OUP user menu

Increasing socio-economic inequality in type 2 diabetes prevalence—Repeated cross-sectional surveys in England 1994–2006

Anne K. Imkampe, Martin C. Gulliford
DOI: http://dx.doi.org/10.1093/eurpub/ckq106 484-490 First published online: 4 August 2010


Background: This study aimed to evaluate the association of rising type 2 diabetes prevalence with socio-economic inequality in diabetes. Methods: Data from the Health Survey for England were analysed for 1994, 1998, 2003 and 2006. This is a nationally representative annual survey of private households. Data for 41 643 individuals aged ≥35 years were included. The prevalence of self-reported diabetes diagnosed by a doctor was analysed in relation to household income, occupational social class and educational qualifications. Data were standardized for age using the European Standard Population for reference. Results: Prevalence of diagnosed diabetes increased in men from 3.74% in 1994 to 7.25% in 2006, and in women from 2.28% to 4.88%. In 1994, there were no associations between social class or educational level and diabetes prevalence evident. In 2006, there was evidence of a negative association in women [prevalence ratio for social class (IV + V vs. I) = 4.54, P-value for trend = 0.005; prevalence ratio for educational level (‘none’ vs. ‘A-levels’) = 1.96, P-value for trend = 0.001]. The Slope Index of Inequality (SII) for social class in women increased from –1.65 in 1994 to –4.95 [95% Confidence Interval (95% CI –8.52 to –1.38)] in 2006 and for level of education from –1.39 to –6.48 (95% CI –9.03 to –3.93). In men, diabetes prevalence was not associated with social class or level of education. Conclusion: Increasing prevalence of type 2 diabetes has been associated with an increase of socio-economic inequality in women. There was no socio-economic gradient observed in men.

  • socio-economic inequality
  • trends over time
  • type 2 diabetes prevalence


The prevalence of type 2 diabetes has increased rapidly over the last three decades.1 The International Diabetes Federation estimated that in 2006, 246 million people were living with diabetes worldwide2 with 80% of people with diabetes now in low- and middle-income countries.3 However, the disease burden is also increasing in high-income countries. In the UK, Gonzalez et al.4 examined data from primary care medical records and reported a rise in type 2 diabetes prevalence from 2.47% to 3.9% between 1996 and 2005. Thomas et al.5 report an age-adjusted prevalence increase from 4.3% between 1979 and 1984 to 11.8% between 2003 and 2005 in British men aged 40–59 years. Type 2 diabetes prevalence is predicted to increase further until 2036.6

The existence of a social gradient in health in general is well recognized and has been defined by Marmot as a stepwise or linear decrease in health with decreasing social position.7 Mackenbach et al.8 reviewed data on socio-economic inequality in health in 22 European countries and confirmed existing health inequalities by socio-economic status in all countries whereby the magnitude of inequality was variable between countries.

Health inequalities in the prevalence of type 2 diabetes have also been demonstrated. The disease affects all socio-economic groups but is generally more frequent in lower socio-economic groups.9,10 Espelt et al.11 reported inequalities in diabetes morbidity and mortality from 13 countries across Europe. Data from the British Women’s Heart and Health Study12 that investigated elderly women aged 60–79 years showed that area deprivation independently influenced the diagnosis of type 2 diabetes. Health outcomes in individuals suffering from diabetes appear similarly related to socio-economic position and Brown et al.13 developed a conceptual framework highlighting possible mechanisms. Their model include proximal mediators such as individual health behaviours, health care access and process of care as well as distant mediators that entail individual, provider, community and health care system factors.

Modifiable risk factors for diabetes, including obesity, physical inactivity and cigarette smoking, are themselves subject to socio-economic inequalities. Socio-economic disadvantage has been associated with high smoking rates and low levels of leisure time physical activity in both, men and women;14–16 a higher obesity prevalence has been described for women of low socio-economic position.17,18 However, the association of diabetes with socio-economic position is only partly explained by known risk factors for diabetes such as obesity, exercise, alcohol, smoking, ethnicity or family history.19,20

In recent years, there have also been efforts to improve case finding and facilitate earlier diagnoses of diabetes. The charity, Diabetes UK, for example, launched a campaign to raise awareness of the disease and its seriousness among the British public as well as among health professionals.21 Consistent with other preventive medical efforts that tend to increase inequalities,22 this might have the effect of increasing the prevalence of diagnosed diabetes in higher socio-economic groups that are better able to access preventive medical services.

The aim of this study, therefore, was to evaluate the association of rising prevalence with socio-economic inequalities in type 2 diabetes. We specifically aimed to determine whether socio-economic inequalities in diagnosed diabetes are reducing or increasing over time.


Data source: the health survey for England

The Health Survey for England (HSE) is carried out annually and aims to provide data about the nation’s health, the prevalence of specific health conditions and the prevalence of risk factors for disease. Cardiovascular risk factors, including diabetes, have been an important focus of the survey in several years. The HSE is a nationally representative survey of the English population. It uses a multi-stage cluster sampling design with post-code sectors as the primary sampling unit.23 The sampling methods have been consistent since its inception in 1991, but sample sizes increased from initially 3000 adults to about 16 000 from 1993 (children excluded). In this study, data from the 1994, 1998, 2003 and 2006 survey were analysed because the survey focused on cardiovascular disease in these years and the collected data provided information on diabetes prevalence and associated risk factors. The Household response rates were 77% in 1994, 74% in 1998, 73% in 2003 and 68% in 2006. The response rates of adults in co-operating households were, respectively, 92, 92, 90 and 88%.

Data preparation

There were 41 643 individuals ≥aged 35 years included in the analyses. Individuals who answered ‘don’t know’ or ‘no answer’ were coded for diabetes status unknown. Subjects <35 years were excluded because diabetes is less frequent, and type 1 diabetes represents a higher proportion of cases, at young ages. The prevalence of diabetes was defined from the interview-administered questionnaire. If a respondent stated he/she had diabetes this was confirmed with the question ‘Were you told by a doctor that you have diabetes?’ Only participants who had answered this question positively were categorized as having diabetes in this analysis. Diabetes during pregnancy was excluded. There was no question differentiating between types 1 and 2 diabetes in the survey and patients with diabetes were, therefore, combined as one group. The distinction between the two types can be difficult owing to the increasing trend of childhood obesity and diagnosis of type 2 diabetes at earlier ages.24 It is known that >95% of patients with diabetes have type 2 disease and that type 1 diabetes is less frequently diagnosed at ages >35 years. It is therefore assumed for the relationships found in this study that they are in the majority reducible to type 2 diabetes.

Socio-economic status was assessed by social class, education and household income. For social class, the ‘Registrar General’s Classification’ was used, based on occupation of the individual respondent. Level of education was categorized as: (i) A-levels or higher degrees, (ii) O-levels/GCSE or equivalent degrees, (iii) other degrees and (iv) no degrees. Household income was considered as equivalized household income in quintiles. Each year some modifications had been made to the survey questionnaire and extra questions had been added. Data on household income were only available from 1998.

Statistical analysis

Stata version SE11 was used for statistical analyses. Diabetes prevalence was calculated separately for men and women. Age-standardized prevalences were estimated using the European Standard Population for reference. For determination of independent associations of risk factors with diabetes prevalence, a log-binomial regression model was fitted with whether or not the participant had diabetes as the binary outcome variable. For this regression model, the ‘binreg’ command in Stata was used.

The Slope Index of Inequality (SII) and the Relative Index of Inequality (RII) was used to quantify socio-economic inequalities. The SII was calculated as described by Mackenbach and Kunst25 and compared by year of survey. The gradient was determined with linear regression whereby the relative rank of the socio-economic measurement factor was taken as predictor variable and the respective age-adjusted diabetes prevalence as outcome variable. The formula: Embedded Image was used to account for heteroskedasticity of the error terms (Y = diabetes rate of respective socio-economic class; a = size of socio-economic class; b = relative rank of socio-economic class).26

For the relative rank of social class, the cumulative proportion of the sample in each category of social class was estimated. The midpoint was used as the code for the respective social class. For example, the lowest social class category in women included 24.16% in 2006 and had therefore a value of 0.121. The highest social class category in this group included 3.3% and was assigned the value 0.984.

For estimation of the RII, the SII was divided by the overall diabetes prevalence by gender of the respective year.


Number of included respondents

After exclusion of individuals <35 years old, 10 600 subjects were included from 1994 of which 346 had diabetes and 12 had diabetes status unknown; in 1998, there were 11 059 subjects of which 421 had diabetes and two had diabetes status unknown and in 2003, 10 890 subjects were included of which 585 had diabetes. In 2006, 1592 subjects were excluded from the diabetes question. A total of 9094 subjects were included, of which 524 had diabetes and two had diabetes status unknown. This resulted in 41 643 subjects of whom 18 699 were men and 22 944 were women, 1876 had diabetes and in 16 diabetes status was unknown.

Diabetes prevalence

Age-standardized prevalence of diabetes diagnosed by a doctor was generally higher in men than in women, and increased over time. In men there was a rise from 3.74% in 1994 to 7.25% in 2006 and in women from 2.28% in 1994 to 4.88% in 2006.

In table 1, diabetes prevalence is compared between age-groups and socio-economic groups: social class, level of education and household income. The results show that the observed increase in prevalence was present in both genders, in all age- and most socio-economic groups considered.

View this table:
Table 1

Diabetes prevalence (crude prevalences) compared by year of survey and sex

Number189/4776 (3.96)223/4980 (4.48)309/4831 (6.40)290/4112 (7.05)157/5824 (2.70)198/6079 (3.26)276/6059 (4.56)234/4982 (4.70)
Age-group (years)
    35–4413/1329 (0.98)21/1305 (1.61)33/1263 (2.61)27/1183 (2.28)13/1520 (0.86)14/1573 (0.89)24/1618 (1.48)17/1494 (1.14)
    45–5428/1127 (2.48)37/1289 (2.87)38/1101 (3.45)61/1050 (5.81)20/1300 (1.54)24/1484 (1.62)32/1279 (2.50)44/1279 (3.44)
    55–6464/1001 (6.39)57/987 (5.78)88/1103 (7.98)93/1126 (8.26)27/1059 (2.55)36/1148 (3.14)61/1307 (4.67)74/1269 (5.83)
    65–7451/877 (5.82)59/837 (7.05)95/807 (11.77)67/437 (15.33)54/1120 (4.82)64/967 (6.62)79/952 (8.30)49/470 (10.43)
    75–8428/370 (7.57)39/464 (8.41)47/460 (10.22)39/251 (15.54)40/649 (6.16)42/714 (5.88)64/719 (8.90)41/355 (11.55)
    ≥855/72 (6.94)10/98 (10.20)8/97 (8.25)4/66 (6.06)3/176 (1.70)18/193 (9.33)16/184 (8.70)9/116 (7.76)
    I10/345 (2.90)10/369 (2.71)18/386 (4.66)24/349 (6.88)1/75 (1.33)1/99 (1.01)0/137 (0.00)1/139 (0.72)
    II47/1397 (3.36)70/1571 (4.46)90/1522 (5.91)85/1435 (5.92)24/1317 (1.82)21/1330 (1.58)51/1655 (3.08)46/1519 (3.03)
    III non-manual31/460 (6.74)19/459 (4.14)30/455 (6.59)22/416 (5.29)44/1961 (2.24)57/2037 (2.80)83/1942 (4.27)67/1572 (4.26)
    III manual63/1632 (3.86)69/1613 (4.28)99/1554 (6.37)95/1173 (8.10)21/522 (4.02)30/550 (5.45)37/468 (7.91)22/369 (5.96)
    IV + V34/886 (3.84)51/911 (5.60)63/848 (7.43)61/684 (8.92)60/1734 (3.46)69/1800 (3.83)87/1602 (5.43)85/1191 (7.14)
    A-levels45/1528 (2.95)59/1802 (3.27)97/1911 (5.08)115/2016 (5.70)14/1100 (1.27)11/1399 (0.79)39/1761 (2.21)37/1912 (1.94)
    GCSE/O-level24/665 (3.61)31/845 (3.67)45/974 (4.62)43/787 (5.46)17/883 (1.93)17/1210 (1.40)44/1369 (3.21)42/1216 (3.45)
    Other25/619 (4.04)28/553 (5.06)45/485 (9.28)23/253 (9.09)17/720 (2.36)21/647 (3.25)24/671 (3.58)19/318 (5.97)
    None94/1952 (4.82)103/1767 (5.83)121/1450 (8.34)108/1035 (10.43)109/3106 (3.51)149/2814 (5.29)169/2244 (7.53)134/1524 (8.79)
Household income quintiles
    I (high)26/921 (2.82)34/934 (3.64)39/844 (4.62)7/881 (0.79)16/904 (1.77)15/848 (1.77)
    II39/913 (4.27)45/914 (4.92)42/782 (5.37)13/983 (1.32)22/980 (2.24)25/858 (2.91)
    III52/1026 (5.07)55/878 (6.26)54/695 (7.77)29/1154 (2.51)44/1129 (3.90)26/841 (3.09)
    IV43/767 (5.61)59/685 (8.61)55/592 (9.29)63/1177 (5.35)72/974 (7.39)68/857 (7.93)
    V (low)34/671 (5.07)60/703 (8.53)45/470 (9.57)52/909 (5.72)64/1040 (6.15)45/620 (7.26)
  • Percentages in brackets. N/D = not documented

Socio-economic inequality and diabetes prevalence

A log-binomial regression model, table 2, showed independent relationships of social class, educational level and household income with diabetes prevalence in women in 2006.

View this table:
Table 2

Association of socioeconomic measures with diabetes prevalence by sex and year. Figures are prevalence ratios (95% CIs)

Model 1Model 2Model 1Model 2Model 1Model 2Model 1Model 2
Social class
    II1.14 (0.59–2.23)1.14 (0.59–2.22)0.90 (0.59–1.39)1.00 (0.66–1.51)1.04 (0.14–7.67)0.85 (0.12–6.17)3.32 (0.46–23.89)2.90 (0.40–21.03)
    III non-manual2.10 (1.05–4.19)2.07 (1.04–4.13)0.74 (0.43–1.29)0.73 (0.43–1.25)1.17 (0.16–8.47)0.94 (0.13–6.67)3.97 (0.55–28.39)3.52 (0.49–25.33)
    III manual1.17 (0.61–2.25)1.17 (0.61–2.26)1.10 (0.72–1.68)1.08 (0.72–1.61)1.80 (0.24–13.36)1.37 (0.19–10.02)5.50 (0.75–40.39)4.24 (0.57–31.36)
    IV + V1.14 (0.57–2.28)1.15 (0.57–2.32)1.25 (0.80–1.95)1.14 (0.75–1.75)1.64 (0.23–11.81)1.17 (0.16–8.30)6.51 (0.91–46.43)4.54 (0.63–32.70)
    P for trend0.9840.9790.0430.2970.0230.134< 0.0010.005
    GCSE/O-level1.07 (0.66–1.73)1.11 (0.69–1.80)0.95 (0.68–1.33)0.89 (0.65–1.24)1.44 (0.72–2.90)1.41 (0.70–2.82)1.74 (1.13–2.69)1.59 (1.03–2.45)
    Other0.92 (0.57–1.49)0.92 (0.56–1.49)1.24 (0.81–1.90)1.01 (0.67–1.53)1.49 (0.74–3.01)1.39 (0.69–2.78)2.31 (1.34–3.98)1.93 (1.13–3.29)
    None1.06 (0.74–1.52)1.06 (0.74–1.53)1.25 (0.96–1.63)0.94 (0.73–1.22)1.64 (0.93–2.90)1.39 (0.79–2.44)2.78 (1.91–4.07)1.96 (1.33–2.87)
    P for trend0.7830.8240.0750.7250.1000.368< 0.0010.001
Household income quintiles
    I (high)Ref.Ref.Ref.Ref.Ref.Ref.Ref.Ref.
    II1.34 (0.82–2.18)1.35 (0.83–2.20)1.09 (0.72–1.66)1.06 (0.71–1.59)1.43 (0.57–3.57)1.27 (0.51–3.16)1.65 (0.88–3.10)1.59 (0.85–2.98)
    III1.38 (0.87–2.21)1.35 (0.84–2.15)1.28 (0.86–1.91)1.19 (0.81–1.74)2.31 (1.01–5.27)1.84 (0.81–4.19)1.45 (0.77–2.71)1.20 (0.64–2.24)
    IV1.23 (0.75–2.01)1.02 (0.62–1.69)1.38 (0.92–2.07)1.02 (0.69–1.51)3.37 (1.52–7.49)2.44 (1.10–5.39)2.99 (1.70–5.26)2.24 (1.28–3.92)
    V (low)1.14 (0.68–1.90)0.87 (0.51–1.46)1.55 (1.02–2.35)1.00 (0.67–1.51)4.51 (2.03–9.99)2.75 (1.24–6.12)3.29 (1.84–5.86)2.12 (1.19–3.77)
    P for trend0.8390.3060.0101.000<0.001<0.001<0.0010.002
  • Model 1: adjusted for age; Model 2: adjusted for age, body mass index, waist hip ratio, physical activity, ethnicity, alcohol, smoking

  • a data on household income were not available for 1994

In 1994, there was no evidence of an association between social class and diabetes prevalence in women. The adjusted prevalence ratio of having diabetes by social class (IV + V vs. I) was 1.17 (P-value for trend = 0.13). In 2006, a negative association was evident with an adjusted prevalence ratio of 4.54 (P-value for trend = 0.005).

Similarly, there was no association of educational level with diabetes prevalence in 1994. The adjusted prevalence ratio for ‘no qualification’ vs. ‘A-levels’ was 1.39 (P-value for trend = 0.37), but there was evidence of a negative association in 2006 (prevalence ratio 1.96, P-value for trend = 0.001).

Data on household income were not available for 1994. In 1998, diabetes prevalence was negatively associated with level of household income (adjusted prevalence ratio for household income level V vs. I = 2.74, P-value for trend <0.001). This relationship was also significant in 2006 although statistically weaker than in 1998 (prevalence ratio 2.12, P-value for trend = 0.002).

In men, there were no associations between diabetes prevalence and the three measures of socio-economic position. A weak association of social class, qualification level and household income with diabetes in 2006 was accounted for by other risk factors for diabetes prevalence.

Slope index of inequalities

Table 3 shows the SII compared by survey year and gender.

View this table:
Table 3

Slope Index of Inequalities (SII) and Relative Index of Inequalities (RII) for diabetes prevalence, compared by year of survey and sex; age-adjusted diabetes rates using the European Standard Population

1994199820032006RII 1994 vs. 20061994199820032006RII 1994 vs. 2006
Social class (95% CI)−0.23 (–5.87 to 5.39)−0.94 (–4.17 to 2.30)−1.54 (–4.90 to 1.82)−2.80 (–7.90 to 2.30)−6% (NS) vs. –39% (NS)−1.65 (–3.61 to 0.32)−2.39 (–4.98 to 0.21)−2.79 (–7.16 to 1.59)−4.95 (–8.52 to –1.38)−72% (NS) vs. –101%
Education (95% CI)−0.41 (–3.41 to 2.58)−1.19 (–3.83 to 1.45)−1.26 (–7.08 to 4.57)−2.29 (–11.28 to 6.70)−11% (NS) vs. –32% (NS)−1.39 (–3.33 to 0.55)−4.47 (–5.95 to –3.00)−3.78 (–9.61 to 2.04)−6.48 (–9.03 to –3.93)−61% (NS) vs. –133%
RII 1998 vs. 2006RII 1998 vs. 2006
Household income (95% CI)−1.08 (–4.39 to 2.24)−4.12 (–6.41 to –1.83)−1.5 (–6.71 to 3.70)−26% (NS) vs. –21% (NS)−4.38 (–7.98 to –0.79)−4.34 (–8.61 to –0.64)−6.89 (–11.42 to –2.35)−154% vs. –141%

In women, there were no differences in diabetes prevalence evident between social classes or between high and low education groups in 1994. Over time, there was a gradual evolution of a gap in diabetes prevalence by social class and by qualification level with evidence of a gradient in 2006. Diabetes prevalence differed by 4.95% between women in social class I and social class V and by 6.48% between women with A-levels or higher education and those with no qualifications. There was evidence of inequality by level of household income from 1998 until 2006, corresponding to the independent associations between household income and diabetes prevalence shown in table 2.

In men, SII values did not significantly differ from zero for the socio-economic measures social class and level of education during the years investigated. For household income, there was evidence of inequality in 2003 with a difference in diabetes prevalence of 4.12% between highest and lowest household income groups. This finding was not persistent in the other years examined.

Diabetes prevalence by age group

Diabetes prevalence increased with age and in an overall binomial regression model, the prevalence ratio was 4.76 in 45- to 54-year-old men and 9.61 in 75- to 84-year-old men (compared to the 35–44 year age group). In women, the prevalence ratio was 4.41 in 45- to 54-year olds and 7.89 in 75- to 84-year olds (compared to the 35- to 44-year reference group). There was no evidence of age acting as an effect modifier when interaction factors (= age with respective socio-economic indicator) were included in the regression models and all P-values remained >0.1. Interaction factors between age group and year of survey also remained non-significant (in men P-value = 0.37; in women P-value = 0.57)

Diabetes diagnosis and glycated haemoglobin

Table 4 shows the prevalence of raised glycated haemoglobin levels (≥6.5%) in individuals not known to be diagnosed with diabetes. Measurements were only available for the years 2003 and 2006. In women, the highest percentages of potentially undiagnosed diabetes were in low social classes, in low household income groups and in lower education groups. In men this trend was less distinct.

View this table:
Table 4

Number (percentage) of individuals not known to be diabetic with a glycated haemoglobin measurement ≥6.5%

Men (2003 and 2006)Women (2003 and 2006)
Social class
    I10 (7.94)2 (2.33)
    II42 (33.33)16 (18.60)
    IIInm11 (8.73)21 (24.42)
    IIIm43 (34.13)8 (9.30)
    IV + V20 (15.87)35 (40.70)
    A-levels37 (29.37)9 (10.47)
    GCSE/O-levels16 (12.70)12 (13.95)
    Other15 (11.90)8 (9.30)
    None58 (46.03)57 (66.28)
Household income quintiles
    I23 (18.25)6 (6.98)
    II18 (14.29)8 (9.30)
    III19 (15.08)14 (16.28)
    IV29 (23.02)20 (23.26)
    V20 (15.87)23 (26.74)


The results of this study confirm that diabetes prevalence has increased over time. In women, the increase in diabetes prevalence was associated with increasing socio-economic inequality measured by social class and educational level. There are several possible explanations for this finding. Diabetes incidence could have increased disproportionately among lower socio-economic groups. Case ascertainment may have improved particularly in lower socio-economic classes. The finding may also be related to an increasingly ageing population with women living longer than men and the higher risk of poverty with older age, or to known inequalities in diabetes morbidity and mortality, improving treatment options for patients with diabetes and therefore potentially longer survival of people with diabetes in lower socio-economic classes.

In men, there appeared to be little change in socioeconomic inequality in diabetes prevalence over time.

Increased diabetes incidence and improved case ascertainment

An increase in type 2 diabetes incidence may be explained by the following risk factors for diabetes: unhealthy diet, obesity and a sedentary lifestyle.27

In high-income countries, socio-economic position has been associated with lifestyle and behavioural risk factors and it is well documented that lower socio-economic status is associated with lower levels of leisure-time physical activity, higher smoking rates and unhealthy diets that contain less fruit and vegetables and more saturated fats.27–29

The growing ‘diabetes pandemic’ is likely to have led to better awareness of the disease and increased case finding in individuals at high risk. The paper ‘Five Years On—Delivering the Diabetes National Service framework’30 reported a significant progress in the detection of patients with diabetes in the UK since 2003. As diabetes prevalence has been shown to be highest among obese and physically inactive people, a general improvement in identifying undiagnosed individuals with diabetes would explain the current findings. An analysis of individuals not diagnosed with diabetes, but with elevated glycated haemoglobin levels in the present study confirmed a higher prevalence of potentially undiagnosed diabetes in women of lower socio-economic position. For the results of the present study, this may mean that differences in inequalities could in fact be larger than suggested by the presented data.

There has been some debate about the appropriateness of using glycated haemoglobin levels for the diagnosis of diabetes. Christensen et al.31 suggest that diabetes prevalence estimated by oral glucose tolerance testing can differ considerably from diabetes prevalence when assessed with A1C assays and agreement may vary between ethnic groups. However, glycated haemoglobin levels are considered a precise measure of long-term glycaemic levels and they correlate well with the risk of diabetes complications.32

Socio-economic inequality in diabetes prevalence by gender over time

Socio-economic inequalities in diabetes prevalence were patterned by gender in the present study: No significant inequalities in diabetes prevalence were seen in men but increasing inequalities over time were seen in women. The reason for this is unclear but the findings are consistent with reports from other studies.33,34 Espelt et al.11 found inequalities in diabetes prevalence in both genders but the degree of inequality was more severe in women. It is possible that higher obesity rates in women of low socio-economic position have mediated the observed effect of increasing inequalities in diabetes prevalence. There are current increasing trends in obesity in men and women but an inverse relationship with socio-economic position is apparent in women only.17,18

However, after adjustment for obesity measures, the differences in diabetes prevalence by socio-economic indicators remained, suggesting that other variables that were not adjusted for, such as psychosocial factors, may be of additional relevance. Norberg et al.35 reported a higher risk of type 2 diabetes for women in passive or tense working situations and with low emotional support and these correlations were not seen in men. This may indicate that stress factors could be contributing and may have increased over time for women in socially disadvantaged groups.

There was a lack of an association between socio-economic position and diabetes prevalence in 1994 in both, men and women. A possible explanation is the lower number of individuals with reported diabetes in the earlier years. In fact, there are possible trends towards inequalities in diabetes prevalence in women also in 1994, but these did not reach statistical significance.


Undiagnosed or unreported diabetes was difficult to establish and to differentiate based on the survey data. The validity of self-reported diabetes may be as low as 75% in individuals with physician diagnosed diabetes36 and this number corresponds with findings from several studies.37,38 Some respondents with diagnosed diabetes may have been missed and an analysis of individuals that were not recorded as having diabetes is likely to reflect both people with undiagnosed diabetes and people with diagnosed diabetes who did not report their illness in the survey. The relationships presented could in reality be more distinct if diabetes prevalence is higher.

Social class, education and household income are frequently used in combination to assess socio-economic position. The opportunities to attain higher qualification levels have changed over time and elderly participants, in whom diabetes is commonest, may have comparatively fewer qualifications than younger participants, particularly women. The measures of household income and social class could be affected by long-term disease and be consequence as well as cause in any associations measured. The lack of data for household income in 1994 allowed only a limited assessment of associations with diabetes prevalence over time.

The nature of self-reported data that were used in the Health Survey for England may have led to varied response rates according to socio-economic status and in general, individuals tend to give answers that are most socially acceptable, leading to social desirability bias. The Health Survey questionnaire contains requests for cross checks in order to confine this bias.


The results of this study show that the rise in type 2 diabetes prevalence over the last decades was associated with an increase in inequalities in type 2 diabetes prevalence in women since 1994, as measured by occupational social class and by level of education. This may be a true increase owing to a rise in risk factors for diabetes in the population or an apparent increase owing to improved case ascertainment in socio-economically disadvantaged groups, earlier onset of type 2 diabetes and longer survival of individuals with diabetes.

Conflicts of interest: None declared.

Key points

What is already known on this subject

  • The prevalence of type 2 diabetes has increased worldwide.

  • Health inequalities in type 2 diabetes have been described with type 2 diabetes being negatively associated with socio-economic status.

What this study adds

  • Type 2 diabetes prevalence increased in the UK between 1994 and 2006 and the increase was associated with an increase in socio-economic inequality in diabetes.

  • This inequality increase was evident for women but not for men.


View Abstract