Socioeconomic inequalities in 29 childhood diseases: evidence from a 1,500,000 children population retrospective study

We carried out a population-based study on a retrospective review of all children under 15 resident in Catalonia during the period 2014–2017 (1,449,816 children, 48.52% girls).

Two different databases were used:

The Central Registry of Insured Persons (RCA, Catalan acronym), to obtain the reference population for the 2014–2017 period. This registry also collects and updates data on the annual income level, employment status, and Social Security benefits of each individual, which are routinely used to calculate citizens’ pharmaceutical copayment levels.

The Registry of the Minimum Basic Dataset (CMBD, Catalan acronym) is an administrative register containing detailed information on sociodemographic characteristics and medical diagnoses (coded using the International Classification of Diseases, 9th Edition). The CMBD encodes all contacts with the public healthcare system at an individual level: primary care, hospital care, emergency, mental health, and long-term care services. All the diagnostic codes registered in the CMBD during the period 2014–2017 were gathered to create the main dataset which was then linked to the RCA through a common personal healthcare ID.

Both registries are considered of high quality as 1) they include the entire population of Catalonia since all citizens are granted universal healthcare coverage (99.6% of all residents are registered); 2) they have a validation system; and 3) they contain historical diagnosis data. We excluded 307 children (0.02% of the entire child population) due to a misclassification of the SEP variable.

The dependent variables were 29 diseases or risk factors representative of the health status of the child population of Catalonia. This wide selection includes both diseases that are strongly influenced by social determinants and others whose influence is not clearly established [6, 27]. This selection also represents diseases of diverse etiology, affecting different organic systems, and of differing severity. Each disease variable was constructed selecting diagnostic codes based on clinical criteria (see Additional file 1). The inclusion criteria at an individual level was having at least one medical diagnosis, being registered on the CMBD, and included in at least one of the defined diseases. Duplicate diagnoses were not taken into account so that each of the disease variables was treated as a dichotomous variable.

The independent variables were SEP, age, and sex. SEP and age were defined at the beginning of the study period. SEP was proxied by the pharmaceutical copayment level of the child’s parents or guardians, obtained from the RCA registry. Pharmaceutical copayment uses income level and employment status information (specially to detect individuals without any income from work) to define four SEP categories as follows: “Very low” (no member of the household employed or receiving welfare support), “Low” (<€18,000/year), “Medium” (€18,000–€100,000/year), and “High” (>€100,000/year). This classification has been used in previous studies [20, 28]. Age was categorised into five groups (0–2 years, 3–5 years, 6–8 years, 9–11 years, and 12–14 years) and sex (boys, girls).

The measure of RII is a relative measure of inequality that involves the following steps: 1) determination of the disease frequency of each SEP category and 2) ranking SEP levels, assigning a hierarchical value from 0 (lowest SEP) to 1 (highest SEP) according to its proportional size in the population by linear regression. The population of each SEP level is assigned a Ridit score based on the mid-point of the cumulative distribution of the population of participants in the given category, and finally 3) this weighted SEP is included as exposure and disease as a dichotomous dependent variable in a generalised linear model (log-binomial regression) with a logarithmic link function. The highest SEP category was used as a reference. The RII reflects the risk ratio of an outcome between those on the social hierarchy who are more advantaged and those who are more disadvantaged [29, 30].

The SII is an absolute measure of inequality that represents the absolute difference in estimated values of a health outcome between the most advantaged and the most disadvantaged. The calculation of SII follows the same steps as RII: 1) and 2) and then, 3) weighted SEP is also included as exposure and frequency of disease as a dependent variable in a generalised linear model adjusted by age with an identity link function. The difference in the predicted values for the extreme SEP categories generates the SII value. If there is no inequality, SII takes the value zero. Greater absolute values indicate higher levels of inequality [29, 30].

Finally, to study intersectionality between sex and SEP across diseases, sex differences were assessed by inserting a two-way interaction term between the SEP Ridit score calculated previously and gender for the RII and SII models. A positive and significant coefficient for the two-way interaction term would indicate a larger increase in RII or SII in girls compared to boys [31].

We calculated population attributable fractions in absolute terms (PAFa) and relative terms (PAF%). To achieve this, medium and high SEP categories were grouped and defined as the reference group and low and very low SEP as the exposure group; the log-binomial model mentioned previously were performed again. PAFa are the estimated cases that would be prevented if low and very low SEP had the same health status as higher SEP levels while other risk factors remain unchanged. PAF% is the associated fraction and its PAF% = pd.[(PRa-1)/PRa], where PRa is the prevalence ratio, pd. is the proportion of those with the specific disease and PRa-1 represents the risk over the reference point. For incidence proportion substitute PRa for IRa. An example of interpretation is, for instance: TB in girls has a PAF% = 51.95% of the total TB events in girls which could be potentially avoided if all the population had the same TB incidence as medium-high SEP, which is translated in PAFa = 171 TB events. The PAFa and PAF% results are presented with a 95%CI and the associated p-values [25, 29, 32].

All analyses were performed separately for girls and boys. All methods were carried out in accordance with relevant guidelines and regulations. We used the STATA IC/15.1 software modules RIIGEN for RII and SII and Punaf for PAF, to perform all the analyses [33, 34].

Prevalences, incidence proportions by sex and SEP, and PR and IR between sexes for each disease are shown in Table 2. The most frequent diseases or risk factors in both sexes were injuries (34.51% in boys and 29.25% in girls), dermatitis (21.22% in boys and 23.30% in girls), and acute bronchitis (18.32% in boys and 15.23% in girls). Congenital anomalies, being overweight, influenza, chronic bronchitis, obesity, and asthma had proportions between 5 and 10% in both sexes.

Of all the 29 pathologies studied, 25 were more frequent in boys. Autism spectrum disorder (ASD) (PR = 4.66, 95% CI: 4.39, 4.94), attention deficit hyperactivity disorder (ADHD) (PR = 2.70, 95% CI: 2.63, 2.76), meningitis (IR = 1.57, 95% CI:1.36, 1.80), and asthma (PR = 1.47, 95% CI:1.45, 1.49) were significantly more frequent in boys than in girls. By contrast, mood disorders (PR = 0.87, 95% CI: 0.83, 0.91) and dermatitis (PR = 0.94, 95% CI: 0.93, 0.94) were significantly more frequent in girls than in boys.

Measures of SEP inequality are presented in Table 3. Overall, 25 of the 29 diseases analysed showed a significant inverse SEP gradient: the lower the SEP, the higher the frequency. Only food allergy showed a significant direct SEP gradient (RII = 0.86, 95% CI: 0.82, 0.90 in boys and RII = 0.87, 95% CI: 0.83, 0.92). The other diseases studied (cancer of the brain and nervous system, leukaemia, cystic fibrosis for both sexes, and meningitis and acute cerebrovascular disease for girls) did not show statistical significance.

The diseases with a more pronounced SEP gradient (RII ≥ 2.0) in both sexes were TB, obesity, adjustment and anxiety disorders, essential hypertension, poisoning, short gestation, low birth weight and foetal growth retardation, and intrauterine hypoxia and birth asphyxia and trauma. Following these closely, but with slight differences depending on sex, were mood disorders, meningitis and paralysis in boys, and mood disorders and overweight in girls (see Fig. 1).

Regarding the differences between sex in inequalities, relative inequalities were significantly higher in girls than in boys for respiratory allergy, asthma, dermatitis, being overweight, and obesity (p < 0.05). In contrast, boys showed significantly higher relative inequalities compared to girls only in congenital anomalies (p < 0.05). Absolute inequalities were significantly greater in boys compared to girls for acute cerebrovascular disease and cancer of the brain and nervous system (p < 0.05).

The population SEP impact measures are presented in Table 4. The diseases that could potentially be prevented more from equating the health status of the childhood population to the medium-high SEP were, in relative terms, TB (PAF% = 56.26, 95% CI: 43.03, 66.41 in boys and PAF% = 51.95, 95% CI: 38.51, 62.45 in girls), intrauterine hypoxia and birth asphyxia and trauma (PAF% = 30.04, 95% CI: 25.40, 34.39 in boys and PAF% = 32.10, 95% CI: 26.54, 37.24 in girls), obesity (PAF% = 27.14, 95% CI: 25.86, 28.39 in boys and PAF% = 30.31, 95% CI: 28.77, 31.80 in girls), and short gestation low birth weight and foetal growth retardation (PAF% = 29.83, 95% CI: 26.78, 32.75 in boys and PAF% = 28.21, 95% CI: 24.98, 31.29 in girls). In absolute terms, the diseases or risk factors with the highest potential of preventable individuals affected due to the impact of SEP were dermatitis (PAFa = 15,345, 95% CI: 14457, 16,227 in boys and PAFa = 19,027, 95% CI: 18128, 19,921 in girls), injuries (PAFa = 19,432, 95% CI: 18303, 20,555 in boys and PAFa = 14,758, 95% CI: 13714, 15,796 in girls), acute bronquitis (PAFa = 14,082, 95% CI: 13174, 14,938 in boys and PAFa = 11,905, 95% CI: 11091, 12,713 in girls), and being overweight (PAFa = 10,905, 95% CI: 10280, 11,524 in boys and PAFa = 10,872, 95% CI: 10299, 11,438 in girls).

In terms of population impact, 7.80% (95% CI: 7.53, 8.06) of the overall disease events analysed in boys and 8.88% (95% CI: 8.58, 9.17) in girls would be avoided if all children had the same disease rates as the medium-high SEP group. That is translated into 80,188 disease events (95% CI: 77456, 82,912) in boys and 74,921 disease events (95% CI: 72429, 77,405) in girls that would be potentially preventable (data for the total burden of diseases studied; for the PAF results of each disease, see Table 4). By contrast, food allergy showed a significant negative attributable population impact (PAF% = − 5.42, 95% CI: − 7.07, − 3.8 in boys and PAF% = − 4.46, 95% CI: − 6.22, − 2.73 in girls).

Figure 2 for boys and Fig. 3 for girls show three different inequality measures for each disease and sex.

This study confirms that socioeconomic inequalities are present in a large number of diseases of different etiologies and levels of seriousness (from malignant neoplasms to dermatitis), in the entire population of children in Catalonia. The systematic pattern of socioeconomic inequality observed highlights that, beyond the nature and idiosyncrasy of each disease, 1) the entire child population is affected by this social phenomenon, 2) SEP plays a common role in their development, and 3) gender differences in inequalities exist in different diseases and risk factors. This is consistent with the findings of different studies where social class and child health status were assessed [6, 7, 9, 27, 35].

To calculate health inequalities and to enrich later interpretations and findings, we calculated four summary measures of health inequalities between SEP levels in each disease and the burden of the diseases attributable to SEP. The purpose was to capture both absolute and relative inequality measures accounting for all SEP groups and their size in the studied population. On the other hand, absolute measures provided the total number of disease events affected by SEP and relative measures tend to quantify the association of SEP and the outcome. Summary measures, based on regression, take into account all SEP groups weighted according to their real size rather than an equal distribution. Therefore, they provided a gradient that includes the entire population, being less biased than using pairwise measures in which only very low and high SEP would be taken into account. This last approach could be more affected by undereported health outcomes, mainly in the high SEP group, due to their use of private healthcare services [26, 29].

SEP, as a social determinant of health, influences the distribution of diseases among the population, as shown in this and previous studies [6, 27, 36]. Although the pathways by which SEP impacts on health are mostly indirect, SEP influences factors such as nutrition, housing conditions, and education, hence the degree of vulnerability and social exclusion of children [1].

Social vulnerabilities in childhood is a broad concept that includes all the negative experiences that could affect children’s development and, therefore, their health. However, income level could exacerbates or buffer the effect of the social vulnerabilities in lifestyle and health outcomes [1, 7]. All these social experiences in childhood impact on a crucial period of development through different complex and inter-related pathways such as biological embedding, material deprivation, psychosocial or structural mechanisms. All these pathways, highly dependent on the family’s relative SEP, act simultaneously creating social gradients in present and future health outcomes [1, 37].

The results observed in this study indicate that the gradients between SEP and health present different intensities depending on the disease. This sheds more light on the identification of those diseases with a high social factor in their construction and distribution. In concordance with other studies, the diseases that show the greatest inequalities are TB, obesity, hypertension, adverse birth outcomes, and mental health disorders [27, 38‐41]. TB is by far the disease with most inequalities in Catalonia and elsewhere [38]. TB is related to poverty (children from very low SEP levels are at least up to 8 times more likely to have it than their most advantaged peers) and it is particularly exacerbated in immigrant populations [42, 43]. Screening programs and an improvement in working and housing conditions stand out as the greatest contributions to eradicating TB [38]. As in other studies which test inequalities in different health outcomes, obesity (the most acute form of overweight) and hypertension, both related to lifestyles, show a high association with SEP [27]. There is a lot of evidence which shows that diseases related to lifestyles are greatly impacted by social determinants such as SEP or a parent’s educational level, or nutrition and physical activity, among others [15, 16, 27, 39]. Therefore, within a global upward trend, the number of children who suffer from this inequality is also increasing [17].

Special attention should be paid to inequalities in adverse birth outcomes. As our results reveal, they are broadly present [40] or even increase in countries similar to Catalonia [44]. These outcomes are related to pregnancy and delivery, which are periods of intense surveillance by the healthcare system. This situation emphasizes that other factors such as employment status or SEP level (both individual and neighbourhood) are key in explaining these inequities beyond the responsability of the health system [40, 44]. Regarding mental health disorders, our findings show that SEP inequalities stand out in concordance with other studies [6, 27, 41]. Other factors as mental health or the employment status of parents (mainly of mothers) also play an important role [28] being particularly incisive in contexts of economic crisis [18]. Significant attention must be given to those diseases that would potentially be most reduced by improving children’s SEP. These diseases are not those showing the highest inequalities but those that account for the largest number of preventable disease events (mainly due to their high prevalence in childhood population), namely: dermatitis, injuries, acute bronchitis, obesity, and being overweight.

Previous studies on children have not taken into account gender as an axis of inequality, assuming that the social role of gender does not influence health inequalities in childhood. By contrast, our study has shown differences in many of the measures calculated. In general terms, the PR or IR of boys compared to girls is higher, to the boys’ detriment. This finding is consistent with the slightly increased vulnerability in boys (focused in the early ages) and the physiological differences between the sexes [21, 45]. However, in terms of inequalities, this study has found that girls suffer from significant differences in inequalities in more pathologies in comparison to boys. This finding points to an intersectionality between sex and SEP where the two axes of inequality jointly create roles of multiplicative discrimination and subordination [46].

This study brings attention to a possible gender bias in paediatrics healthcare. Gender bias in paediatrics involves a number of different actors: the infant as a patient, the parents who present the child’s symptoms, and the medical professional who interprets the patient’s narratives and behaviours. Several studies have revealed that girls versus boys receive less effort to obtain an advanced diagnostic, a therapeutic intervention, or an adequate pain assessment, have different hospitalisation rates for the same conditions and experience more preconceptions about pathology patterns [22, 23, 47, 48]. SEP, sex, gender bias, and their intersections simultaneously influence the process of disease construction via diverse pathways and in differing intensities [21, 24]. Gender bias is well known in adulthood but little research has been done on the phenomenon in childhood, as if it did not exist [31, 49]. Future research should explore further on gender-related health inequalities in childhood.

With regard to food allergy, other studies have also highlighted its direct SEP gradient [27]. An increasing global trend awareness and the widespread screening of food allergy supports our finding of a direct SEP gradient [50]. High-SEP parents tend to be more concerned with identifying food allergies in their children and seek out more tests than the lowest-SEP parents, contributing to overdiagnosis in socioeconomically advantaged children [51]. Other studies also add that part of these socioeconomic differences may be caused by underdiagnosis in disadvantaged groups [27], eventhough in our study population low SEP children do a higher number of visits to the healthcare system than their advantatged peers [20].

Our results confirm that, despite the important role of universal healthcare coverage in guaranteeing free and equal access to the health system, it does not assure achieving equity in health status. Thus, similar findings have encouraged governments to include social determinants as a main cause of health inequalities [9].

This study included a broad representation of the childhood pathology; 29 diseases or risk factors of different etiologies, magnitudes, seriousness, and affecting diverse systems of the human body. It provides high-value evidence of the broad impact of social determinants on health and a useful variety of inequality measures, absolute and relative, with the intention of influencing population and political decisions.

Calculating the same inequality measures in a broad set of diseases on the same population increases the robustness of comparisons between diseases as an homogeneous methodology has been used, and as all the population is affected by the same common external factors.

This population-based study included the total population under 15 years old in Catalonia, nearly 1,500,000 individuals over a 4-year period providing a large capacity to detect more individuals with low-prevalence diseases.

The study used individual-level data, which allowed economic and employment information for one of a child’s guardians to be linked with the child’s health status. However, this data does not provide complete information about household economic status or a more detailed segmentation of the SEP variable but it has been used successfully in different studies [20, 28].

The SEP indicator used is based on the parent’s economic and employment situation, which is available from the Catalan Department of Health databases. Other commonly used indicators such as parental education or occupation were not available. Besides, income has proven to have a strong association with health and to influence a wide range of material circumstances with direct implications for health [52].

Medical diagnoses are based on contacts with the public healthcare system. Even where universal healthcare access is granted, private and public health services co-exist, especially for high SEP families. It is estimated that 30% of all children under the age of 15 use both private and public healthcare (88% from medium and high SEP) [53]. This can lead to an underreporting of some diseases, especially those among the most frequent, less severe and acute profile, such as respiratory diseases or dermatitis.