 Australian census. Other variables included PHI (including holders of a Department of Veterans’ Affairs [DVA] card), BMI, sex, age group, area of residence (based on the Accessibility/Remoteness Index of Australia Plus,20 derived from postcode), marital status, country of birth, self-rated health, diabetes (ever diagnosed by a doctor), number of other doctor-diagnosed chronic conditions, smoking, tertile of physical activity (based on number of weekly sessions of walking and moderate and vigorous activity, weighted for intensity) and alcohol intake.

Negative binomial regression was used to estimate bariatric surgery rates according to baseline characteristics and to model inequality estimates. We used separate multivariable regression models for the two main SES variables of interest — household income and area-level disadvantage. We calculated rate ratios (RRs) for each socioeconomic level using the lowest level as the reference group, adjusting for all other non-SES variables (Model 1). In Model 2, we added PHI. We then quantified the extent to which PHI explained any socioeconomic variation in bariatric surgery rates by testing for equality of the SES coefficients across Models 1 and 2. Stata version 12.1 (StataCorp) was used for all analyses.

Ethics approval for this project was obtained from the NSW Population and Health Services Research Ethics Committee and the Australian National University Human Research Ethics Committee.

Survey and linked hospital and death data were available for 266 724 of the 266 848 current participants in the 45 and Up Study. After excluding those who had BMI data missing (20 262 participants; 7.60%) and those who had had bariatric surgery before recruitment (17 participants), there were 55 038 participants (22.33%) with BMI ≥ 30 kg/m2 who were eligible for this study. As no one over the age of 74 in this sample had bariatric surgery in the follow-up period, we confined our analysis to those aged less than 75 years (49 364 participants).

A total of 312 participants had surgery over 111 757 person-years (py) of follow-up (mean, 2.26; SD, 0.86), giving a rate of 27.92 (95% CI, 24.91–31.19) per 10 000 py. Of these, only one was treated as a public patient and four as DVA patients, with the remainder treated as private patients. The mean BMI (at baseline) of those having surgery was 39.15 kg/m2. The principal diagnosis was recorded as obesity (ICD-10 code E66) in 261 patients (84%) and as diabetes (E10 or E11) in 45 patients (14%). Only six of the 312 procedures were bypass procedures, the remaining 98% being gastric banding or gastroplasty.

Descriptive data showing bariatric surgery rates in relation to participant baseline characteristics are shown in Box 1 and Box 2. Rates of surgery increased with increasing BMI, ranging from 3.72 per 10 000 py (BMI 30–32.49 kg/m2) to 227.77 per 10 000 py (BMI 45–50 kg/m2). Rates varied significantly in relation to all participant characteristics except country of birth and marital status (P > 0.05). Higher rates were associated with being female, younger, a resident in a major city, in poorer health, a non-smoker, a non-drinker and being in the lowest tertile of physical activity.

With regard to SES, unadjusted rates (Box 2 and Box 3) and age–sex- adjusted RRs (Box 2) show that bariatric surgery rates were higher among those who were relatively advantaged. There was a clear socioeconomic gradient with household income; for IRSD, the most notable difference was between the top quintile (low disadvantage) and the other quintiles; for education, rates were highest among those with post-school (non-trade) qualifications and lowest in those with no qualifications; and rates were much higher among those with PHI than among those without.

The degree of socioeconomic in-equality in bariatric surgery rates, after adjusting for all variables except PHI (Model 1), was substantial (Box 4). The adjusted RRs for household income show a clear gradient, with those in the highest bracket (≥ $70 000) five times more likely to have surgery than those in the lowest bracket (< $20 000) (RR, 5.27; 3.18–8.73). After adjusting for PHI (Model 2), the RRs decreased by 35%–62% (P < 0.001 for all income levels), confirming that PHI explained a substantial proportion of income-related inequality. Nevertheless, significant inequality remained, with those in the highest income bracket still being almost twice as likely to have bariatric surgery as those in the lowest bracket (RR, 1.98; 1.15–3.41). When income and education were jointly modelled, this made virtually no difference to the income inequality estimates, while education inequality estimates were not significant in either Model 1 or 2 (results not shown).

Rates of surgery by IRSD quintile show that those living in areas of least disadvantage were four times more likely to have surgery than those living in the most disadvantaged areas (RR, 4.01; 2.41–6.67), after taking into account potential confounding factors (Model 1). After adjusting for PHI, the RR for each quintile of disadvantage decreased by 12%–40% (P < 0.001 for all quintiles). However, significant inequality remained, with those in the least disadvantaged areas still being over twice as likely to have bariatric surgery than those in the most disadvantaged areas (RR, 2.41; 1.48–3.93).

There is significant inequality in the uptake of bariatric surgery among obese people in Australia, with the likelihood of surgery increasing with increasing SES. Even when measured using an area-level measure of disadvantage, and adjusting for remoteness and other factors, the magnitude of inequality is substantial. Of particular note is the fivefold higher rate of surgery in those with household incomes of $70 000 or more, compared with that of those with household incomes less than $20 000. PHI accounted for some but not all of the observed SES inequalities. While people with higher education qualifications were twice as likely to have surgery as those with no qualifications, much of this was because of the association between education and income.

Our inequality findings differ from a previous report that showed that bariatric surgery rates in the middle SES quintile of area disadvantage were more than double those of any other SES quintile;14 however, this report was based on the whole population, not the obese population, and hence did not take into account the “need” for surgery. Our findings that bariatric surgery is more common among women, middle-aged rather than older people, and among those living in major cities are consistent with previous reports.14 In addition, the variation in rates we found in relation to health characteristics was in keeping with the indications for surgery7 — the likelihood of surgery increased with increasing BMI, and was greater among those with poor health, diabetes and other chronic conditions. We also found that current smokers were less likely to have surgery than non-smokers.

Strengths of this study include its grounding in a very large population-based cohort, allowing a relatively rare event to be examined; investigation of a large range of factors not recorded in routine data; and use of linked administrative records, allowing virtually complete and objective ascertainment of surgery. A limitation is that BMI was based on self-reported weight and height. However, a validation study involving participants in the 45 and Up Study found that the mean difference between self-reported and measured BMI was not large (on average, 0.74 kg/m2), with sensitivity for classifying obesity of 79%, and importantly, specificity of 99%.21 Although the relatively low response rate and the potential for a “healthy cohort effect” mean that the estimates of surgery rates in our sample may be different to those of the general population, relative comparisons of surgery rates among groups within the cohort remain valid.22,23 Some caution must be applied, however, in generalising the size of the inequality estimates to younger ages, and beyond NSW, which has the highest proportion of private hospital weight loss procedures of all Australian jurisdictions.14

There are many potential barriers to bariatric surgery, apart from cost, that may underlie variations in uptake of surgery. These include patients’ preferences and clinical decisions regarding the suitability of patients for surgery, and possibly views by some that bariatric surgery is largely cosmetic. However, the observed SES-related inequality in rates of surgery is also likely to reflect system-wide issues, including the mix of public and private care, out-of-pocket costs, limited resources and cost-sharing between state and federal governments. Moreover, the current situation is that there is very limited availability of bariatric surgery in public hospitals, while Medicare subsidises bariatric surgery and post-surgical care for private patients, effectively restricting access to people with PHI and those who can afford to pay what are usually large associated out-of-pocket costs.

In 2009 the House of Representatives Standing Committee on Health and Ageing Inquiry into Obesity recommended that equity in access be ensured by publicly funding bariatric surgery.24 Our findings suggest that bariatric surgery, an MBS-listed procedure, is currently largely available only to those who can afford PHI and the associated out-of-pocket costs, with poor access to these cost-effective procedures in the section of the population that is most in need. Continuing inequity in access is likely to exacerbate existing inequalities in obesity and related health problems. However, if bariatric surgery came to be less discretionary over time, particularly for the treatment of type 2 diabetes,25 such inequalities could decline. While resource issues may limit the total number of patients that can have bariatric surgery, there is scope to consider how the distribution of limited supply can be improved, and the potential savings that could be made from increasing supply and improving health outcomes.

3 Rates of bariatric surgery in relation to household income, education level and area-level disadvantage* in 49 364 participants with body mass index ≥ 30 kg/m2

* Area-level disadvantage was based on the Socio-Economic Indexes for Areas Index of Relative Socio-Economic Disadvantage (IRSD),19 derived from postcode of residence and categorised into quintiles using cut-off scores from the 2006 Australian census.