Article Text
Abstract
Background To analyse the validity of diabetes declarations in a health interview survey in order to evaluate the appropriateness of using health interview surveys to understand the relationship between diabetes and social groups.
Methods People with self-reported diabetes are those who report to have diabetes in the health survey. People with diabetes (gold standard) are those who were identified with fasting blood glucose level ≥126 mg/dl or those who were treated with oral antidiabetic drugs or insulin. Independent variables were educational level, social class and gender. The authors calculated sensitivity, specificity and κ coefficient.
Results The majority of indicators for evaluating the adequacy of using health interviews surveys to analyse inequalities in self-reported diabetes by social groups are good. The worst indicator was sensitivity, although it did not vary according to social groups.
Conclusion Health surveys are a good instrument to evaluate the social inequalities in the prevalence of diabetes.
- Diabetes DI
- inequalities SI
Statistics from Altmetric.com
Background
Diabetes is increasing all over the world and has become an important concern for all countries. Recently, several studies or reviews have attempted to understand the relationship between gender or socioeconomic position (SEP) and diabetes.1 2 In this sense, the majority of studies have relied on surveys based on self-reported diabetes. All these studies have found that people with deprived SEP (using educational level or occupational class) are at a high risk of suffering from diabetes.1 2 It is well known that a high percentage of people with diabetes are undiagnosed.3–7 It is very important to determine whether this proportion changes by SEP. The objective of this study was to analyse the validity of diabetes declarations in a health interview survey in order to evaluate the appropriateness of using health interview surveys to understand the relationship between diabetes, gender and SEP.
Methods
This study was based on a cross-sectional design. The study population consisted of a subsample of people included in the 2006 Catalan Health Interview Survey (ESCA). The survey was conducted in 2006 in Catalonia (Spain) by interviewing a representative sample of people at home. For our study, the sample population included men and women who were resident in Catalonia, aged 15 years or older. The study population (n=1774) was recruited by random sampling among people who accepted to participate in the health examination. People were contacted at their home, and the interviewers carried out the health examination including a blood extraction. It should be noted that quality control the health examination ensured representativeness, quality of the information and confidentiality. The study was approved by the ethical committee of Advisory Commission for the Treatment of Confidential Information of Catalonia Health Department.
Variables
People with self-reported diabetes are those who report to have diabetes in the health survey, while people with diabetes (gold standard) are those who were identified with fasting blood glucose level ≥126 mg/dl or those who were treated with oral antidiabetic drugs or insulin (according to the 1997 ADA Criteria).
Independent variables were educational level, social class, age and gender. Educational level was grouped into two categories: less than primary education and primary education or more. Social class was assigned according to the current or the most recent occupational status of the reference person of the household (respondent, partner or household member with the highest salary) and measured with a widely used Spanish adaptation of the British Registrar General Classification. Social class levels were reclassified into two categories: non-manuals (includes managerial and senior technical staff and freelance professionals; occupations and managers in commerce and skilled non-manual workers) and manuals (skilled and partly skilled manual workers and unskilled manual workers). Values for social class were missing in 12.1% of cases. Educational level and social class were categorised in two groups because of the low number of people with diabetes in some categories.
Data analysis
First, all the variables were described for people with self-reported diabetes and for people with diabetes separately. To test our hypotheses, some common statistical indicators of the validity of a diagnostic test were used, and a χ2 test was used to compare the two social categories: (1) sensitivity, which measures the proportion of people with diabetes and are correctly identified in the health survey and (2) specificity, which measures the proportion of people without diabetes and are correctly identified (ie, do not report diabetes in the health survey). Another indicator of the agreement between two diagnostic tests (κ coefficient) was calculated. Finally, the prevalence of known diabetes (people who self-report diabetes) and unknown diabetes (people who are identified by fasting blood glucose level or taking antidiabetic medication and have not self-reported diabetes) was also calculated. These indicators were also calculated by educational level, social class and sex. All the estimates were obtained by applying weights to restore the representativeness of the Catalonia general population due to the complex sample design of ESCA. The statistical analysis was carried out with Epidat 3.1.
Results
Table 1 presents a cross-tabulation of self-reported diabetes against having diabetes, classified according to SEP and gender (when tested, age was not a confounder). The prevalence of diabetes in Catalonia was 13.5% (95% CI 11.7% to 14.9%) with important socioeconomic inequalities. People with less than primary education have 2.69 high risk to have diabetes than those with primary education or more. The table also presents sensitivity, specificity and κ coefficient along with the corresponding 95% CIs of self-reported diabetes. Sensitivity does not vary significantly according to SEP. The sensitivity of self-reported diabetes is 24.0% (95% CI 15.4% to 32.7%) in non-manual workers and 30.2% (95% CI 20.5% to 39.9%) in manual workers. No differences are observed between men and women. Specificity varies according to social class dimensions. For example, the specificity of self-reported diabetes in manual workers is 98.2% (95% CI 97.1% to 99.3%), while in non-manual workers, it is 99.6% (95% CI 99.1% to 100.0%) showing that people in advantaged social groups without diabetes are better classified in health interview surveys than those in disadvantaged social groups. However, the specificity of social classes is high.
Discussion
This is the first study to present the indicators of validity, specificity and κ coefficient of self-reported diabetes in health interview surveys in terms of SEP and gender. Moreover, the study shows the prevalence of known and unknown diabetes for each social group. The main findings of this study are: (1) in general, most of these indicators do not vary by social variables in health interview surveys. However, in advantaged social groups, a lower punctual estimation of sensitivity than in deprived social groups was found and (2) educational level showed inequalities in the prevalence of diabetes (known and unknown diabetes).
In the USA, approximately one-third of all elderly people with diabetes are undiagnosed,3 whereas in Germany, half of all cases with diabetes are undiagnosed.5 It should be noted that while the prevalence of diabetes is rising, the prevalence of undiagnosed diabetes does not present a significant increase.4 Several studies have been conducted which did not find any differences in the patterns of social inequalities among people who are diagnosed, those who are undiagnosed or in the accuracy of self-reports of diabetes.7 8
These studies support our findings and, in addition, find that social patterns of diagnosed status in terms of educational level or social class do not differ substantially. However, in another study carried out in the USA, where insurance status varies across individuals, an increase of social inequalities in undiagnosed diabetes was found;9 people with higher SEP being better diagnosed than more disadvantaged people. This result would lead to underestimation of the risk of having diabetes among deprived SEP groups in health interview surveys without biological markers. However, the conclusions about diabetes inequalities drawn from surveys that use biological markers to assess diabetes seem not to differ compared to other international studies focused on social inequalities and self-reported diabetes.2 10 Our study adds new knowledge regarding this issue and presents some of the most commonly used indicators for evaluation of health interview surveys. We show that in Catalonia, a region of Spain with National Health System, the sensitivity of self-report diabetes in health interview surveys does not vary by social characteristics. In this sense, access and use of healthcare services could be an important predictor for being diagnosed with diabetes. We found that sensitivity is slightly lower in advantage social groups and in men. A possible explanation for these is that in Spain, women and people with deprived SEP visited the general practitioner more often than people with advantage SEP or men, regardless of their health needs.11
A limitation of this study is the small sample size that does not allow stratifying by sex, and consequently, we are not able to make conclusions for men and women into different social categories separately. Due to the limited statistical power of the study, we have to be cautious with the conclusions. However, our results are in accordance with other studies that analyse the probability of being undiagnosed of diabetes by SEP.5 7 Another limitation is that a 2-h post-glucose load test would be a better gold standard than the one used in this study. In our study, diabetes diagnosis was defined when the test was positive in one occasion, and it would be better to consider only when the test was positive in two occasions. However, the confirmation diagnosis with two tests is mainly used in clinical practice, and in epidemiological studies, a single occasion is standard.12 Moreover, there are no reasons to think that the results of these different gold standards could change by SEP or gender.
Furthermore, this is the first study to measure sensitivity, specificity and index κ of self-reported diabetes among social classes and gender, which gives a new and relevant point of view for future researchers on this topic.
We conclude that health surveys are a good instrument to evaluate the social inequalities in the prevalence of diabetes in terms of educational level, occupational class and gender inequalities.
What is already known on this subject
Previous studies reported no differences in the proportion of unknown diabetes by socioeconomic position, ethnicity or gender, using the fasting blood glucose level. Many studies on social inequalities in diabetes are based on health interview surveys, which could lead to bias if the proportion of unknown diabetes varies for social position.
What this study adds
When comparing fasting blood glucose level with self-reported diabetes in a health interview survey in Catalonia (Spain), we found that there were no differences in the proportion of known and unknown diabetes for educational level, occupational class or gender. The study shows that health interview surveys without blood fasting glucose could be a valid tool to analyse the social inequalities in diabetes.
Acknowledgments
This study was based on data of the Catalonia Health Survey, and the authors are very grateful to the Departament de Salut de la Generalitat de Catalunya for providing them. The authors especially thank Dra. Conxa Castell and Dave Macfarlane of the Institut Municipal d'Atenció Sanitària (IMIM) for assistance with English revision.
References
Footnotes
This article forms part of the doctoral dissertation of Albert Espelt at the Universitat Pompeu Fabra of Barcelona.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Departament de Salut de Catalunya.
Provenance and peer review Not commissioned; externally peer reviewed.