Skip to main content
Download PDF

Combining education and income into a socioeconomic position score for use in studies of health inequalities

BMC Public Health volume 22, Article number: 969 (2022) Cite this article

Abstract

Background

In studies of social inequalities in health, there is no consensus on the best measure of socioeconomic position (SEP). Moreover, subjective indicators are increasingly used to measure SEP. The aim of this paper was to develop a composite score for SEP based on weighted combinations of education and income in estimating subjective SEP, and examine how this score performs in predicting inequalities in health-related quality of life (HRQoL).

Methods

We used data from a comprehensive health survey from Northern Norway, conducted in 2015/16 (N = 21,083). A composite SEP score was developed using adjacent-category logistic regression of subjective SEP as a function of four education and four household income levels. Weights were derived based on these indicators’ coefficients in explaining variations in respondents’ subjective SEP. The composite SEP score was further applied to predict inequalities in HRQoL, measured by the EQ-5D and a visual analogue scale.

Results

Education seemed to influence SEP the most, while income added weight primarily for the highest income category. The weights demonstrated clear non-linearities, with large jumps from the middle to the higher SEP score levels. Analyses of the composite SEP score indicated a clear social gradient in both HRQoL measures.

Conclusions

We provide new insights into the relative contribution of education and income as sources of SEP, both separately and in combination. Combining education and income into a composite SEP score produces more comprehensive estimates of the social gradient in health. A similar approach can be applied in any cohort study that includes education and income data.

Peer Review reports

Background

An extensive empirical literature has documented a positive association between individuals’ socioeconomic position (SEP) and their health, commonly referred to as the social gradient in health [1, 2]. The social gradient reflects that individuals’ structural location in society is an important determinant of the likelihood of experiencing health-damaging exposures, or of holding certain health-enhancing resources [3]. This is largely built on the theoretical contribution of Max Weber, who argued that society is stratified into hierarchies along various dimensions, creating groups based on different sets of skills, knowledge, and assets. These factors, which Weber defined as individuals’ “life chances”, produce social stratification and will, as such, determine individuals’ position in the marketplace [4]. Measures of SEP aim to reflect these life chances [5].

However, there is no single measure that best identifies SEP [4]. Therefore, SEP is most commonly captured by three proxy measures: education, occupation and/or income [6]. Through various mechanisms, these measures produce status that is considered health-enhancing (see e.g., Marmot [7]). While closely related, the three measures are not interchangeable [8, 9].

A growing literature suggests that subjective SEP measures are also powerful determinants of health [10]. Rather than focusing solely on objective indicators of SEP, inequalities in subjective SEP could be as important, or even more strongly linked to health than objective SEP measures [11, 12]. This builds on the hypothesis that subjective SEP captures socioeconomic dimensions not measured by objective SEP indicators [13, 14]. For example, in The English Longitudinal Study of Ageing, it was found that subjective SEP mediated the association between objective SEP measures and mortality, as well as independently predicting mortality [10]. Additionally, none of the objective SEP indicators directly measure socially derived attributions of prestige or status [15]. This suggests that objective SEP measures should be complemented by a measure of subjective SEP.

The association between SEP and health has been observed with each of the three objective SEP indicators, which could indicate that SEP represents a broader, underlying construct related to social stratification [16]. Therefore, if these SEP variables capture different aspects of the same concept [6], a composite measure could better represent SEP when estimating social inequalities in health [17]. Additionally, a composite SEP measure may capture multiple aspects of relevance when estimating how individuals’ SEP influences health inequalities, thus simplifying interpretation [18] and communication of results [19].

In the literature on social inequalities in health, composite indicators of SEP are applied in different ways. The focus here will be on individual-level composite indicators. Early examples include the Hollingshead index of social status [20], using a priori defined weights for education and occupation; the Duncan’s socioeconomic index for occupational prestige; and the Nam-Powers occupational status score [19, 21]. These indicators are not as relevant today due to changes in eduation and the labour market [22]. However, the Nam-Powers score has in later years been updated and refined into the Nam-Powers-Boyd occupational score, using data from the 2010–12 American Community Surveys, in which median education and median earnings of different occupations are used as the basis for the score [23]. In the UK, occupation is widely used for socioeconomic classifications [6]. In application today is the National Statistics Socio-economic classification, which incorporates employment relations and conditions of occupations, into non-hierarchical occupational classes [24]. The latter two examples are limited to the US and UK contexts, and would need to be adjusted to fit other contexts. Other recent examples of composite SEP indicators most frequently use education, occupation and income for composite SEP indices (see e.g., [25, 26]), as well as education and income only (e.g., [27, 28]).

A common critique against composite SEP measures is that they conceal the relative influence of their components [29]. However, it can also be argued that a composite indicator of SEP can capture the synergies between its different components [17]. In this paper, we propose a composite SEP score that compiles several SEP indicators into one, that still allows for disaggregation of the score’s components.

In the literature on social inequalities, the most common health indicators are mortality or disease-specific health outcomes (e.g., [30,31,32]), or self-rated health [33]. In this paper, we use two measures of health-related quality of life (HRQoL): the multidimensional EQ-5D-5L descriptive system, and a visual analogue scale (VAS).

The current paper is based on data from a general adult population and aims to: i) develop a composite SEP score from empirically derived weights that reflect individuals’ subjective SEP; and ii) test how the composite SEP score predicts inequalities in HRQoL. We regress subjective SEP on education and income levels. The resulting weights are used to predict a SEP value for each individual based on combinations of their education and income levels. We further demonstrate how the composite SEP score predicts inequalities in HRQoL. This study contributes to the literature by proposing a simple composite SEP score based on the two most widely collected objective indicators of SEP using derived weights according to their influence on subjective SEP.

Conceptual framework

The concept of SEP is complex. It is therefore necessary to describe its components and the hypothesised relationships between them.

Education proxies an individual’s cognitive resources and the ability to process health information [4]. In addition, education has been found to be strongly associated with childhood socieconomic conditions (see e.g., [34]), and can, as such, be understood as a representation of early-life circumstances. Education is often measured as the highest level of educational attainment, or as years of education.

Occupation mirrors educational achievement, yields income, and reflects individuals’ social standing [35]. Occupation indicators can capture the prestige associated with specific professions; environmental exposures on the job (e.g., pollution); or psychosocial aspects, such as job strain and satisfaction [6].

Income is hypothesised to impact health through the ways in which individuals’ resources provide a healthy physical environment, healthier lifestyle and/or ease of access to health services. Additionally, income itself can entail a higher SEP [6].

These three indicators can be conceptualised as components of the latent construct of objective SEP. This is shown in our conceptual framework (Fig. 1), which demonstrates the hypothesised links between the key concepts included in this paper. Education provides skills and knowledge that qualify people for specific occupations. The higher education level that an occupation requires, the more cognitive resources and skills does the individual possess, all of which are associated with objective SEP. However, occupations with similar levels of educational attainment (e.g., a physician vs a priest) differ immensely in terms of income levels: having a high income, then, reflects that the individual has an occupation that society values more highly. Thus, education, occupation, and income represent the concept of objective SEP, displayed in Fig. 1, as encompassing these three indicators. In this framework, objective SEP predicts subjective SEP, which in turn, determines HRQoL. Additionally, age and sex are added as covariates, as they are assumed to influence both subjective SEP and HRQoL. They will also likely influence objective SEP, but this model focuses on how they relate to subjective SEP. Lastly, although not included in this paper, it is important to acknowledge the role played by the intergenerational transmission of both socioeconomic factors and health. It is widely established that parents not only transfer their genes to their offspring, but also their SEP and health (behaviours) [36,37,38].

Fig. 1
figure 1

Conceptual framework of the relationship between the components of SEP and HRQoL. Note: Arrows reflect hypothesised associations between key concepts. SEP: socioeconomic position

Full size image

In the current study, occupational category is not directly included in the composite SEP score. However, it is indirectly captured, in that occupation (to a large extent) is determined by education, and (to an even larger extent) a determinant of income. As opposed to education, measured in years; and income, measured in money, occupational categories can be more difficult to hierarchically order. This is because the categories include individuals with large differences in skills, prestige, power, and/or incomes, and are arguably not originally developed as a SEP measure [8]. Moreover, occupational measures vary widely in what they proxy and are likely to differ substantially between countries and different contexts [4]. In this sense, education and income are more consistently available from surveys and registers than occupation.

Methods

Data

The Tromsø Study is a prospective cohort study from a general adult population residing in the municipality of Tromsø. With approximately 77,000 inhabitants, Tromsø is the largest city in Northern Norway. The current paper is based on data from the seventh wave conducted in 2015/16. Of the 32,591 people that were invited (aged 40 years and older), N = 21,083 (65%) completed the survey. The study design is described in detail elsewhere [39].

The study was approved by the Regional Committee for Medical Research Ethics Northern Norway (REK North; ID 2019/607). The Tromsø Study complies with the Declaration of Helsinki and all participants gave written informed consent before admission. Data access was granted by the Data and Publication Committee of the Tromsø Study. All methods were carried out in accordance with relevant guidelines and regulations.

Variables

Education was recorded as the highest completed education level, categorised into four: primary education up to ten years; upper secondary and vocational school; undergraduate (less than four years of higher education); and postgraduate degree (four years or more of higher education).

Income was recorded as the combined gross income of adults in the household, in eight income brackets. These were collapsed to approximate quartiles. Income groups were (per NOK 1,000): Low: ≤ NOK 450 (20.9%); Lower middle: NOK 451–750 (29.3%); Upper middle: NOK 751–999 (24.2%); and High: NOK ≥ 1 million (25.6%).

Inspired by the seminal work of Marmot on the crucial role of social status [7], we used subjective SEP to develop the composite SEP score. Subjective SEP was obtained from the statement ‘I consider my occupation to have the following social status (if you are currently out of work, think about your latest occupation)’, which was rated using a five-level scale (very high; fairly high; middle; fairly low; very low). With few respondents in the lowest category (< 1%), we collapsed the bottom one into the category for ‘low’ status, leaving subjective SEP as a four-level ordinal variable. The subjective SEP measure is framed in terms of the perceived SEP of respondents’ occupation, as an individual’s occupation is thought to largely shape the perception of own social standing. It is a variant of the more commonly applied MacArthur scale of subjective social status [11].

HRQoL was the main outcome variable and was measured in two ways: directly on a VAS, and indirectly with the EuroQol EQ-5D. The EQ-5D is the most widely applied generic preference-based descriptive system [40, 41]. It describes health along five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression [42]. We applied the most recent version with five severity levels along each dimension (EQ-5D-5L) [43]: ‘no problems’, ‘slight problems’, ‘moderate problems’, ‘severe problems’, and ‘unable’. In the absence of a Scandinavian value set for the EQ-5D, we used an amalgam tariff, the Western preference pattern (WePP), representing a hybrid of four Western countries’ published value sets [44]. The VAS asks respondents to rate their health today on a scale from [0–100]. The VAS was converted into a [0–1] interval for reasonable comparison with the EQ-5D value.

Age and sex were included as covariates.

Statistical analyses

Descriptive statistics

Means, proportions, and standard deviations (SD) of the included variables were reported for the full sample and stratified by sex. We excluded respondents above the age of 80 (N = 761) due to a disproportionately low response rate and to diminish the impact of cohort effects on the education variable, leaving a sample of N = 20,322. For the analyses, respondents with missing observations for education and income were excluded, corresponding to 4.5% of the sample.

Regression-based approach to develop a composite SEP score

To develop the composite SEP score, we applied subjective SEP as the dependent variable, proxying SEP. We used adjacent-category logistic regression, which is an alternative to classical ordered logistic regression. This method compares each category (level) of the dependent variable with the next larger response category [45]. We modelled the four-level subjective SEP (\(sSEP\)) variable as a function of education (\(Educ\)), and income (\(Inc\)) (Eq. 1). The education and income variables were dummy-coded, with the lowest level serving as the reference level for each variable. Sex and age (in years) were included as control variables (\(X\)):

$$sSEP=f\left(Educ,Inc,X\right)$$
(1)

The resulting regression coefficients from Eq. 1 were used as education and income weights in the composite SEP score. Each of the education and income levels was multiplied with their corresponding regression coefficient, resulting in a composite SEP score that predicts individuals’ SEP, demonstrated in Eq. 2:

$${SEP}_{i}={\sum }_{j=1}^{k}{\beta }_{j}*{Educ}_{ij}+{\sum }_{j=1}^{k}{\gamma }_{j}*{Inc}_{ij}$$
(2)

This approach was inspired by Mehta et al. [46]: instead of using the risk ratios to construct a summary score, the composite SEP score was generated based on the regression coefficients modelled in Eq. 1. The SEP summary score from Eq. 2 was rescaled into a [1–10] interval to form the composite SEP score: first, the coefficients for all combinations of education and income level \(j\) were added together. Second, each value of the composite SEP score was rounded to the nearest integer, resulting in a predicted SEP value [1–10] for each individual \(i\) based on their combinations of income and education levels. As such, we identified how different levels of education and income influence subjective SEP.

Predicting variation in HRQoL

To evaluate how the composite SEP score predicted variation in HRQoL (EQ-5D and VAS), we ran ordinary least squares (OLS) regression of HRQoL on the composite SEP score, adjusted for age and sex. Further, we calculated the age-adjusted predicted mean HRQoL values (EQ-5D and VAS) for all values of the SEP score.

As an alternative analysis of variation in HRQoL, we applied the concentration index (CI). The CI measures the degree of socioeconomic inequality in HRQoL [47]. The CI's range is [-1,1], with the value 0 indicating perfect equality. A positive (negative) value indicates that the distribution of HRQoL is ‘pro-high SEP’ (‘pro-low SEP’) [48]. We compared CIs using the SEP score as the variable from which to rank individuals, with education and income.

Sensitivity analyses

We performed age-stratified analyses to assess whether the education and income weights derived for the composite SEP score differed across age groups. These analyses were conducted by running separate adjacent-category logistic regression analyses stratified by age groups (40–49; 50–65; and 66–79). Sex-specific analyses were also conducted, as well as analyses including only respondents who were currently in the labour force (full or part time). Lastly, we tested equivalising the household income variable with marital status.

We randomly split the sample in equal halves (referred to as Subsamples 1 and 2), before rerunning the adjacent-category logistic regression as in Eq. 1 on both samples. Next, we conducted the same procedure as in Eq. 2, using regression coefficients from Subsample 1, generating an alternative composite SEP score. With OLS regression, we tested how well the composite SEP score with weights from Subsample 1 performed in predicting HRQoL (EQ-5D and VAS) in Subsample 2. We assessed how these estimates (composite SEP score coefficients and the R2) differed from the analyses on HRQoL run on the full sample.

All statistical analyses were performed with Stata© version 15.1 (Stata Corporation, College Station, Texas).

Results

Descriptive statistics

Table 1 reports respondent characteristics. Given that this is a community sample, respondents were healthy in general, with a mean EQ-5D value of 0.89 and a mean VAS score of 0.76. Among them, 28.6% can be classified as in ‘full health’, i.e., they reported ‘no problems’ in all five EQ-5D-5L dimensions. The proportion of the sample with tertiary level education was larger than in the corresponding age group from the Norwegian population (50.0% compared to 33.1%, respectively) [49].

Table 1 Sample characteristics
Full size table

Weights for the composite SEP score

Table 2 displays the adjacent-category logistic regression output. Education was the main driver for subjective SEP, as demonstrated by the clear increase in the size of the education coefficient for each level change, particularly so for a postgraduate degree. For income, there was a non-linear increase in the size of the coefficients, with the highest income coefficient being thrice as large as the upper-middle income coefficient.

Table 2 Adjacent-category logistic regression on subjective SEP: weights for the composite SEP score based on education and income
Full size table

The weights derived were used to generate the composite SEP score (Eq. 2). The rescaled and rounded SEP score is reported in a ‘4X4 SEP table’ (Table 3). This table indicates that the observed non-linearities reported in Table 2 are reinforced when combining education and income levels.

Table 3 ‘4X4 SEP’ table, combining education and income levels
Full size table

Predicting variation in HRQoL with the composite SEP score

Table 4 provides the results from the OLS regression of the composite SEP score on both HRQoL measures (EQ-5D and VAS). A one-unit increase in SEP is associated with an average increase of 0.006 in the case of EQ-5D and 0.010 for VAS. Comparing the results with OLS regression of education and income separately led to a similar model fit based on the R2 (output not shown).

Table 4 Ordinary least squares regression on HRQoL (EQ-5D-5L values and VAS scores) with the composite SEP score as the independent variable
Full size table

Figure 2 presents age-adjusted mean EQ-5D values and VAS scores by SEP score levels. There was a clear linear increase in the reported HRQoL scores as the SEP values increased from 1 to 10, with EQ-5D values consistently higher than the VAS scores. The gradient was steeper for VAS (range: 0.72–0.82) than for EQ-5D (range: 0.87–0.92).

Fig. 2
figure 2

Age-adjusted mean EQ-5D values and VAS scores by composite SEP score. Mean VAS scores (left bars) and EQ-5D values (right bars) for each SEP score. SEP scores 6 and 9 are empty due to no data for these SEP score values. SEP: socioeconomic position

Full size image

The concentration indices of HRQoL are reported in Additional file 1. The CIs using the SEP score were 0.020 and 0.040 for EQ-5D and VAS, respectively. The CIs using education and income were slightly larger. The positive values of the CI indicate that better HRQoL were concentrated among respondents with a higher SEP.

Sensitivity analyses

Age-group stratified analyses of the determinants of subjective SEP indicated that the importance of education increased with age, whereas income became less important with age (Additional file 2). In terms of sex differences, the second-lowest education level was not statistically different from the reference among women. The patterns are the same as in the main model, with increasing coefficient sizes for each level increase in education and income (Additional file 3). Restricting the sample to respondents who stated being actively employed led to similar results as the main model, except for a non-significant upper secondary education coefficient (Additional file 4). Analysing household income equivalised for marital status did not lead to substantially different estimates. This output was therefore not included.

Using the composite SEP score with weights generated from Subsample 1 (Additional file 5) to predict both EQ-5D and VAS on Subsample 2, the coefficients remained similar and the change in R2 was marginal (Additional file 6).

Discussion

This paper has proposed a composite SEP score by modelling individuals’ subjective SEP based on four education and four income levels. The derived weights demonstrated how education and income influenced subjective SEP. There were non-linearities in determining subjective SEP, with greater importance placed on the higher education and income levels. These non-linearities became more evident when combining the different education and income levels, indicating that higher levels of education and income reinforced each other. The score was used to estimate inequalities in HRQoL based on combinations of education and income, and for each level of the composite SEP score. We found a clear gradient in HRQoL, with a linear increase from the bottom to the top of the score.

The proposed composite SEP score was derived from a measure of subjective SEP. This is in line with research recognising the added value of supplementing objective measures with subjective measures [12, 13, 50]. We contribute to the literature with a composite SEP score that captured both subjective and objective aspects of SEP, in which the objective indicators (education and income) estimated the subjective component (subjective SEP). The subjective SEP measure applied here differs from the more commonly used MacArthur scale of subjective social status [11]. Whereas the MacArthur scale is framed in terms of education, occupation and income, the subjective SEP measure is closely tied to occupation. Moreover, it should not be confused with occupational prestige, since the measure applied in this paper captures individuals’ perception of their own occupation’s social status, not the society’s judgement of the status of specific occupations [35].

The composite SEP score was estimated by education and income. Occupational category was not included in the estimation because we assumed that its influence on SEP was captured in its intermediate role between education (the determinant of occupation) and income (the reward of occupation). Moreover, in contrast to education (years) and income (money), the occupational categories are not as easily hierarchically ordered, in line with the arguments presented by Braveman et al. [8]. Education and income are more often consistently measured and available across different surveys and registers [27]. Besides, social standing derived from occupation is arguably more context dependent: a fisherman’s standing is likely judged higher in his local community than in the big city. We therefore followed Freeman et al. in omitting occupation [27].

Furthermore, the role of parental and early-life SEP when determining adult SEP must be acknowledged, the importance of which is consistently found to be substantial in the literature: children born to parents with higher SEP are more likely to prosper both in terms of socioeconomic achievements and in terms of health (e.g., [34, 51]). These factors are essential in the understanding of SEP.

The observed non-linear relationship of education and income in determining subjective SEP was evident from Table 3, with large marginal increases in subjective SEP from the highest education level, regardless of income. These non-linearities are likely to have different explanations. For example, Norway has a relatively egalitarian income distribution and a generous welfare state, which is likely to contribute to income being of less importance for most people. For the richest, however, income could matter more for SEP, potentially because social success can be signalled through various types of conspicuous consumption [52], such as living in a posh neighbourhood.

Age-stratified analyses added additional insights on cohort effects. Education appeared to matter more for the older age groups, whereas the size of the income coefficients decreased with age (Additional file 1). This could imply that education was a relatively stronger determinant of subjective SEP for those who did have higher education in the oldest age group (66–79). Indeed, the share of people opting for higher education has dramatically increased over the past generations, suggesting that higher education was more important for SEP when it was more of a privilege for the few. Cohort effects are also relevant in the case of sex differences (Additional file 3), in that women constitute a larger share of those taking higher education. The non-significant upper secondary/vocational coefficient in women could reflect that taking higher education is more important for women’s SEP than for men’s.

The relative importance of education and income in predicting SEP will likely vary between countries [53, 54]. If a similar analysis had been performed in a country with larger income inequalities than Norway, there would likely be starker differences between all the income categories, not only the top one as in this sample. Therefore, it is important to consider international differences in the relative importance of socioeconomic factors as determinants of SEP.

Our results indicated that the composite SEP score predicted considerable variation in HRQoL. Although there was no difference in the predictive power of the composite SEP score model compared to analysing education and income separately, it is arguably a more convenient way to calculate the combined impact of education and income on health inequalities, rather than conducting separate analyses [17, 55]. Moreover, the composite SEP score allowed us to demonstrate a linear increase in the age-adjusted HRQoL value by SEP score level for both the EQ-5D and the VAS (Fig. 2). This indicates a clear social gradient in both HRQoL measures, a message that would be hard to communicate with separate indicators.

The use of an alternative measure of inequality, the concentration index, suggested that inequalities in HRQoL are concentrated among higher-SEP groups, although the degree of inequality is relatively low (Additional file 1). The CIs of education and income were slightly larger than those of the composite SEP score, which could suggest that the combination of education and income somewhat compensates for differential variation in these two SEP variables. The order of magnitude of these results is comparable to other studies investigating inequalities in HRQoL [48, 56].

For the split-sample analyses, the estimates from the OLS analysis with the alternative composite SEP score (Additional file 6) did not differ greatly from the results in Table 4. This suggests that our estimates were internally valid.

Strengths and limitations

The key contribution of the current paper is that we provide a new application of a regression-based method for developing a composite SEP score with empirically derived education and income weights along a [1-10] SEP scale. Since education is grouped into the standard four levels, and income is approximately grouped into quartiles, our proposed approach can be replicated in any cohort study that collects these data, on any health outcome. Second, we provide new insights into the relative importance of different education and income levels as sources of SEP. Third, we have shown how SEP in the form of a composite score can be applied in analyses of health inequalities.

Some limitations should be acknowledged. First, the sample consists of respondents aged 40–79, leaving out the younger segment of the adult population. Second, since the subjective SEP measure targets people in the labour force, there is a risk that respondents who did not work at the time of the survey did not answer the question ‘correctly’, although the question specified that those who were not currently working should think about their latest occupation, assuming that an individual’s previous occupation is important for their current SEP. Sensitivity analyses indicated that including only currently employed respondents did not dramatically differ from the main results. Third, although a missing rate of observations of 4.5% is relatively small, there could still be systematic differences between the included and excluded shares of the sample. Missing value analysis indicated that those not reporting education or income were older and had a larger proportion of women compared to the full sample. We therefore cannot rule out that our results could underestimate inequalities, since older respondents would be more likely to report a lower HRQoL.

Conclusions

Our results suggest that a composite SEP score should be considered when studying social inequalities in health. We have proposed a model for a composite SEP score that predicts individuals’ SEP based on empirically weighted combinations of education and income levels, which identified a clear social gradient in HRQoL. This approach could be used when data on education and income are collected, either in cohort studies or through registers, potentially predicting the SEP of the entire population. The weights derived in this paper are relevant in a Norwegian context. Research from other countries is needed to compare the relative importance of education and income as determinants of SEP across countries, and to investigate how a composite SEP score would predict health inequalities in other institutional contexts.

Availability of data and materials

The data that support the findings of this study are available from The Tromsø Study but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. In order to get access to the data on which the present study is based, permission from the Tromsø Study is required. The Data and Publication Committee of the Tromsø Study evaluates all applications for access to data, and upon approval of application, an agreement is made between The Tromsø Study and the project manager of the project in question. Questions regarding access to data may be directed towards tromsous@ism.uit.no.

Abbreviations

CI:

Concentration Index

EQ-5D:

EuroQol Five-dimensional Measure of Health-related Quality of Life

HRQoL:

Health-Related Quality of Life

OLS:

Ordinary Least Squares

SD:

Standard Deviation

SE:

Standard Error

SEP:

Socioeconomic Position

VAS:

Visual Analogue Scale

WePP:

Western Preference Pattern

References

  1. Marmot M, Friel S, Bell R, Houweling TAJ, Taylor S. Closing the gap in a generation: health equity through action on the social determinants of health. The Lancet. 2008;372(9650):1661–9. https://doi.org/10.1016/S0140-6736(08)61690-6.

    Article  Google Scholar 

  2. Marmot M. Social justice, epidemiology and health inequalities. Eur J Epidemiol. 2017;32:537–46. https://doi.org/10.1007/s10654-017-0286-3.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Donkin AJ. Social Gradient. The Wiley Blackwell Encyclopedia of Health, Illness, Behavior, and Society. 2014:2172–8. https://doi.org/10.1002/9781118410868.wbehibs530

  4. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Davey SG. Indicators of socioeconomic position (part 1). J Epidemiol Community Health. 2006;60:7–12. https://doi.org/10.1136/jech.2004.023531.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Lynch J, Kaplan G. Socioeconomic Position. In: Berkman LF, Kawachi I, editors. Social Epidemiology. New York: Oxford University Press; 2000. p. 13–35.

    Google Scholar 

  6. Galobardes B, Lynch JW, Smith GD. Measuring socioeconomic position in health research. Br Med Bull. 2007;81:21. https://doi.org/10.1093/bmb/ldm001.

    Article  PubMed  Google Scholar 

  7. Marmot M. The Status Syndrome: How Social Standing Affects Our Health and Longevity. New York City: Holt Paperback; 2004.

    Book  Google Scholar 

  8. Braveman PA, Cubbin C, Egerter S, Chideya S, Marchi KS, Metzler M, et al. Socioeconomic status in health research: one size does not fit all. JAMA. 2005;294:2879–88. https://doi.org/10.1001/jama.294.22.2879.

    Article  CAS  PubMed  Google Scholar 

  9. Geyer S, Hemström Ö, Peter R, Vågerö D. Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. J Epidemiol Community Health. 2006;60:804–10. https://doi.org/10.1136/jech.2005.041319.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Demakakos P, Biddulph JP, de Oliveira C, Tsakos G, Marmot MG. Subjective social status and mortality: the English Longitudinal Study of Ageing. Eur J Epidemiol. 2018;33:729–39. https://doi.org/10.1007/s10654-018-0410-z.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Adler NE, Epel ES, Castellazzo G, Ickovics JR. Relationship of subjective and objective social status with psychological and physiological functioning: Preliminary data in healthy. White women Health Psychol. 2000;19:586. https://doi.org/10.1037/0278-6133.19.6.586.

    Article  CAS  PubMed  Google Scholar 

  12. Wilkinson RG. Health, hierarchy, and social anxiety. Ann N Y Acad Sci. 1999;896:48–63. https://doi.org/10.1111/j.1749-6632.1999.tb08104.x.

    Article  CAS  PubMed  Google Scholar 

  13. Demakakos P, Nazroo J, Breeze E, Marmot M. Socioeconomic status and health: the role of subjective social status. Soc Sci Med. 2008;67:330–40. https://doi.org/10.1016/j.socscimed.2008.03.038.

    Article  PubMed  Google Scholar 

  14. Diemer MA, Mistry RS, Wadsworth ME, López I, Reimers F. Best practices in conceptualizing and measuring social class in psychological research. Anal Soc Issues Public Policy. 2013;13:77–113. https://doi.org/10.1111/asap.12001.

    Article  Google Scholar 

  15. Blakemore T, Strazdins L, Gibbings J. Measuring family socioeconomic position. Australian Social Policy. 2009;8(121–168):121–68.

    Google Scholar 

  16. Adler NE, Boyce T, Chesney MA, Cohen S, Folkman S, Kahn RL, et al. Socioeconomic status and health: the challenge of the gradient. Am Psychol. 1994;49:15. https://doi.org/10.1037/0003-066X.49.1.15.

    Article  CAS  PubMed  Google Scholar 

  17. Moreno-Maldonado C, Rivera F, Ramos P, Moreno C. Measuring the Socioeconomic Position of Adolescents: A Proposal for a Composite Index. Soc Indic Res. 2018;136:517–38. https://doi.org/10.1007/s11205-017-1567-7.

    Article  Google Scholar 

  18. Joint Research Centre of the European Commission. Handbook on constructing composite indicators: Methodology and user guide. Paris: OECD Publishing; 2008. p. 13–4.

    Google Scholar 

  19. Oakes JM, Rossi PH. The measurement of SES in health research: current practice and steps toward a new approach. Soc Sci Med. 2003;56:769–84. https://doi.org/10.1016/s0277-9536(02)00073-4.

    Article  PubMed  Google Scholar 

  20. Hollingshead AB. Four factor index of social status. Yale Journal of Sociology. 1975;8:11–51.

    Google Scholar 

  21. Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health. 1997;18(1):341–78. https://doi.org/10.1146/annurev.publhealth.18.1.341.

    Article  CAS  PubMed  Google Scholar 

  22. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Davey SG. Indicators of socioeconomic position (part 2). J Epidemiol Community Health. 2006;60:95–101. https://doi.org/10.1136/jech.2004.028092.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Boyd M, Nam CB. The newest Nam-Powers-Boyd occupational scale: Development and insights. Southern Demographic Association Annual Meeting; 2015; San Antonio, Texas. Available from: http://www.npb-ses.info/latest-scores.html. Accessed 18 January 2021.

  24. Office for National Statistics. The National Statistics Socio-economic classification (NS-SEC). 2010. https://www.ons.gov.uk/methodology/classificationsandstandards/otherclassifications/thenationalstatisticssocioeconomicclassificationnssecrebasedonsoc2010 . Accessed 15 January 2020.

    Google Scholar 

  25. Darin-Mattsson A, Fors S, Kåreholt I. Different indicators of socioeconomic status and their relative importance as determinants of health in old age. Int J Equity Health. 2017;16(1):173. https://doi.org/10.1186/s12939-017-0670-3.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Hoebel J, Maske UE, Zeeb H, Lampert T. Social Inequalities and Depressive Symptoms in Adults: The Role of Objective and Subjective Socioeconomic Status. PLoS ONE. 2017;12(1): e0169764. https://doi.org/10.1371/journal.pone.0169764.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Freeman A, Tyrovolas S, Koyanagi A, Chatterji S, Leonardi M, Ayuso-Mateos JL, et al. The role of socio-economic status in depression: results from the COURAGE (aging survey in Europe). BMC Public Health. 2016;16(1):1098. https://doi.org/10.1186/s12889-016-3638-0.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Chichlowska KL, Rose KM, Diez-Roux AV, Golden SH, McNeill AM, Heiss G. Individual and Neighborhood Socioeconomic Status Characteristics and Prevalence of Metabolic Syndrome. The Atherosclerosis Risk in Communities (ARIC) Study. Psychosom Med. 2008;70(9):986. https://doi.org/10.1097/PSY.0b013e318183a491.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Saisana M, Saltelli A, Tarantola S. Uncertainty and sensitivity analysis techniques as tools for the quality assessment of composite indicators. J R Stat Soc Ser A Stat Soc. 2005;168:307–23. https://doi.org/10.1111/j.1467-985X.2005.00350.x.

    Article  Google Scholar 

  30. Kinge JM, Modalsli JH, Øverland S, Gjessing HK, Tollånes MC, Knudsen AK, et al. Association of Household Income With Life Expectancy and Cause-Specific Mortality in Norway, 2005–2015. JAMA. 2019;321:1916–25. https://doi.org/10.1001/jama.2019.4329.

    Article  PubMed  PubMed Central  Google Scholar 

  31. Ernstsen L, Strand BH, Nilsen SM, Espnes GA, Krokstad S. Trends in absolute and relative educational inequalities in four modifiable ischaemic heart disease risk factors: repeated cross-sectional surveys from the Nord-Trøndelag Health Study (HUNT) 1984–2008. BMC Public Health. 2012;12(1):266. https://doi.org/10.1186/1471-2458-12-266.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Tetzlaff F, Epping J, Tetzlaff J, Golpon H, Geyer S. Socioeconomic inequalities in lung cancer – a time trend analysis with German health insurance data. BMC Public Health. 2021;21(1):538. https://doi.org/10.1186/s12889-021-10576-4.

    Article  PubMed  PubMed Central  Google Scholar 

  33. McFadden E, Luben R, Bingham S, Wareham N, Kinmonth A-L, Khaw K-T. Social inequalities in self-rated health by age: Cross-sectional study of 22 457 middle-aged men and women. BMC Public Health. 2008;8(1):230. https://doi.org/10.1186/1471-2458-8-230.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Case A, Fertig A, Paxson C. The lasting impact of childhood health and circumstance. J Health Econ. 2005;24(2):365–89. https://doi.org/10.1016/j.jhealeco.2004.09.008.

    Article  PubMed  Google Scholar 

  35. Fujishiro K, Xu J, Gong F. What does “occupation” represent as an indicator of socioeconomic status?: Exploring occupational prestige and health. Soc Sci Med. 2010;71(12):2100–7. https://doi.org/10.1016/j.socscimed.2010.09.026.

    Article  PubMed  Google Scholar 

  36. Ahlburg D. Intergenerational Transmission of Health. Am Econ Rev. 1998;88(2):265–70.

    Google Scholar 

  37. Black SE, Devereux PJ, Salvanes KG. Why the apple doesn’t fall far: Understanding intergenerational transmission of human capital. Am Econ Rev. 2005;95(1):437–49. https://doi.org/10.1257/0002828053828635.

  38. Dolton P, Xiao M. The intergenerational transmission of body mass index across countries. Econ Hum Biol. 2017;24:140–52. https://doi.org/10.1016/j.ehb.2016.11.005.

    Article  PubMed  Google Scholar 

  39. Jacobsen BK, Eggen AE, Mathiesen EB, Wilsgaard T, Njølstad I. Cohort profile: The Tromsø Study. Int J Epidemiol. 2011;41:961–7. https://doi.org/10.1093/ije/dyr049.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Brazier J, Ratcliffe J, Saloman J, Tsuchiya A. Measuring and valuing health benefits for economic evaluation. New York: Oxford University Press; 2017.

  41. Wisløff T, Hagen G, Hamidi V, Movik E, Klemp M, Olsen JA. Estimating QALY Gains in Applied Studies: A Review of Cost-Utility Analyses Published in 2010. Pharmacoeconomics. 2014;32:367–75. https://doi.org/10.1007/s40273-014-0136-z.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Rabin R, Gudex C, Selai C, Herdman M. From translation to version management: a history and review of methods for the cultural adaptation of the EuroQol five-dimensional questionnaire. Value Health. 2014;17:70–6. https://doi.org/10.1016/j.jval.2013.10.006.

    Article  PubMed  Google Scholar 

  43. Herdman M, Gudex C, Lloyd A, Janssen M, Kind P, Parkin D, et al. Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L). Qual Life Res. 2011;20:1727–36. https://doi.org/10.1007/s11136-011-9903-x.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Olsen JA, Lamu AN, Cairns J. In search of a common currency: A comparison of seven EQ-5D-5L value sets. Health Econ. 2018;27:39–49. https://doi.org/10.1002/hec.3606.

    Article  PubMed  Google Scholar 

  45. Fagerland MW. adjcatlogit, ccrlogit, and ucrlogit: Fitting ordinal logistic regression models. Stata J. 2014;14:947–64. https://doi.org/10.1177/1536867X1401400414.

    Article  Google Scholar 

  46. Mehta HB, Mehta V, Girman CJ, Adhikari D, Johnson ML. Regression coefficient–based scoring system should be used to assign weights to the risk index. J Clin Epidemiol. 2016;79:22–8. https://doi.org/10.1016/j.jclinepi.2016.03.031.

    Article  PubMed  Google Scholar 

  47. Wagstaff A, Paci P, van Doorslaer E. On the measurement of inequalities in health. Soc Sci Med. 1991;33(5):545–57. https://doi.org/10.1016/0277-9536(91)90212-U.

    Article  CAS  PubMed  Google Scholar 

  48. JieAnNaMu, Xu X, You H, Gu H, Gu J, Li X, et al. Inequalities in health-related quality of life and the contribution from socioeconomic status: evidence from Tibet, China. BMC Public Health. 2020;20(1):630. https://doi.org/10.1186/s12889-020-08790-7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Statistics Norway: Educational attainment of the population. https://www.ssb.no/en/statbank/table/08921/ (2016). Accessed 8 Mar 2021.

  50. Singh-Manoux A, Adler NE, Marmot MG. Subjective social status: its determinants and its association with measures of ill-health in the Whitehall II study. Soc Sci Med. 2003;56:1321–33. https://doi.org/10.1016/S0277-9536(02)00131-4.

    Article  PubMed  Google Scholar 

  51. Cohen S, Janicki-Deverts D, Chen E, Matthews KA. Childhood socioeconomic status and adult health. Ann N Y Acad Sci. 2010;1186(1):37–55. https://doi.org/10.1111/j.1749-6632.2009.05334.x.

    Article  PubMed  Google Scholar 

  52. Veblen T. The Theory of the Leisure Class. New York: MacMillan; 1899.

    Google Scholar 

  53. Präg P, Mills MC, Wittek R. Subjective socioeconomic status and health in cross-national comparison. Soc Sci Med. 2016;149:84–92. https://doi.org/10.1016/j.socscimed.2015.11.044.

    Article  PubMed  Google Scholar 

  54. Quon EC, McGrath JJ. Subjective socioeconomic status and adolescent health: a meta-analysis. Health Psychol. 2014;33:433. https://doi.org/10.1037/a0033716.

    Article  PubMed  Google Scholar 

  55. Parrott MD, Shatenstein B, Ferland G, Payette H, Morais JA, Belleville S, Kergoat MJ, Gaudreau P, Greenwood CE. Relationship between diet quality and cognition depends on socioeconomic position in healthy older adults. J Nutr. 2013;143(11):1767–73. https://doi.org/10.3945/jn.113.181115.

    Article  CAS  PubMed  Google Scholar 

  56. Gundgaard J, Lauridsen J. A decomposition of income-related health inequality applied to EQ-5D. Eur J Health Econ. 2006;7(4):231–7. https://doi.org/10.1007/s10198-006-0360-3.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We thank all the participants in the Tromsø Study. We are thankful to colleagues and conference/seminar participants for helpful comments to previous versions.

Funding

Open Access funding provided by UiT The Arctic University of Norway. This project was part of the Tracing causes of inequalities in health and wellbeing project funded by the Research Council of Norway, grant 273812. The Tromsø Study of UiT – The Arctic University of Norway (UiT) provided the data, and the Department of Community Medicine at UiT funded the study. The study sponsors had no role in the design and implementation of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. The publication charges for this article have been funded by a grant from the publication fund of UiT.

Author information

Authors and Affiliations

  1. Department of Community Medicine, UiT – The Arctic University of Norway, 9037, Tromsø, Norway

    Marie Hella Lindberg, Jan Abel Olsen & Birgit Abelsen

  2. Centre for Health Economics, Monash University, Melbourne, Australia

    Gang Chen & Jan Abel Olsen

  3. Division of Health Services, Norwegian Institute of Public Health, Oslo, Norway

    Jan Abel Olsen

Authors
  1. Marie Hella Lindberg
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Gang Chen
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Jan Abel Olsen
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Birgit Abelsen
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: MHL, GC, JAO, BA. Material preparation, analysis and interpretation of data: MHL. Drafting of the manuscript: MHL. Critical revision of the manuscript for important intellectual content: MHL, GC, JAO, BA. All authors read and approved the final version to be published.

Corresponding author

Correspondence to Marie Hella Lindberg.

Ethics declarations

Ethics approval and consent to participate

This study was approved by the Regional Committee for Medical Research Ethics Northern Norway (REK North; ID 2019/607). The Tromsø Study complies with the Declaration of Helsinki and all participants gave written informed consent before admission.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1.

Concentration index of EQ-5D and VAS

Additional file 2.

Adjacent-category logisticregression on subjective social status: weights for composite SEP score,stratified by age groups.

Additional file 3.

Adjacent-category logistic regression onsubjective SEP: stratified by sex.

Additional file 4.

Adjacent-category logistic regression onsubjective SEP, including only currently employed respondents (full or parttime).

Additional file 5.

Adjacent-category logistic regression onsubjective social status: weights for composite SEP score with sample randomlysplit in two.

Additional file 6.

Ordinary least squares regression analysis to test internalvalidity.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lindberg, M.H., Chen, G., Olsen, J.A. et al. Combining education and income into a socioeconomic position score for use in studies of health inequalities. BMC Public Health 22, 969 (2022). https://doi.org/10.1186/s12889-022-13366-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12889-022-13366-8

Keywords

BMC Public Health

ISSN: 1471-2458

Contact us