Socioeconomic determinants of myalgic encephalomyelitis/chronic fatigue syndrome in Norway: a registry study

doi:10.21203/rs.3.rs-2755999/v1

Background: Previous research has shown that socioeconomic status (SES) is a strong predictor of chronic disease. However, to the best of our knowledge, there has been no studies of how SES affects the risk of Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) that has not been based upon self-reporting or retrospectively screening of symptoms. As far as we know, this is therefore the first study that isolate and describe socioeconomic determinants of ME/CFS and calculate how these factors relate to the risk of ME/CFS diagnosis by utilizing individual level registry data, which allows for objective operationalization of the ME/CFS population, and the utilization of different control groups.

Data and methods: We utilize health registry data from all adult patients diagnosed with ME/CFS from 2016-2018 in Norway, coupled with socioeconomic data from statistics Norway from 2009-2018. We operationalize SES as household income and educational attainment fixed at the beginning of the study period. We compare the effects of SES on the risk of ME/CFS diagnosis to a population of patients with hospital diagnoses that share clinical characteristics of ME/CFS and a healthy random sample of the Norwegian population. Our models are estimated by logistic regression analyses.

Results: When comparing the risk of ME/CFS diagnosis with a population consisting of people with four specific chronic diseases, we find that high educational attainment is associated with a 19% increase (OR: 1.19) in the risk of ME/CFS and that high household income is associated with a 18% increase (OR:0.82) in risk of ME/CFS. In model 2, when comparing with a healthy population sample, we find that low educational attainment is associated with 69% decrease (OR:0.31) in the risk of ME/CFS and that low household income is associated with a 53% increase (OR: 1.53).

Conclusion: We find statistically significant associations between SES and the risk of ME/CFS. However, our more detailed analyses shows that our findings vary according to which population we compare the ME/CFS patients with, and that the effect of SES is larger when comparing with a healthy population sample, as opposed to controls with selected hospital diagnoses.

In this paper we study the relationship between socioeconomic status (SES) and the risk of receiving ME/CFS diagnosis in the Norwegian population. Myalgic encephalomyelitis (ME), or chronic fatigue syndrome (CFS), is regarded as a complex and highly disabling chronic disease with currently unknown pathophysiology, which has no approved cure. Those suffering from ME/CFS have a complex assortment of immunological, neurological, and psychological symptoms, resulting in high levels of comorbidity. The physiological symptoms are diverse and include chronic pain, sensory hypersensitivity, and severe fatigue (e.g., post exertional malaise) (Fluge et.al, 2021). These symptoms often persist in combination with psychological ailments, such as cognitive decline and depression. The burden from ME/CFS is severe both in an individual and societal perspective (Jason et al. 2008).

A systematic review of the prevalence and incidence of ME/CFS in Europe from 1994-2019 shows that prevalence ranges from 0.1% to 2.2% depending on diagnosis criteria, study designs and populations (Lopez et.al, 2020). Women are approximately 3 times higher at risk of ME/CFS diagnosis and that the risk is highest in the age groups 15-25 and 35-45 (Hilland & Anthun, 2022; Muller et al., 2014). The comorbid and chronic nature of the disease results in a very high and sustained use of various health services from those affected by the disease, and yearly economic costs of ME/CFS has been estimated to be substantial (Jason & Mirin, 2021). In addition, studies find that caregivers of those affected are also highly burdened (Fluge et.al, 2021).

Therefore, the various aspects of ME/CFS should be understood as much as possible. Foremost to reduce the suffering of those afflicted by ME/CFS and their caregivers, but also to lower the societal costs associated with the disease. The disease is not well enough understood to implement effective preventative measures and treatments, especially when it comes to those that are moderately or severely afflicted by it. As of today, there is a lack of studies about the relationship between SES and ME/CFS, even though we know that socioeconomic status is a strong predictor for the onset of an array of diseases, especially diseases with comorbid facets and complex pathologies between somatic and psychological symptoms.

We argue that the relationship between SES and ME/CFS is understudied and that the few studies that have been conducted on the subject have utilized data that are prone producing biased results, mainly stemming from a small number of participant and self-reporting. In addition, there is a rather unique problem of selection bias when studying ME/CFS due to the lack of medical tests and biomarkers for the disease. This is an essential factor that underpins the whole discourse, one that is often overlooked in individual studies of the disease. The lack of objective medical tests results in heterogeneity challenges in defining the ME/CFS populations in epidemiological studies, which in turn reduce the transferability of findings from studies on ME/CFS populations across study settings (e.g., across time or between countries). This further strengthens the rationale for our study of the relationship between SES and the risk of ME/CFS in Norway, using comprehensive registry data.

The relationship between socioeconomic status and chronic disease

Socioeconomic status can be defined in several ways. We follow Psaki and colleagues' (2014) definition of the concept:

"Socioeconomic status (SES) is a theoretical construct encompassing individual, household, and/or community access to resources. It is commonly conceptualized as a combination of economic, social, and work status, measured by income or wealth, education, and occupation, respectively" (Psaki et al., 2014).

Several studies find correlations between low SES and various health outcomes, including a higher risk of chronic diseases, the disease group of which ME/CFS is a part of (Pollack et al., 2007; Black et al., 2008). It is therefore vital to account for the multifaceted theoretical construction of SES when operationalizing the concept for quantitative analyses. Not doing so, for instance if operationalizing SES as only income, or only educational attainment, could lead to results with questionable validity. Diemer at al. (2013) problematize the utilization of incomplete SES measures and reviews the literature for best practice when operationalizing SES. They conclude that the "gold standard" of measuring SES as an objective quantification of social class involves three parts: occupation, educational attainment, and income (Diemer et al., 2013).

This conceptualization of SES does not capture subjective social status (SSS), which is also a relevant predictor for both physical and mental health (Adler et al. 2000; Singh-Manoux et al., 2005). SSS is often explored using qualitative data to gain knowledge of the subjective aspect of social standing. It is however outside the scope and objectives of this study, and we therefore explicitly focus on objective social class affiliation operationalized as various aspects of quantifiable SES.

We follow Diemer and colleagues (2013) and operationalize SES as comprised of both income level and educational attainment. This approach is also supported in other studies, for instance by Callahan & Eyberg (2010) which found that a model containing separate variables for income and educational attainment explained three times more of the variance observed in the outcome of study (Diemer et al., 2013). Therefore, instead of utilizing composite measures such as indexes that collapse different SES categories, we model the effect of each SES variable separately, which is argued to be the best approach to preserve the individual effects of each SES variable (Duncan & Magnuson, 2003; Williams, 2009). Unfortunately, we do not have the data that would be required test occupation as a SES-indicator.

Previous research on SES and ME/CFS

To our knowledge, the first study of the link between SES and the risk of ME/CFS was initiated by the U.S Centre for Disease Control (CDC) in the latter part of the 1980s and published in 1993 (Gunn et al., 1993). In this epidemiological study, the authors concluded that highly educated Caucasian women, with the potential for high income, comprised most of the ME/CFS cases. However, later studies did not find statistically significant differences between various socioeconomic strata and the risk of developing ME/CFS (Lloyd et al., 1990; Katz et al., 2009; Karfakis, 2018). The findings in the discourse are divergent, as other studies again do find that SES status is a predictor for ME/CFS. Interestingly, several studies find that high SES status is a predictor (Rangel et al., 2000; Song et al., 1999; Viner & Hotopf, 2004), which contradicts research on the relationship between SES and the onset of chronic disease in general (Mair & Jani, 2020). While other studies find that the onset of ME/CFS is associated with middle-to-low socioeconomic status, which contradicts these findings (Jason et al., 1999; Kocalevent et al., 2011).

Furthermore, there are several studies showing an association between unexplained fatigue and various psychological disorders, disorders that are like the psychological ailments associated with ME/CFS (Addington & Ford, 2001; Skapinakis et al., 2003; Leone, 2010; Huibers et al., 2007; Corfield et al., 2016). This is a strain of the literature that is worth mentioning, given the heterogeneity challenges posed by lack of diagnostic tests for ME/CFS. However, it is important to mention that we do not equate between ME/CFS and these psychological diseases or with "unexplained" fatigue in general, even though these disorders share many of the same symptoms. However, we do believe they are worth including in this overview of previous research, as the literature on SES and ME/CFS specifically is so scarce and full of divergent findings.

In sum, the literature shows that socioeconomics potentially influences the probability of receiving an ME/CFS diagnosis, even though these variables alone far from adequately explains the onset of the disease. Nevertheless, research shows that further understanding of the socioeconomic status of people with ME/CFS is needed to understand the relationship between socioeconomics and the disease especially given the challenges posed by selection bias when studying ME/CFS.

We tested the relationship between SES and the risk of ME/CFS diagnosis by utilizing individual level registry data from two sources. Hospital visits and diagnoses were collected from the Norwegian Patient Register (NPR), and SES-related data were collected from Statistics Norway. Data from the Norwegian Patient Register has been used in this publication. The interpretation and reporting of these data are the sole responsibility of the authors, and no endorsement by the Norwegian Patient Register is intended or should be inferred. The project was reviewed and approved by the Regional Committees for Medical and Health Research Ethics (reference 2018/293) and it is not possible to identify individual patients from the data presented in this paper.

Study design and population

This study was designed as a pooled cross-sectional analysis where we analysed the probability for ME/CFS diagnosis by pooling two different populations with the ME/CFS population. One of the major challenges when studying how different factors relate to the risk of ME/CFS diagnosis is the disease' long and complex diagnostic process. Therefore, our study was designed with an early measurement/exposure of SES, combined with a long ME/CFS "washout" period of 5-7 years. As the objective of our study was to test how levels of SES prior to ME/CFS diagnosis affects the probability of diagnosis, we modelled how the 2011 values of our SES variables affected the probability of receiving an ME/CFS diagnosis 5-7 years later. This way, the effects of ME/CFS onset on each individual' values of SES (i.e., reverse causality) was reduced. In addition, we conducted a washout of individuals with G93.3 diagnosis from 2009-2015, meaning that we removed all individuals with ME/CFS diagnosis prior to 2016. This long washout period considers the complex diagnostic process of ME/CFS and reduce the probability that factors related to the onset of the disease affects the levels of SES for each individual.

Another major challenge when studying the relationship between SES and ME/CFS has often been a lack of viable comparison groups, or control groups, in small N studies based on self-reporting. Our data allowed for the utilization of objectively defined control groups consisting of randomly sampled healthy individuals and selected individuals with specific ICD-10 diagnoses that share some of the characteristics of ME/CFS disease.

The ME/CFS population was defined as all patients registered with ICD-10 code G93.3 at least once in the specialized hospital care during the years 2016-2018 but not once in the years 2009-2015. The random sample of the healthy population was created by randomly selecting 0.1 percent of the total Norwegian population from the registries. As we wanted to study the effect of SES on the risk of ME/CFS relative to a healthy, randomly selected population, we removed all individuals with hospital diagnoses in the Norwegian patient registry (NPR). The chronically ill patient population was created by drawing individuals registered in NPR with the following ICD-10 diagnoses: C50 "Malignant neoplasm of breast", M79.7 "Fibromyalgia", G35 "Multiple sclerosis", and A962 "Lyme disease". These diagnoses were chosen because they include other chronically diseases with increased disease burden for a long but limited time frame. In addition, they either share clinical characteristics with ME/CFS and/or disproportionately affect women, which previous research has shown is the case also for ME/CFS (Muller et al., 2014; Hilland & Anthun, 2022). Table 1 shows the number of individuals in each population.

Table 1: Number of individuals per study population.

Population	Number of individuals
ME/CFS population after washout	5,478
Healthy random sample population	4,661
Controls with the four selected NPR diagnoses	113,971

Household income

We operationalized income as a categorical variable consisting of three categories, from low income to high income. Low income was defined as household income below 40 percent of the median household income, whilst high income was defined as above 70 percent of the median. We followed Diemer and colleagues (2013) review of best practices when conceptualizing SES by operationalizing income level as a function of household income, as opposed to individual income. Individual income can lead to inaccurate depictions of social class affiliation if individuals work part-time but are part of a two- person high-income household, thus being affiliated with a higher social class than what is evident from the individual income level. Medium household income was the reference category in our regression analyses.

Educational Attainment

Educational attainment was constructed at the family level as the highest level of education for either the individual, or either of the parents in 2011. We measured educational attainment as a categorical variable consisting of three categories: low education, medium education, and high education. Low education was defined as having no education or having completed only pre-school and elementary school. Medium educational attainment was defined as having completed lower secondary school and secondary school, whilst high educational attainment was defined as having completed a university degree, at any level. Medium educational attainment was the reference category in our regression analyses.

Control variables

In addition to our independent SES variables, we controlled for occupational status, marital status, gender and age. Occupational status is operationalized as a dichotomous variable where the effect of being employed is included in the model. Marital status is operationalized as a categorical variable, and we model the effect of being married and separated/divorced relative to being single. Gender is operationalized as a dichotomous variable and the effect of female sex is included in the model. Age is a categorical variable consisting of 7 categories, and the effect of each age group is modelled relative to the 0-17 age group.

Our variables are operationalized as categories, and table 2 shows the weighted distributions of individuals in each category of our variables. In table 2, we only include individuals between 18 and 75 years to reduce the effects of age on the distributions, making comparisons between the different population more variable. However, we also show the number of individuals per age group, divided by gender and population, to give an overview of the age distribution in each population.

Table 2: Summary statistics per category of each variable, as weighted percentage, by gender and population.

Hospital controls		Healthy population		ME/CFS
Men	Women	Men	Women	Men	Women
Low household income	5.1	7.7	6.0	6.4	6.7	10.7
Medium household income	85.0	85.4	85.8	89.7	87.3	85.4
High household income	10.0	6.9	8.2	3.9	6.1	3.9
Low educational attainment	3.5	5.6	12.6	13.4	8.3	10.4
Medium educational attainment	89.2	88.7	79.8	78.2	83.1	83.0
High educational attainment	7.3	5.7	7.7	8.4	8.7	6.6
Not working	48.2	33.8	14.9	14.2	27.8	26.7
Working	51.8	66.2	85.2	85.8	72.2	73.3
Single	16.3	24.0	50.3	44.2	57.5	51.4
Married	60.1	50.8	38.6	40.3	31.5	33.5
Divorced/separated	23.6	25.2	11.2	15.5	11.0	15.1
Age 0-17	0.2	0.1	13.1	8.2	17.8	5.7
Age 18-24	0.2	1.2	12.1	14.4	13.7	17.4
Age 25-34	0.6	4.7	17.9	14.9	17.1	18.3
Age 35-44	2.0	12.8	17.2	20.3	17.8	26.0
Age 45-54	7.9	27.6	19.3	23.9	20.3	24.7
Age 55-64	16.6	20.1	10.9	11.8	10.7	6.8
Age 65+	72.4	33.5	9.4	6.5	2.7	1.1
N	40,719	73,252	2,368	2,293	1,174	4,382

Statistical analyses

We tested how the 2011 values of our independent variables predicted the risk of receiving an ME/CFS diagnosis 5-7 years later through logistic regression analyses. We modelled how socioeconomic status affected the probability of receiving a ME/CFS diagnosis (ICD10, G93.3) in specialized healthcare. Our outcome variable was dichotomous with the value of 0 if a patient did not receive G93.3 and 1 if an individual in our dataset had a ME/CFS diagnosis. This meant we had a binary outcome, which rendered a normal OLS regression useless as its core assumptions of linearity and homoscedasticity were breached, and we also risked predicting values that are outside the 1-0 dichotomy (Mehmetoglu & Jakobsen, 2017). Therefore, we conducted logistic regression analyses. We were interested in estimating how much the natural logarithm of the odds for Y being equal to 1 change for each unit change in the covariates. The equation of a logistic model can be written as

L_i = β₀ + β₁X_1i+ β₂ X_2i + β₃X_3i+ β_xZ_i+ β_k-1X_{k-1, i}

The total logit (L_i) was a linear function of the X-variables, and β_kthe number of parameters included in our model and the constant and X_k the number of variables. As previous research has shown that there are strong effects of gender on the risk of ME/CFS, we included interaction terms between gender and both household income and educational attainment. Z_i denotes the interaction term, where the effect of coefficients β_xZ_ifor Y, regressed on X at values of the moderator Z_i. In our analyses, we estimated the odds ratio instead of the logit coefficient, which meant that the logit was exponentiated to the change in odds for Y=1. An odds ratio of exactly 1 for one of our covariates meant that the specific covariate was associated with no change in Y, regardless of statistical significance level. An odds ratio above 1, for instance 1.5, meant that the odds for Y=1 increase by 50% for each unit increase in X. Conversely, negative odds ratios indicated a decrease in the odds of Y=1. As all our covariates were categorical, we estimated the effect of each category relative to a given reference category for each variable. All statistical analyses were conducted in Stata, version 16 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.).

In table 3 we present our results in a joint table consisting of two models. In each model we tested the relationship between SES and risk of ME/CFS diagnosis relative to 1) a healthy random sample of the general population and 2) chronically ill individuals with the ICD-10 diagnoses G35 (multiple sclerosis), C50 (malignant neoplasm of breast/breast cancer), M79.7 (fibromyalgia) and A962 (Lyme disease).

Table 3: Logistic regression analyses of the relationship between SES and ME/CFS. Estimations are relative to hospital diagnosed controls (Model 1) and a healthy randomly selected sample of the Norwegian population (Model 2). Odds ratio (standard deviations in parentheses).

	(1)	(2)
	Risk of ME/CFS relative to a hospital diagnosed controls	Risk of ME/CFS relative to a healthy population sample
Low educational attainment	0.894 (0.0512)	0.318^*** (0.0241)
High educational attainment	1.198^** (0.0800)	0.955 (0.0960)
Low household income	0.988	1.535^***
	(0.0577)	(0.155)
High household income	0.822^*	1.260
	(0.0649)	(0.152)
Working	0.715^***	0.420^***
	(0.0282)	(0.0263)
Married	0.836^***	0.870^*
	(0.0355)	(0.0605)
Divorced/separated	0.907	0.992
	(0.0495)	(0.0955)
Women	1.809^***	4.278^***
	(0.0748)	(0.215)
Age 18-24	0.434^***	1.796^***
	(0.0415)	(0.179)
Age 25-34	0.121^***	1.175
	(0.0110)	(0.117)
Age 35-44	0.0661^***	1.133
	(0.00603)	(0.117)
Age 45-54	0.0295^***	0.802^*
	(0.00271)	(0.0872)
Age 55-64	0.0101^***	0.350^***
	(0.00101)	(0.0426)
Age 65+	0.000730^***	0.0371^***
	(0.0000994)	(0.00571)
Constant	0.568^***	1.276^***
	(0.0667)	(0.0753)
N	119 449	10 137
R²	0.2834	0.1454

Exponentiated coefficients; Standard errors in parentheses

^* p < 0.05, ^** p < 0.01, ^*** p < 0.001

In model 1 we compared the risk of ME/CFS diagnosis with a population consisting of people with four specific chronic diseases. Our findings in model 1 showed that there was a statistically significant effect from low educational attainment and the risk of ME/CFS. Our models showed that high educational attainment increased the risk of ME/CFS by 19% (OR: 1.19), relative to medium educational attainment. We found no statistically significant effect from low educational attainment. There was a statistically significant relationship between household income and risk of ME/CFS diagnosis in Model 1. We found that high household income decreases the risk of ME/CFS by 18% (OR: 0.82). There was no effect from low household income in model 1. In model 1, we also found that employment reduced the risk of ME/CFS diagnosis by 29% (OR: 0.71). Another control variable in our models was marital status, and being married was associated with a 17% decrease (OR: 0.83) in risk of ME/CFS diagnosis, relative to being single. Furthermore, when comparing with the population consisting of chronically ill individuals, we found that women have an 80% increased risk of ME/CFS diagnosis (OR: 1.8), compared to men. For age, we modelled the risk of ME/CFS per age group relative to the 0-17 group. When comparing with the hospital diagnosed population sample, a population that skewed strongly towards older age, we saw that the risk of ME/CFS was highest in the age group 0-17 years, as the risk of ME/CFS was strongly reduced in all other age groups in the model.

In model two, we modelled the risk of ME/CFS relative to a healthy, randomly selected sample of the Norwegian population. We found a statistically significant relationship between educational attainment and risk of ME/CFS relative to the healthy population sample. Low educational attainment was statistically significantly associated with a 69% reduction (OR: 0.31) in the risk of ME/CFS relative to medium educational attainment, while there was no statistically significant relationship between high educational attainment and the risk of ME/CFS. The result from our logistic regression in model 2 showed a statistically significant relationship between low household income and the risk of ME/CFS (OR: 1.53) relative to medium household income. Low household income increased the risk of ME/CFS by 53%, relative to belonging to the medium household income group. Employment is associated with decreased risk of ME/CFS, as individuals that were employed had a 59% (OR=0.41) reduction in risk of ME/CFS diagnosis. We found that being married decreased the risk of ME/CFS diagnosis, by 14% (OR: 0.86), relative to being married. Furthermore, we found that women had a 4.27 time (OR: 4.27) greater risk of G93.3 diagnosis than men. For age, we modeled the risk of ME/CFS per age group relative to the 0-17 group. When comparing with the healthy random population sample, we saw that the risk of ME/CFS was highest in the age groups 18-24 (OR: 1.79).

Figure 1 shows the confidence intervals from each variable in the analysis, including the interaction effects.

Robustness tests

We conducted two robustness tests to account for the strong effects of gender and age on the risk for ME/CFS diagnosis. Firstly, we estimated the effect of age as a cubic spline to test if the effect of age would be better fitted as a non-linear continuous association. It has been argued that cubic spline interpolation is imperative in some cases to achieve a good fit between a continuous variable such as age, and a dichotomous outcome such as ME/CFS diagnosis (Gauthier et al. (2020. Our tests indicated that while there where strong drawbacks from modelling age either as a continuous or a dichotomized variable, there were no improvements in fit from modelling age as a cubic spline as opposed to a categorical variable consisting of 7 categories. We therefore present our findings with age as a categorical variable as these results are much more intuitive to interpret for the reader, as opposed to cubic spline knots.

We also tested if there were interaction effects between gender and our two SES variables (see appendix 1, table 4 and appendix 2, figures 2 and 3), as previous research has shown that there are strong effects from gender on the risk of ME/CFS diagnosis (cf. Hilland & Anthun, 2022). The model showed that both education and income interacted with gender. However, the interactions were barely statistically significant and only for specific age groups, which makes it hard to give meaningful interpretations of the interaction effects. We therefore only included the interaction effect in the appendix. As a final robustness check we conducted a Hosmer –Lemeshow goodness-of-fit test (Hosmer et al., 2013). The results from the goodness of fit test were statistically insignificant (not reported in tables), which is an indication of reasonable model fit.

There are two important factors that underpin the entire discussion of our findings, which is important to present before we discuss the results of our analyses. The first factor is related to how we operationalize ME/CFS in this study. We only include ME/CFS patients that are diagnosed in the specialized healthcare services in Norway with the ICD code G93.3 which excludes all individuals that are either diagnosed in primary care or that suffer from undiagnosed ME/CFS.

Secondly there is a lack of biomarkers for ME/CFS, rendering traditional medical tests unusable and resulting in ME/CFS having to be diagnosed clinically by the individual physicians, based on a combination of various diagnostic tools or guidelines and subjective assessments of the patient' symptoms. As we discussed in the introduction, this opens for a problem of selection bias due to heterogeneity stemming from differences in both individual patients and consulting physicians. This also increases the possibility of misdiagnosing individuals with ME/CFS. Taken together, these factors leave the interpretation of our findings vulnerable to type I and II errors, if one disregards these constraints posed by the operationalization of ME/CFS as a hospital diagnosis and the problem of selection bias that stems from clinically diagnosing the disease. However, we do believe that our study design reduces the probability of misinterpreting our findings, as we will discuss in this chapter.

We operationalized SES as educational attainment and household income. For educational attainment, the findings of our analyses diverged from the established discourse on the relationship between education level and the risk of chronic disease in general. Taken at face value, there seems to be a reverse effect of education for ME/CFS as we find that low educational attainment strongly reduces the risk of ME/CFS diagnosis, relative to medium educational attainment, when comparing with the healthy population sample controls. However, our analyses do not indicate that this is an epidemiological effect of low educational attainment that is unique for the risk of ME/CFS, rather it seems to reflect the relationship between educational attainment and the risk of chronic disease in general. When comparing with a population consisting of hospitalized controls this effect is already accounted for in the model, and we see that the effect of low education is reduced to 11% and just becomes statistically insignificant (p=0.055). However, this is very close to being statistically significant and we argue that this suggests that there are effects of educational attainment and the risk of ME/CFS even when comparing with a population that innately controls for the effect of educational attainment on the likelihood of hospital diagnosis. This finding should be taken together with the fact that we do find a statistically significant increase in risk of ME/CFS diagnosis for people with high educational attainment when comparing with hospital diagnosed controls.

This suggests that there is an effect of educational attainment on the risk of ME/CFS diagnosis, even when comparing with a population that innately controls for the existing relationship between SES and chronic disease in general. It is, however, unlikely that this is an epidemiological effect of educational attainment on the risk of ME/CFS, though it is theoretically possible given how poorly the disease is understood. What is more likely, is that given the unique characteristics of ME/CFS, specifically the lack of medical tests and the need for clinical diagnosis that naturally follows from the lack of objective tests and biomarkers, the cohort consisting of people with low educational attainment is probably less likely to receive an ME/CFS diagnosis, compared to people with medium educational attainment. Individuals from families in the low educational attainment group or individuals with low educational attainment does not necessarily have a reduced risk of the disease as a function of their education level. Our findings could rather be interpreted as that they are less able to ask for, or less likely to receive a G93.3 diagnosis in meeting with the specialized healthcare system.

Previous research has shown that clinical perceptions of patients with low SES affects the decisions made by clinicians in meeting with their patients (Arpey et al., 2017). Given the historical stigma surrounding ME/CFS, this bias might be uniquely present for this disease. Furthermore, the patients with low educational attainment are likely to have lower degrees of health literacy. Previous research has shown that low levels of health literacy is correlated with less knowledge about medical conditions (Gazmarian et al., 2003) and asking fewer questions during medical visits (Parikh et al., 1996). Previous research has also shown that people with higher SES are more likely to ask for and receive diagnoses compared to people with low SES (Arpey et al., 2017).

A large study gives further credence to this interpretation (Jason et al., 1999). Jason and colleagues screened a random sample of 18 675 individuals from 1995 to 1998 for ME/CFS symptomatology, and their analyses showed that almost 90% of people with middle to lower SES that were eligible for ME/CFS when retrospectively screening their symptoms, did not receive ME/CFS diagnosis by a physician. Therefore, it is highly possible that our findings reflect the underlying differences between high and low levels of education in meeting with the healthcare services, and not an epidemiological effect of education in and of itself. Furthermore, our analyses could reflect unequal access to healthcare services, as previous research has shown that higher SES is associated with better healthcare access (McMaughan & Smith, 2020; Arpey et al., 2017). The finding of an association between low educational attainment and reduced likelihood of ME/CFS diagnosis could therefore reflect unequal access to healthcare services. However, it is less likely that this effect is very prominent in our analyses, given the fact that Norway has a publicly funded healthcare system focused on egalitarianism. Nevertheless, it is a factor that could theoretically explain this finding.

This explanation becomes less likely as we do not find similar pattern in terms of the effect of household income. In fact, our findings related to household income reflect the large body of research on the relationship between wealth and the risk of chronic disease in general. For income, our analyses indicated that the same mechanisms that explain the relationship between income and reduced risk of chronic illness in general is prevalent for the risk of ME/CFS as well. When we compare to a randomly sampled population consisting of individuals that do not have hospital diagnosis, we find that low household income increases the risk of ME/CFS by 53%. When comparing with a population consisting of hospitalized controls, we find no statistically significant effect of comparatively low household income, but high household income reduces the risk of ME/CFS, relative to medium household income when comparing with the hospital diagnosed control population. These findings are in line with previous research on the relationship between SES and the risk of disease in general, and specifically research on the effect of wealth on the risk of chronic illness (Louwman et al., 2010).

It is important to note that we do not interpret our findings as evidence of substantial underdiagnosis of ME/CFS for people with low educational attainment in Norway as this explorative and rather descriptive study is not designed to draw this conclusion. Nevertheless, we argue that the rather unique clinical characteristics of ME/CFS, the lack of objective medical testing and the stigma surrounding the disease increases the likelihood that low educational attainment affects the clinical perceptions of these patients, which could reduce the probability of diagnosing these patients with ME/CFS. Taken together with the known effects of health literacy and healthcare access on the likelihood of hospital diagnosis, this could explain our findings. However, more research is needed to establish whether this is the case. As previously stated, our findings could also reflect an unknown epidemiological effect of low educational attainment that protects individuals from ME/CFS, or that the effect of age is not adequately captured in our models despite passing robustness checks for how we fit the effect of age in our models. We therefore encourage more research on the topic of healthcare utilization for the ME/CFS population as the implications for the healthcare system are important. If it is the case that individuals with low educational attainment are less likely to receive due to factors related to health literacy, clinical perceptions of low SES patients or healthcare access, then this needs to be addressed in order to avoid underdiagnosing individuals with ME/CFS.

Strengths and limitations

To the best of our knowledge, this is the first study about the relationship between SES and the risk of ME/CFS that utilize individual level health registry data. Individual level registry data allows for analyses with high internal validity due to objective measures for ME/CFS diagnosis, as opposed to small-N studies that rely on self-reporting, thus being prone to selection bias. The Norwegian registry data also gives us the opportunity to assess the relationship between SES and risk of ME/CFS diagnosis relative to both healthy and hospital diagnosed controls, which increase the validity of our findings. An additional strength related to our data is the fact that we can utilize a long "washout" period and model the effect of SES several years before ME/CFS diagnosis is confirmed. This way, the effects of ME/CFS on SES, i.e., reverse causality is almost completely removed. We measure SES at the family/household level, which also reduces problems of measuring income and education for young cohorts. Another strength of our study is that we operationalize SES as separate variables instead of a composite measure. Research shows that the mechanisms behind the correlation between high socioeconomic status and good health are diverse and complex. Operationalizing SES status should be done meticulously to capture the different mechanisms, which can be problematic.

As with all studies, ours also has weaknesses. One weakness is the lack of detailed occupational data that corresponds with the required washout period for our SES variables. It would be very interesting to include this as a SES variable as previous research has shown that occupational status, operationalized as profession or types of professions, is an important socioeconomic risk factor for chronic disease. However, the major weakness, or caveat, of this study, is the fact that we can only operationalize and define the ME/CFS population as G93.3. This means that all patients that are diagnosed by a primary care physician is not included in our analyses. This creates a form of selection bias that may lead us to underestimate the number of ME/CFS patients, as have been discussed both in this paper in previous research. We know that there is a risk of excluding immigrants and people with low societal status when studying the relationship between SES and chronic illness, as they are less likely to receive hospital diagnosis due to worse healthcare access and less resources in general (Dutton, 1986). This effect may even be more prevalent for ME/CFS, specifically due to the lack of biomarkers for the disease.

To the best of our knowledge, this is the first study that utilizes individual level registry data to test the relationship between socioeconomic status and the risk of being diagnosed with ME/CFS. The key takeaway from our analyses is that the effect of SES on the risk of ME/CFS diagnosis varies according to which population we use as a basis of comparison. We find a statistically significant relationship between SES operationalized as educational attainment and household income and the risk of ME/CFS diagnosis in all models. When comparing with a population consisting of the four selected hospital diagnosis, we find that high household income reduces the risk of ME/CFS diagnosis, while high educational attainment increases the risk. When comparing with a healthy sample of the random population, low educational attainment becomes statistically significant and strongly predicts a decreased risk of ME/CFS diagnosis. Furthermore, when comparing to a random population sample, we find that low household income increases the risk of ME/CFS diagnosis.

While our study confirms that SES affects the risk of ME/CFS, there is a need for more research on ME/CFS in general, and specifically on the mechanisms behind the relationship between SES and ME/CFS. Future research on SES and ME/CFS should seek to further unveil these mechanisms, for instance by including subjective social status (SSS) and more detailed occupational data as SES variables. More research on the healthcare utilization of ME/CFS patients is also needed, to further understand how SES variables such as educational attainment affects the diagnostic process pre- ME/CFS diagnosis.

Competing interests

We have no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethics approval and consent to participate

This study was part of a larger project on ME/CFS in Norway. Approval of individual level health registry data was obtained prior to data collection, by Norway' Regional Committees for Medical and Health Research Ethics (REK), reference: 2018/293. All methods were performed in accordance with the The World Medical Association' (WMA) Declaration of Helsinki, and it is not possible to identify individual patients from the aggregated data presented in this paper. The need for written informed consent was waived by the Norwegian Medical and Health Research Ethics (REK) ethics committee due to the retrospective nature of the study.

Consent for publication

Not applicable. Data was collected retrospectively from national registries and patient administrative systems, and it was not possible for the authors to directly identify individuals (name, address).

Availability of data and materials

Data used in this study are available from national registries pending approval and application. We are authorized to publish findings from these data as per Norway' Helseforskningsloven § 35. Linked health data from registries is classified as highly sensitive, and the authors of this study are therefore not authorized to share our data with the public due to the sensitive nature of individual health records.

Funding

Funding for this study was provided by the Research Council of Norway under the following grant number: 272674

Authors' contributions

The authors confirm contribution to the paper as follows: study conception and design: Geir Haakon Hilland and Kjartan S. Anthun; data collection: Kjartan S. Anthun; analysis and interpretation of results: Geir Haakon Hilland and Kjartan S. Anthun; draft manuscript preparation: Geir Haakon Hilland and Kjartan S. Anthun. All authors reviewed the results and approved the final version of the manuscript.

Acknowledgements

Not applicable

Authors' information (optional)

Not applicable

Addington, A.M., Gallo, J.J., Ford, D.E., Eaton, W.W., 2001. Epidemiology of unexplained fatigue and major depression in the community: The Baltimore ECA Follow-up, 1981–1994. Psychological Medicine 31, 1037–1044. doi:10.1017/s0033291701004214
Adler, N. E., Epel, E. S., Castellazzo, G., & Ickovics, J. R. (2000). Relationship of subjective andobjective social status with psychological functioning: Preliminary data in healthy Whitewomen.Health Psychology,19, 586–592.
Arpey, N.C., Gaglioti, A.H., Rosenbaum, M.E., 2017. How Socioeconomic Status Affects Patient Perceptions of Health Care: A Qualitative Study. Journal of Primary Care & Community Health 8, 169–175. doi:10.1177/2150131917697439
Black, R. E., Allen, L. H., Bhutta, Z. A., Caulfield, L. E., De Onis, M., Ezzati, M., ... & Maternal and Child Undernutrition Study Group. (2008). Maternal and child undernutrition: global and regional exposures and health consequences. The lancet, 371(9608), 243-260.
Brurberg KG, Fønhus MS, Larun L, Flottorp S, Malterud K. Case definitions for chronic fatigue syndrome/myalgic encephalomyelitis (ME/CFS): a systematic review. BMJ Open. 2014 Feb 7;4(2): e003973. doi: 10.1136/bmjopen-2013-003973. PMID: 24508851; PMCID: PMC3918975.
Callahan, C. L., & Eyberg, S. M. (2010). Relations between parenting behavior and SES in a clinicalsample: Validity of SES measures, Child & Family Behavior Therapy,32(2), 125–138.
Choi, H., Marks, N.F., 2011. Socioeconomic Status, Marital Status Continuity and Change, Marital Conflict, and Mortality. Journal of Aging and Health 23, 714–742. doi:10.1177/0898264310393339
Corfield, E. C., Martin, N. G., & Nyholt, D. R. (2016). Co-occurrence and symptomatology of fatigue and depression. Comprehensive psychiatry, 71, 1-10.
Diemer, M. A., Mistry, R. S., Wadsworth, M. E., López, I., & Reimers, F. (2013). Best practices in conceptualizing and measuring social class in psychological research. Analyses of Social Issues and Public Policy (ASAP), 13(1), 77–113. https://doi.org/10.1111/asap.12001
Duncan, G. J., & Magnuson, K. A. (2003). Off with Hollingshead: Socioeconomic resources, parenting and child development. In M. H. Bornstein & R. H. Bradley (Eds.), Socioeconomic status, parenting, and child development(pp. 83–106). Mahwah, NJ: Erlbaum
Dutton, DB. Social class, health and illness. In: Aiken, L.; Mechanic, D., editors. Applications of Social Science to Clinical Medicine, and Health Policy. New Brunswick, N. J: Rutgers University Press; 1986.
Fluge, Ø. R, Tronstad, K.J., Mella, O., 2021. Pathomechanisms and possible interventions in myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). Journal of Clinical Investigation 131. DOI: https://doi.org/10.1172/jci150377
Galobardes, B., Lynch, J. W., & Davey Smith, G. (2004). Childhood socioeconomic circumstances and cause-specific mortality in adulthood: systematic review and interpretation. Epidemiologic reviews, 26(1), 7-21.
Gauthier, J., Wu, Q.V., Gooley, T.A., 2020. Cubic splines to model relationships between continuous variables and outcomes: a guide for clinicians. Bone Marrow Transplantation 55, 675–680.. https://doi.org/10.1038/s41409-019-0679-x
Gazmararian JA, Williams MV, Peel J, Baker DW. Health literacy and knowledge of chronic disease. Patient Educ Couns. 2003;51(3):267-275.
Geraghty KJ, Adeniji C. The Importance of Accurate Diagnosis of ME/CFS in Children and Adolescents: A Commentary. Front Pediatr. 2019 Jan 21; 6:435. doi: 10.3389/fped.2018.00435. PMID: 30719431; PMCID: PMC6348258.
Gunn WJ, Connell DB, Randall B. Epidemiology of chronic fatigue syndrome: The Centers for Disease Control study. In: Rock BR, Whelan J, editors. Chronic Fatigue Syndrome. New York: John Wiley & Sons; 1993. pp. 83–101.
Hilland G.H, Anthun S.K: Estimert insidens av kronisk utmattelsessyndrom/myalgisk encefalopati i Norge mellom 2016-2018. Nordisk Tidsskrift for Helseforskning 2.22.
Jann, Ben (2014). Plotting regression coefficients and other estimates. The Stata Journal 14(4): 708‐737.
Jason, L.A., Richman, J.A., Rademaker, A.W., Jordan, K.M., Plioplys, A.V., Taylor, R.R., Mccready, W., Huang, C.-F., Plioplys, S., 1999. A Community-Based Study of Chronic Fatigue Syndrome. Archives of Internal Medicine 159, 2129. doi:10.1001/archinte.159.18.2129
Jason LA, Taylor RR, Kennedy CL, Jordan K, Huang CF, Torres-Harding S, Song S, Johnson D. A factor analysis of chronic fatigue symptoms in a community-based sample. Social Psychiatry & Psychiatric Epidemiology. 2002a;37(4):183–189. [PubMed] [Reference list]
Karfakis, N. The biopolitics of CFS/ME, Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences,Volume 70, 2018, Pages 20-28, ISSN 1369-8486, https://doi.org/10.1016/j.shpsc.2018.05.009.
Katz, B.Z., Shiraishi, Y., Mears, C.J., Binns, H.J., Taylor, R., 2009. Chronic Fatigue Syndrome After Infectious Mononucleosis in Adolescents. Pediatrics 124, 189–193. doi:10.1542/peds.2008-1879
Kocalevent, R.D., Hinz, A., Brähler, E., Klapp, B.F., 2011. Determinants of fatigue and stress. BMC Research Notes 4, 238. doi:10.1186/1756-0500-4-238
Larun, L., Brurberg, K.G., Odgaard-Jensen, J., Price, J.R., 2017. Exercise therapy for chronic fatigue syndrome. Cochrane Database of Systematic Reviews. doi: 10.1002/14651858.cd003200.pub7
Leone, S.S., 2010. A disabling combination: fatigue and depression. The British Journal of Psychiatry 197, 86–87. doi:10.1192/bjp.bp.109.076604
Mehmetoglu, Mehmet & Jakobsen, Tor. (2017). Applied Statistics Using Stata - A Guide for the Social Sciences.
Louwman, W.J., Aarts, M.J., Houterman, S., Van Lenthe, F.J., Coebergh, J.W.W., Janssen-Heijnen, M.L.G., 2010. A 50% higher prevalence of life-shortening chronic conditions among cancer patients with low socioeconomic status. British Journal of Cancer 103, 1742–1748. https://doi.org/10.1038/sj.bjc.6605949
Lloyd AR, Hickie I, Boughton CR, Spencer O, Wakefield D. Prevalence of chronic fatigue syndrome in an Australian population. Med J Aust. 1990; 153:522–528.
López, F., et al. (2020). "Systematic Review of the Epidemiological Burden of Myalgic Encephalomyelitis/Chronic Fatigue Syndrome Across Europe: Current Evidence and EUROMENE Research Recommendations for Epidemiology." Journal of Clinical Medicine 9(5): 1557. DOI: https://doi.org/10.3390/jcm9051557
Hosmer, D. W., Jr., S. A. Lemeshow, and R. X. Sturdivant. 2013. Applied Logistic Regression. 3rd ed. Hoboken,NJ: Wiley
Huibers, H., Stephanie S. Leone, Ludovic G.P.M. van Amelsvoort, IJmert Kant, J. André Knottnerus. Associations of fatigue and depression among fatigued employees over time: A 4-year follow-up study, Journal of Psychosomatic Research, Volume 63, Issue 2, 2007, Pages 137-142. https://doi.org/10.1016/j.jpsychores.2007.02.014.
Mair, F.S., Jani, B.D., 2020. Emerging trends and future research on the role of socioeconomic status in chronic illness and multimorbidity. The Lancet Public Health 5, e128–e129. doi:10.1016/s2468-2667(20)30001-3
McMaughan DJ, Oloruntoba O, Smith ML. Socioeconomic Status and Access to Healthcare: Interrelated Drivers for Healthy Aging. Front Public Health. 2020 Jun 18; 8:231. doi: 10.3389/fpubh.2020.00231. PMID: 32626678; PMCID: PMC7314918.
Muller, A.E., Tveito, K., Bakken, I.J., Flottorp, S.A., Mjaaland, S., Larun, L., 2020. Potential causal factors of ME/CFS: a concise and systematic scoping review of factors researched. Journal of Translational Medicine 18. doi:10.1186/s12967-020-02665-6
Natelson, B.H., 2019. Myalgic Encephalomyelitis/Chronic Fatigue Syndrome and Fibromyalgia: Definitions, Similarities, and Differences. Clinical Therapeutics 41, 612–618. doi: 10.1016/j.clinthera.2018.12.016
Nacul, L., Lacerda, E.M., Kingdon, C.C., Curran, H., Bowman, E.W., 2019. How have selection bias and disease misclassification undermined the validity of myalgic encephalomyelitis/chronic fatigue syndrome studies? Journal of Health Psychology 24, 1765–1769. doi:10.1177/1359105317695803
Parikh NS, Parker RM, Nurss JR, Baker DW, Williams MV. Shame and health literacy: the unspoken connection. Patient Educ Couns. 1996;27:33-39.
Psaki, S.R., Seidman, J.C., Miller, M., Gottlieb, M., Bhutta, Z.A., Ahmed, T., Ahmed, A.S., Bessong, P., John, S.M., Kang, G., Kosek, M., Lima, A., Shrestha, P., Svensen, E., Checkley, W., 2014. Measuring socioeconomic status in multicountry studies: results from the eight-country MAL-ED study. Population Health Metrics 12, 8. doi:10.1186/1478-7954-12-8
Pollack, C. E., Chideya, S., Cubbin, C., Williams, B., Dekker, M., & Braveman, P. (2007). Should health studies measure wealth? A systematic review. American journal of preventive medicine, 33(3), 250-264.
Rangel, L., Garralda, M.E., Levin, M., Roberts, H., 2000. The course of severe chronic fatigue syndrome in childhood. Journal of the Royal Society of Medicine 93, 129–134. doi:10.1177/014107680009300306
Singh-Manoux A, Marmot MG, Adler NE (2005) Does subjective social status predict health and change in health status better than objective status? Psychosomatic Medicine 67(6): 855–861.
Skapinakis P, Lewis G, Mavreas V: Unexplained fatigue syndromes in amultinational primary care sample: specificity of definition andprevalence and distinctiveness from depression and generalized anxiety.Am J Psychiatry2003,160:785-787
Song S, Jason LA, Taylor RR. The relationship between ethnicity and fatigue in a communitybased sample. Journal of Gender, Culture, and Health 1999;4:255–268.
StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.).
Twisk FN, Maes M. A review on cognitive behavorial therapy (CBT) and graded exercise therapy (GET) in myalgic encephalomyelitis (ME) / chronic fatigue syndrome (CFS): CBT/GET is not only ineffective and not evidence-based, but also potentially harmful for many patients with ME/CFS. Neuro Endocrinol Lett. 2009;30(3):284-99. PMID: 19855350.
Viner, R., Hotopf, M., 2004. Childhood predictors of self-reported chronic fatigue syndrome/myalgic encephalomyelitis in adults: national birth cohort study. BMJ 329, 941. doi:10.1136/bmj.38258.507928.55
Vink M, Vink-Niese F. Graded exercise therapy does not restore the ability to work in ME/CFS - Rethinking of a Cochrane review. Work. 2020;66(2):283-308. doi: 10.3233/WOR-203174. PMID: 32568149.
Williams, W. R. (2009). Struggling with poverty: Implications for theory and policy of increasing research on social class-based stigma. Analyses of Social Issues and Public Policy,9(1), 37–56.

No competing interests reported.

SupplementaryMaterial.docx

Socioeconomic determinants of myalgic encephalomyelitis/chronic fatigue syndrome in Norway: a registry study

Status:

Version 1

Abstract

Figures

Introduction

Background

The relationship between socioeconomic status and chronic disease

Previous research on SES and ME/CFS

Data And Methods

Study design and population

Household income

Educational Attainment

Control variables

Statistical analyses

Results

Robustness tests

Discussion

Strengths and limitations

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1