Article Text

Download PDFPDF

Prevalence of diagnosed depression in South Asian and white European people with type 1 and type 2 diabetes mellitus in a UK secondary care population
  1. S Ali1,
  2. M J Davies2,
  3. N A Taub3,
  4. M A Stone3,
  5. K Khunti3
  1. 1
    University of Warwick, Coventry, UK
  2. 2
    University of Leicester, University Hospitals of Leicester, Leicester, UK
  3. 3
    University of Leicester, Leicester, UK
  1. S Ali, Health Sciences Research Institute, Warwick Medical School, Gibbet Hill Campus, University of Warwick, Coventry CV4 7AL, UK; S.ali{at}


Aim: To examine the prevalence and correlates of diagnosed depression among South Asians and white Europeans with type 1 and type 2 diabetes mellitus, attending a specialist diabetes clinic in the UK.

Study design and methods: A cross-sectional study was conducted using the hospital clinic’s computerised database. Medical and demographic data were extracted for 6230 people with diabetes attending the clinic between 2003 and 2005. Multiple logistic regression was used to model ethnic differences in the probability of diagnosed depression after controlling for demographic and diabetes related factors. Analyses were conducted separately for type 1 and type 2 diabetes.

Results: The unadjusted prevalence of depression in people with type 1 and type 2 diabetes was 8.0% and 9.3%, respectively. Risk factors for depression in type 1 diabetes included female gender, diabetes related complications, and comorbidities. In people with type 2 diabetes the risk factors for depression included younger age, diabetes related complications, comorbidities, insulin use and deprivation. In addition, white Europeans were significantly more likely to be diagnosed with depression compared to South Asians (odds ratio (OR) 1.59, 95% confidence interval (CI) 1.21 to 2.08; p<0.001). Further interaction analyses revealed no evidence that the association between ethnicity and depression differed according to any of the other factors examined in this study.

Conclusions: The findings add to the limited body of knowledge regarding ethnic differences in depression and diabetes. Among those with type 2 diabetes, white Europeans had nearly 60% higher adjusted odds of diagnosed depression compared to South Asians. Disparities may be due to differences in presentation or identification of depression between these two ethnic groups.

  • depression
  • diabetes
  • ethnic groups
  • psychology
  • epidemiology

Statistics from

A growing body of literature continues to demonstrate increased rates of depression in people with both type 1 (T1DM) and type 2 diabetes mellitus (T2DM) relative to those without.1 Although the exact aetiology of this relationship is still a subject of research, evidence to suggest negative associations between depression and glycaemic control and the development of diabetes related complications is presented in recent meta-analyses.2 3 Examining risk factors for depression in diabetes is key in terms of improving our understanding of the relationship as well as enabling healthcare professionals to identify high risk groups. However, despite an emerging interest in racial and ethnic variations in the rates of this comorbidity in recent years,4 the literature has so far failed to examine the association in migrant South Asian populations (people of Indian, Pakistani, Bangladeshi or Sri Lankan descent). A steady increase in the incidence of T1DM diabetes has been observed UK South Asian children, with the rising rate also demonstrated to be higher in comparison to other ethnic groups.5 Epidemiological studies conducted in various parts of the world have also observed a dramatic increase in the prevalence of T2DM in South Asians, with reports of up to a fourfold increased risk in comparison to white Europeans.6 Furthermore, poor glycaemic control, microalbuminuria, retinopathy and cardiovascular disease mortality have been shown to be higher in this group compared to white Europeans.7 8 The present study aims to address this gap in the literature by examining the prevalence and ethnic differences in the risk of diagnosed depression between South Asian and white European people with T1DM and T2DM. In addition we aim to determine whether the factors associated with depression vary between these two ethnic groups.


A cross-sectional study was conducted using the computerised database (clinical workstation) at a hospital diabetes and endocrinology clinic based in Leicestershire, UK. Leicestershire has one of the largest diabetes services in the UK, with a population of approximately 1 million, of whom 36 600 are registered as having diabetes. The proportion of the city’s population from ethic minority backgrounds is almost five times the average for England and Wales (28% vs 6%), with people of South Asian origin comprising the largest minority group (24%).9 The hospital clinic serves a population with mixed ethnic and demographic backgrounds and thus provides an ideal opportunity to investigate ethnic differences in people with diabetes.

The clinical workstation

The clinical workstation is a purposively created electronic patient record system, used for recording clinical data and routine correspondence, as well as for audit and research purposes. The database includes routinely collected demographic and clinical data including medical history, diagnosis and treatment for all patients attending the clinic, either through referral or for their annual diabetes review. Also integrated within the system are links to specialist datasets for the latest as well as previous laboratory results.

Data collection

Electronic patient data were retrieved retrospectively using the search function of the clinical workstation. Data most recent in time to the patient’s clinic attendance were extracted for all patients visiting the clinic between January 2003 and March 2005.

Demographic data included age, gender, postcode and ethnicity. Postcodes were obtained in order to derive indices of multiple deprivation (IMD) scores as a proxy for socioeconomic status, with greater scores indicating higher levels of deprivation.10 Ethnicity was classified as South Asian or white European based on that given in the patient’s record or by the analysis of their name, making use of the South Asian name recognition software “Nam-Pehchan”11 supplemented by a visual inspection of surnames and forenames. This method for ascribing South Asian ethnicity has been shown to have high reliability in UK populations.11 Other racial/ethnic groups were not included in the final analyses of ethnic differences as the group was considered to be too heterogeneous and the numbers (n = 318) too small for meaningful analysis.

Medical data included type and duration of diabetes, current use of insulin and oral anti-diabetic medication, smoking status, body mass index (kg/m2), glycosylated haemoglobin (HbA1c) (%), the presence of one or more comorbid conditions (coronary heart disease, cerebrovascular disease, asthma, chronic obstructive pulmonary disease, heart failure, peripheral vascular disease, inflammatory bowel disease, irritable bowel syndrome, epilepsy, hypothyroidism, transient ischaemic attack, fibromyalgia and malignant neoplasm), one or more diabetes related complications (peripheral neuropathy, diabetic retinopathy, nephropathy, neuropathy and microalbuminuria), psychotic disorder, and psychiatric disorder.

Incomplete data is a common problem in clinical databases and missing data were observed for BMI (30%) and HbA1c (3.2%). Type of diabetes was not specified in 448 cases (7.2%). For these cases, patients were classified as having T1DM if they had been diagnosed under the age of 35 years and were receiving insulin therapy.

Depression identification

Patients with a case documentation of depression or those in receipt of antidepressant medication at the therapeutic dosage recommended for depression12 were classified as depressed. This ensured the inclusion of patients who may have been diagnosed with depression but without a case documentation, a problem which is known to occur within medical records. In contrast, a valuable feature of the clinical workstation is the availability of complete data regarding pharmaceutical prescriptions, including receipt of antidepressants. However, it is recognised that the therapeutic dosage for many antidepressants overlaps with that prescribed for a number of psychiatric complaints as well as neuropathic pain. Such disorders include obsessive compulsive disorder, post-traumatic stress disorder, bulimia, panic disorder and social anxiety; as such, patients with a diagnosis of any one of these problems were excluded.

Ethics committee permission was not required as the study involved secondary analysis of routinely collected audit data.

Statistical analysis

Demographic and medical characteristics were compared between South Asians and white Europeans with T1DM and T2DM. All subsequent analyses were performed separately by diabetes type.

Continuous variables were divided into categories using conventional cut-points (age, duration of diabetes, HbA1c). As no standard cut-points exist for IMD scores, the variable was categorised into four equal groups, with 0 representing the lowest level of deprivation and 4 the highest. Patients with and without depression were compared using χ2 tests.

Multiple logistic regression modelling was conducted in order to examine the variation in the risk of depression according to ethnicity, while controlling for age, gender, comorbidities, complications, insulin and oral anti-diabetic medication use (T2DM), BMI, HbA1c, duration of diabetes, and deprivation. Each of these variables has been shown to be associated with depression in previous studies.13 Due to the high proportion of missing data for BMI, the multiple imputation technique proposed by Rubin 198714 was applied with missing at random assumptions, in order to generate a complete dataset for analyses. Five sets of imputed data for BMI were produced using STATAv9 (StataCorp 2005, Stata Statistical Software, College Station, Texas, USA). Individual logistic regression models were carried out with each imputed dataset. The method described by Rubin14 was used in order to pool the results from the analyses and produce odds ratios (ORs) and confidence intervals (CIs) that incorporate missing data uncertainty.

To examine whether factors associated with depression differed between South Asian and white European patients, logistic regression was performed separately for each ethnic group. Controlling for the effects of age, gender, comorbidities and complications, each variable was entered into the model individually in six separate analyses. The regression models were repeated with each independent variable entered separately with the interaction between the variable in question and ethnicity. Each model was adjusted for age, gender, comorbidity and complications. Hosmer–Lemeshow tests were conducted with each model to assess the degree of goodness of fit.


Of the 6230 study sample, 22.5% (n = 1405) had T1DM and 76.7% (n = 4781) had T2DM. Forty-four (0.7%) of the sample were unclassified due to insufficient information.

Type 1 diabetes mellitus

Table 1 shows medical and demographic characteristics of the study population. The majority of patients were aged ⩽59 years (80%) and had had diabetes for 15 years or longer (57.4%). Fifty-seven per cent were female and similar proportions of the sample were represented in each group for IMD scores. Twenty-five per cent of the sample had one or more comorbid condition and 54.9% had one or more diabetes related complication. HbA1c values ⩾7% were observed in 1123 patients (84%) and 298 (24.9%) had a BMI >30 kg/m2.

Table 1 Demographic and medical characteristics of patients by diabetes type and ethnicity

A greater proportion of South Asians had an IMD score within the highest quartile for deprivation in comparison to white Europeans (28.2% vs 16.9%). South Asians were more likely to have HbA1c values ⩾7% (88.5% vs 83.5%). White Europeans were significantly more likely than South Asians to be smokers (30.5% vs 16.9%) and to have had diabetes for at least 15 years (59.7% vs 45.4%).

The prevalence of depression was higher in females than in males and this pattern was mirrored when examined separately by ethnicity for white Europeans; however, a significant gender difference was not found in South Asians. Rates of depression were also higher in patients with one or more comorbidity or diabetes related complication and in smokers (table 2).

Table 2 Association between depression and demographic and medical variables

Table 3 presents the results from the multiple logistic regression. Female gender, the presence of one or more diabetes related complication, or comorbidity were significantly and independently associated with depression.

Table 3 Odds of depression calculated using multiple logistic regression, separately for type 1 and type 2 diabetes mellitus

Results from the logistic regression models conducted separately by ethnicity showed that in white Europeans, female gender (OR 2.1, 95% CI 1.3 to 3.2, p = 0.001), comorbidities (OR 1.8, 95% CI 1.15 to 2.96, p = 0.011), complications (OR 1.7, 95% CI 1.1 to 2.7, p = 0.027) and smoking (OR 1.7, 95% CI 1.1 to 2.6, p = 0.025) were significantly associated with depression; however, none of the variables examined showed a significant association with depression in South Asians. No significant interactions were observed between ethnicity and the independent variables examined.

Type 2 diabetes mellitus

The majority of the sample were aged ⩾60 (65.5%) and similar proportions of men and women (52.8% and 47.2%) and each IMD score group were represented. Fifty-six per cent were using insulin therapy and 68.6% were prescribed oral hyperglycaemic medication. One or more comorbid conditions were present in 46.4% of the sample, and one or more diabetes related complications were present in 47.2%. Sixty eight per cent had an HbA1c ⩾7% and 55.1% had a BMI >30 kg/m2 (table 1).

In comparison to white Europeans, a greater proportion of South Asians scored within the most deprived IMD score quartile group (30.8% vs 24.5%), had been diagnosed with diabetes for 15 years or more (28.4% vs 21.6%), and had HbA1c ⩾7% (73.5% vs 65.4%). White Europeans with T2DM, in comparison to South Asians, had a higher mean (SD) BMI (32.2 (6.8) vs 29.9 (5.9)) (table 1).

In bivariate comparisons (table 2), the prevalence of depression was 10.3% in white Europeans compared to 7.7% in South Asians (p = 0.008). The rates were also higher in patients aged ⩽59 years, female patients with a longer duration of diabetes, patients on insulin, those with one or more comorbidity, those with one or more complication, and patients who were smokers. A significant association was also demonstrated with BMI.

In multiple logistic regression analyses (table 3), white European ethnicity was a significant independent risk factor for depression. The odds of depression were also increased in those aged ⩽59 years, with complications or comorbidities, in those using insulin, and in patients with a higher score for deprivation.

In models conducted separately by ethnicity, being aged <59 years (white European: OR 1.7, 95% CI 1.3 to 2.2, p<0.001; South Asian: OR 2.31, 95% CI 1.5 to 3.6, p<0.001), female gender (white European: OR 1.5, 95% CI 1.2 to 1.9, p<0.001; South Asian: OR 1.7, 95% CI 1.14 to 2.7, p = 0.009), comorbidity (white European: OR 1.5, 95% CI 1.2 to 2.0, p<0.001; South Asian: OR 1.9, 95% CI 1.23 to 2.9, p = 0.004), and insulin use (white European: OR 1.4, 95% CI 1.1 to 1.8, p = 0.017; South Asian: OR 1.9, 95% CI 1.2 to 3, p = 0.009) were associated with depression. Complications (white European: OR 1.6, 95% CI 1.3 to 2.1, p<0.001) and BMI (white European change in 5 units: OR 1.02, 95% CI 1.00 to 1.03, p = 0.029) were associated with depression in white Europeans but not South Asians. No significant association was demonstrated between ethnicity and any of the independent variables examined.

Hosmer–Lemeshow tests demonstrated no indication of poor model fit to the data (p>0.05 in all instances).


To our knowledge this study represents the first report of ethnic differences in the prevalence and risk factors for depression in a large population with both T1DM and T2DM. In addition, the study is unique in that it presents data for migrant South Asians, a previously under-researched group.

Our unadjusted rates for depression are similar to reports from a large population based survey of participants from over 60 countries, which identified prevalence rates of depression to be between 7.3–11.3% in people with diabetes.15 Depression prevalence estimates vary widely, however, depending on a range of factors including the medical and demographic characteristics of the population studied.1 13 Comparisons between our data and previous studies should therefore be exercised with caution. For example, a meta-analysis of 43 studies of the prevalence of depression in people with T1DM and T2DM concluded that the rates were generally higher in studies conducted in clinical settings (32%) as opposed to population settings (20%).1 Furthermore, differences have been associated with the method used for depression identification, with a higher prevalence observed in studies using self report questionnaires (31%) than with diagnostic interviews.1 It is important, however, to make a distinction between these findings and the present study, in that our observations of depression prevalence are based on cases both recognised by a physician and treated using antidepressant medication. A similar process for depression identification has been applied in only one previous study. Nichols and Brown reported unadjusted rates of diagnosed depression to be 17.9% in a large US population sample with T2DM.16 Although our rates are considerably lower, population characteristics and geographical differences make comparisons with these data problematic.

Nevertheless, factors associated with depression are consistent with past reports,1 4 13 16 17 suggesting that cases with diagnosed depression as identified in our study may not be markedly distinct from cases identified by Nichols and Brown, or previous studies utilising diagnostic interviews or self report tools. In T1DM and T2DM, the odds of depression were increased in females, smokers and those with complications and comorbidities. Increased odds for depression were also associated with the duration of diabetes, insulin use and cholesterol values in T2DM, but not in T1DM.

Risk factors associated with depression varied in T1DM and T2DM and also by ethnicity, demonstrating the importance of observing these subgroups separately when examining the relationship between depression and diabetes. The findings also suggest that research is needed to identify additional risk factors that are associated with depression in South Asians, particularly with T1DM.

Rates of depression in people with diabetes have previously been found to be either similar or higher in ethnic minorities.4 17 In contrast, our results showed that white Europeans exhibited significantly higher rates of diagnosed depression relative to South Asians, although this difference was not observed in T1DM. Interaction analyses revealed that these findings are not modified by any of the variables we examined, including socioeconomic status.

It is possible that our findings reflect true differences in the underlying prevalence of depression between these two ethnic groups. However, our data relate to diagnosed depression, which may differ from actual rates of this condition. Although a small number of studies have also identified lower rates of depression in the South Asian population in the UK in comparison to white Europeans,18 these findings have been criticised on methodological grounds. For example, it is argued that traditional westernised methods for depression assessment may be culturally inappropriate and therefore fail to recognise depression in South Asian populations in whom the manifestation and expression of depression may differ from their white European counterparts.18

Similarly, lower rates of observed depression in South Asians with T2DM may be due to difficulties in case recognition of depression and/or antidepressant treatment due to problems associated with language, stigma, beliefs about health care as well as cultural differences in the presentation of depression.19 20 These issues may be particularly accentuated in the South Asian elderly (those who migrated to the UK in late childhood or adulthood) rather than the generally younger “second generation” South Asians who may have a greater cultural affinity to westerns idioms of depression19 as well higher rates of English literacy.21 This in turn may explain why ethnic disparities in the rates of depression were observed in people with T2DM diabetes (the majority of who were aged 60+ years) and not in people with T1DM.

Limitations of the present study include the fact that the data are from a single geographically located hospital and thus results may not be generalisable to other settings. Furthermore, our data are limited to the coding strategies implemented within the clinical workstation and therefore without explicit categorisation of various diagnostic subgroups of depression, we are unable to provide estimates specifically for the rate of major depression. Our categorisation of ethnicity also represents broad groups and therefore any differences which may exist within subgroups are not accounted for in this analysis.

Main messages

  • Few data are available regarding the rates and risk factors for depression in patients with diabetes in the UK, particularly in UK South Asians who are at increased risk of developing diabetes and diabetes related complications.

  • Identification and treatment of depression in patients with diabetes may have a favourable effect on glycaemic control and perhaps even prevent or delay diabetes related complications.

  • In people with type 2 diabetes, the rate of diagnosed depression is significantly lower in South Asians compared to white Europeans.

  • Factors associated with depression vary between South Asian and white European people with type 1 and type 2 diabetes.

Current research questions

  • Is the prevalence of depression lower in South Asians or are there ethnic differences in the diagnosis and/or treatment of depression in this population?

  • What are the risk factors for depression in South Asians with type 1 and type 2 diabetes and how may better clinical practice be developed?

The accuracy of our prevalence rates is also dependent on the completeness of routinely recorded information and there is likely to be under recording of depressive symptoms. In order to maximise sensitivity we incorporated the receipt of antidepressant medication into our definition of depression. However, our data cannot account for cases in which physicians had failed to record a diagnosis and patients in question were in receipt of non-drug management only. Our figures are therefore likely to be an underestimation of the prevalence of diagnosed depression. Furthermore our rates are likely to underestimate the prevalence of depression in people with diabetes, as it is well established that rates of recognition of depression are low in both primary and secondary care services.22 Finally, our data are cross-sectional which in turn limits analysis concerning direction or causation.

Despite these limitations, the present study offers the first insight into the prevalence of depression in a large multi-ethnic population with both T1DM and T2DM diabetes in the UK. Depression in diabetes represents a significant impediment to the interests of public health, with research documenting its contribution to poor glycaemic control, a reduction in functional status and quality of life, as well as the development of both micro- and macrovascular disease.2 13 23 These issues may be of particular concern in South Asian communities within the UK who are at increased risk for both diabetes and adverse outcomes associated with comorbidity.68 The findings from this study identify important ethnic disparities in the rates of diagnosed depression in people with diabetes. Once recognised, depression in people with diabetes can be effectively treated, with improvements demonstrated in glycaemic control as well as reduced risk of mortality.24 25 Further research is needed in order to identify the underlying reasons for these ethnic disparities. In the meantime, health care providers should continue to be alert to the possibility of depression in all patients with diabetes.


The authors would like to thank Ismail Gangat, Clinical Information Analyst, Leicester Royal Infirmary, for his much appreciated support with the data collection for this project.


View Abstract


  • Competing interests: None.

  • Patient consent: Patient consent not required

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.