WHAT IS ALREADY KNOWN ON THIS TOPIC

Low family socioeconomic position is a well-established determinant of poor health in youth, but research evidence on social inequalities in medication use is inconsistent and mostly lacks information on important confounders.

WHAT THIS STUDY ADDS

Data on nearly 1.5 million Finnish individuals showed that the use of antibiotic, painkiller and allergy and asthma medications was more common among youth from high-income families whereas psychotropic use was less common. These income differentials were likely due to unobserved familial confounding.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

Socioeconomic resources of the family, including parental education, income and wealth, are well-established determinants of offspring health.1 Even in relatively egalitarian Nordic welfare states adolescents and young adults from disadvantaged socioeconomic backgrounds have worse health than their peers across multiple domains such as mortality, self-rated health, chronic and acute somatic illnesses, and mental health.1 2 Less in known about the social patterning of medication use in youth and current evidence is mixed.

While poor health is more common in families with low socioeconomic position, the lack of socioeconomic resources may hinder treatment seeking and access and result in less medication use.3 In line with the well-established socioeconomic gradient health, many studies show low family socioeconomic position to be related to more medication use.4–13 However, some studies have found youth from low socioeconomic backgrounds to have less medication use,9 12 14–19 while a number of studies find no association between socioeconomic background and medication use.7–9 13–15 20–26

The methodological diversity across studies hampers meta-analytical assessment of the prior evidence, but the differing results seem not to be strongly driven by differences in medication type, age range, socioeconomic measures, type of data or country context. However, studies that address the need for treatment, that is, the underlying health status, yield somewhat more consistent results. These studies, with some exceptions,4 14 15 show that, at the same level of need, antibiotic use was less common among youth from low socioeconomic backgrounds,16–18 whereas the use of painkillers,5–7 25 26 asthma and allergy medications,23 24 27 and psychotropics5 7 8 15 20 21 was either more common among youth from low socioeconomic backgrounds or unrelated to socioeconomic background.

Currently little is known on the causal processes behind these associations. Important confounders have mostly been lacking from previous analyses. For example, poor parental health may impact both family income and offspring health and related medication use. Poor health is also a strong predictor of not filling prescriptions3; single parenthood is related both to low family income28 and lower likelihood of filling prescriptions3; immigrants have lower incomes and use less health services and medications.29 Only a handful of studies have addressed these factors11 15–17 21 with parental health only adjusted for by one prior study.11 Furthermore, a prior sibling comparison study found that unobserved familial confounding may explain the association between childhood socioeconomic background and health in adolescence and young adulthood.30 Unobserved familial factors, such as genetic risks, parenting styles, as well as parental values and attitudes regarding health and health care may also confound the association between socioeconomic background and medication use.

This study uses data on 25 full birth cohorts from Finland to assess the use of the most common prescription medications used in youth. The aim is to describe the differences in medication use in ages 16–20 according to childhood family income and assess whether these associations are due to other observed factors, often neglected in prior studies, such as single parenthood, immigration status and parental poor health. As a first study to date, we also use between-sibling comparison as a quasi-experimental design to assess whether socioeconomic differences in medication use are further explained by other unobserved familial factors shared by siblings. We use income as an indicator for childhood family socioeconomic position because, unlike more stable family characteristics such as parental education, it varies between siblings, so that children are exposed to differing income levels in childhood, and thus enables a quasi-experimental sibling comparison analysis. Furthermore, income is more directly linked to the economic resources available for purchasing private healthcare and medication.

The study is set in Finland, a Nordic welfare state with a universal, public health insurance that during the study period covered public healthcare and 40%–65% of the most common prescription medication costs. After exceeding the annual limit for out-of-pocket payments of around 600 Euros, the patient only paid a fixed non-reimbursable sum of around 1–3 Euros per filled prescription.31 Use of complimentary, voluntary private health insurance is, however, common, particularly among families with children, and these fully cover private healthcare and medication expenses. In 2019, around 45% of Finnish children aged 0–17 years were covered by private health insurance and this was strongly related to higher household income.32 Finland has a larger share of households not filling prescriptions due to cost-related barriers than the other Nordic or European countries,3 possibly because of relatively high medication copayments. Not filling a prescription was strongly related to low household income.3

MethodsStudy population

We used administrative register data on the full population of Finland. Socioeconomic and demographic information was acquired from Statistics Finland, and information on purchases of prescription medication and rights to special reimbursement of medications from the Social Insurance Institution. These data were linked using pseudonymised personal identification numbers assigned to all permanent residents of Finland with the permits granted by the Ethics Committees of Statistics Finland (TK-53-1490-18) and Findata (THL/2180/14.02.00/2020).

We assessed full birth cohorts 1979–2003 (n=1 798 582), only including individuals with information on both biological parents to allow for identification of biological full siblings (n=1 581 028, figure 1). Only children who had lived in private households in two-parent or single-parent families throughout ages 11–15 (n=1 501 860) were included to be able to determine childhood household income. In addition, children who emigrated from Finland or died before their 16th birthday (n=11 194) were excluded, thus leaving a final total population sample of 1 490 666 individuals. In the sibling comparison, we further excluded all children without biological full siblings in the studied cohorts (n=206 960), thus leaving a final sibling population sample of 1 283 706.

Figure 1

Flowchart on the formation of the samples for the total population and sibling population analyses.

These individuals were followed from their 16th birthday to the 21st birthday between years 1995 and 2019. As exact dates of birth were unavailable, we set them on the 15th of the month of birth. The follow-up was censored at the first purchase of each of the most common prescription medications, 21st birthday, end of study period (31 December 2019), date of emigration, the last day a person was registered as a permanent resident in Finland, or death, whichever came first.

Outcomes

We assessed the most commonly purchased prescription medications in our cohorts between ages 16 and 20. The medications were classified according to third level of the Anatomical Therapeutic Chemical Classification System.31 For this study, we included medications purchased by at least 10% of the cohort during the follow-up, and psychotropics purchased by 9% of the cohort. Psychotropics were included despite their slightly lower prevalence, because of their quickly increasing use among youth in Finland and elsewhere.33 34 We further classified these medications into four wider medication types based on their main indications (table 1).

Table 1

The most common medications purchased by the study population between ages 16 and 20, their prevalence in follow-up (%) and classification into wider medication types, which were used as outcomes in the study

During the study period, all permanent residents of Finland were entitled to a reimbursement of prescription medication expenses covering 40%–50% of costs for all other studied medication types except asthma medications (R03) for which reimbursement covered 65%.31 Over-the-counter purchases were not reimbursed and thus not registered.

Exposure

Household income was used as an indicator for childhood socioeconomic resources. It was measured as the yearly gross taxable household income in 2019 Euros divided by the square root of household size and averaged across ages 11–15 (see online supplemental file for details).

Covariates

We adjusted for important demographic confounders. Age in days since 16th birthday was used as time scale in regression analyses. Sex was registered at birth as female/male. Calendar year was a time-varying categorical covariate that captures any secular trends in medication use, for example, due to changing prescribing practices.

We further measured various factors that may confound the relationship between childhood income and medication use in youth, many of which previous research have been unable to address:

Geographical region (NUTS3, 20 categories35) measured as mode across ages 11–15 captured potential regional differences in prescription practices.

First language registered at birth or immigration as Finnish/Swedish/other captured immigration-related language and cultural barriers to healthcare.

Parental chronic illnesses were measured as 10 separate indicators (yes/no): heart disease, diabetes, lung disease, cancer, rheumatism, neurological illness, psychosis, Crohn’s disease, kidney disease, and other chronic illnesses. These were based on whether either parent had a registered right for special reimbursement for medicine purchases granted for a given chronic doctor-diagnosed illness31 between offspring ages 11–15. Individuals could be exposed to several parental chronic illnesses simultaneously.

Proportion of years lived in a single-parent household (%, continuous) across ages 11–15 captured features of the family environment that may impact medication use, such as parental time investment.

Birth order was categorised as first/second/third/fourth and subsequent children and accounted for possible confounding by birth-order-related health differences and income changes.

Analyses

For the descriptive analysis, we used unadjusted Poisson regression models to estimate the incidence of having a medication purchase between ages 16 and 20 by medication type and childhood household income vigintile, that is, 5% income group, as rates per 100 person-years. We plotted these rates against the median income of each income vigintile. The medians were calculated by averaging across the yearly inflation-adjusted median incomes of each vigintile over the entire study period.

To model the association between childhood income and the likelihood of purchasing the four most common medication types, we used Cox proportional hazards models (proportional hazards assumption visually inspected in online supplemental figure 1). Due to the non-linearity of the association, we used a logarithmic transformation of income with a base of 1.1. The estimated HR1.1 thus reflect changes in outcomes per 10 percentage increase in childhood household income. To demonstrate relative differences across the income distribution, we translated the HRs and corresponding 95% CIs into differences between children in the highest and lowest income vigintiles, HRvig (online supplemental file).

Model 1 controlled for age, sex and calendar year. Model 2 additionally controlled for other childhood exposures between ages 11 and 15: parental chronic illnesses, the share of years lived in a single-parent household, region, first language and birth order. Finally, in model 3, we added sibling fixed effects by allowing families to have differing baseline hazards, thus comparing the likelihood of medication use between siblings exposed to different levels of family income between ages 11 and 15. As there were practically no differences between siblings in the exposures to parental chronic illnesses, we removed these from the model. However, sibling fixed effects control for everything shared by siblings thus, effectively, also parental health. Model 3 only included individuals with full siblings in the data. The sample reduction is unlikely to cause selection bias, as the distributions of covariates (table 2) and the results of models 1 and 2 (online supplemental table 1) were highly similar between the sibling and total populations. As sensitivity analyses, we ran model 2 additionally controlling for highest parental education between ages 11 and 15, excluding youth with any years lived in single-parent households, and models 1–3 stratifying by income tertile (online supplemental tables 2 and 3). All analyses were performed using STATA V.16.1.36

Table 2

Baseline characteristics of the study populations (total and sibling), Finnish cohorts born in 1979–2003

Results

The use of the three most common medication types, antibiotics, painkillers and allergy and asthma medications was more common among youth with a higher childhood family income (figure 2). Psychotropic use was more common in low-income families.

Figure 2

Incidence of medication purchase between ages 16 and 20 per 100 person-years by medication type and childhood household income vigintile (in 1000 Euros) in Finnish birth cohorts 1979–2003.

In the total population analyses, for each 10% increase in childhood income, there was an increase in 0.8% in the likelihood of using antibiotics, 0.6% for painkillers and 1.7% for allergy and asthma medications between ages 16 and 20, when controlling for age, sex and year (table 3, model 1). Across the income distribution, these translate to HRvig of 1.23, 1.15 and 1.54, respectively, among children in the top 5% of family income compared with children in the lowest 5%. These income gradients were not explained by other childhood exposures in model 2. In contrast, for psychotropics, a 10% increase in childhood income was related to a 2.2% decrease in the likelihood, which translates to a HRvig=0.57 between the top and bottom 5%. About a quarter of this difference was attenuated when controlling for other childhood exposures in model 2. The results were highly similar when only including the population with siblings (online supplemental table 1, models 1–2), when additionally controlling for parental education (online supplemental table 2), and when excluding youth with any years lived in a single-parent household (online supplemental table 2).

Table 3

HR of prescription medication use at ages 16–20 per 10% increase in family income (HR1.1) and between the highest and lowest family income vigintile (HRvig), Finnish birth cohorts born in 1979–2003

In the sibling comparisons, childhood income was not related to the use of any medication type (table 3, model 3). The associations were similar across income tertiles (online supplemental table 3).

Discussion

We used register data on nearly 1.5 million Finnish individuals to assess whether childhood family income predicts use of common prescription medications at ages 16–20. We found that antibiotic, painkiller, and allergy and asthma medication use was more common among youth from high-income families whereas psychotropic use was less common. Our sibling comparisons showed no income differentials in medication use, suggesting that the associations may be explained by unobserved familial factors shared by siblings.

The differences in medication use across the income distribution were sizeable. Youth in the highest 5% of family income was around 15%–55% more likely to use antibiotics, painkillers and allergy and asthma medications, and around half as likely to use psychotropics compared with youth in the lowest. Given that youth is a time of relatively good health and small health inequalities, a difference of 50% between the highest and lowest income groups can be considered substantial. Furthermore, following the life course perspective, early life experiences shape later life outcomes and inequalities in healthcare and medication use in youth may have long-lasting influences on inequalities in adult health and healthcare use.

Our results on antibiotics are similar to prior evidence suggesting more use among youth from high socioeconomic backgrounds, even at the same level of need.16–18 Our results on painkillers and allergy and asthma medications, however, somewhat contrast prior studies: most of the studies that account for the need for treatment, that is, health status or specific health problems, either find less use among youth from highly educated or high-income families or no differences across social groups,5–7 23–27 whereas we find more use in high-income families. This difference may relate to us being unable to control for underlying health. However, it is unlikely that our results merely reflect (unobserved) need for treatment as prior evidence indicates less health problems among youth from more affluent families in Finland.37 Rather, our results are likely to reflect the underuse of antibiotics, painkillers and allergy and asthma medications among youth from low-income families or overuse in high-income families.

The widespread use of private health insurances, particularly in affluent Finnish families,32 is one potential pathway leading to unequal use of common medications in youth. Private health insurances enable swift access to private sector primary care, whereas access to public primary care might be more limited. Having better access to primary care not only increases the likelihood of receiving prescriptions in primary care but also increases the likelihood of referral to specialised care. With our data, we were unable to assess whether families had private health insurances, or whether the medications were prescribed at the public or private sector or in primary or specialised care. Further research is thus needed to clarify where in young people’s care paths are inequalities in medication use established. Are there differences in the rates of first contact, diagnosis, referral—or all of these?

Regarding psychotropics, our results are in line with most of the prior evidence showing less use by increasing income.5 7 8 15 20 21 This is likely to reflect income differences in need, as more mental health problems have been found among Finnish adolescents from low-income families.38 Several prior studies have indeed shown that the excess psychotropic use among youth with lower family socioeconomic position is largely due to having more psychiatric symptoms.7 20 Furthermore, psychotropics are less expensive and more easily available than non-pharmacological treatments and could thus be more commonly used in low-income families even at similar levels of need.

Only a handful of previous studies have assessed confounders of the association between family socioeconomic position and youth medication use.11 15–17 21 We showed that controlling for potentially important observed confounders, such as poor parental health, single parenthood and immigrant status had little impact on the associations, possibly due to methodological limitations. However, our sibling comparison suggests that childhood income may not be an independent predictor of youth medication use and that income differences in medication use are likely to be caused by other, unobserved, family factors shared by siblings, such as patterns of treatment-seeking and adherence. In countries where medication copayments are lower, non-financial influences may be even more important than in Finland where copayments varied between 50% and 60% for most of the studied medications.31

Strengths and limitations

A major strength of our study is the longitudinal population data with reliable register-based information on childhood family income and all prescription medication purchases. The data avoid common problems of non-response, misreporting and recall bias on childhood measures. The large data with linkages between family members also enable sibling comparison analyses, which have not previously been used to study determinants of youth medication use. However, we excluded 17% of the original cohort mostly due to unknown parents. Excluded individuals were more likely than the study population to be male (53% vs 51%) and predominantly (82%) had a first language other than Finnish/Swedish. Based on the language they are likely to be predominantly immigrants and thus on average from low-income families with less healthcare use than the native population. Our results therefore presumably underestimate the income differences in medication use among the full cohorts.

The main limitation of our register-based study is the lack of measurement for underlying health status in youth. We were thus unable to measure and control for the need for medications. Data linking individual-level health measures from large surveys with register-based measures on socioeconomic factors and medication use would be ideal for further assessing this issue.

While we addressed many previously neglected confounders, our measures have their limitations. Our measures for parental health captured some of the most common chronic physical conditions among the Finnish adult population, such as hypertension, asthma and diabetes.39 However, our measures poorly captured other common health problems such as musculoskeletal diseases and psychological symptoms. Parental psychological symptoms could be particularly important in explaining socioeconomic differences in youth medication use, and require further study. Furthermore, reliably differentiating between single parenthood and shared physical custody was not possible because in Finland children can only be registered at one permanent address. Youth categorised as having lived in single-parent families may thus have actually alternated between two parental households. However, excluding these youth had negligible impact on the results (online supplemental table 2) so even a more accurate measure of single parenthood would be unlikely to explain the income differences in medication use.

Another limitation is the potential lack of between-sibling variability in childhood income. While 20% of the total sample deviated from the population mean income by 20 000 Euros or more, 20% of the sibling population deviated from the family mean by only 5300 Euros or more (table 2). The share of the sibling population deviating from the family mean by more than 10% (ie, one unit of our income measure) was nevertheless more than 20%, and the share deviating by more than 5% (ie, half a unit) was more than 50%, so the between-sibling variation was non-negligible. Given the reduced variation in income compared with the total population, however, it may be difficult to identify income effects in the sibling fixed-effects analyses, and the results should thus be considered suggestive.

Conclusions

Apart from psychotropics, our results may indicate underuse of the most common medications among youth from low-income families, because previous studies have shown no corresponding differences in underlying health conditions. The sibling comparisons, however, suggest that moderate differences in childhood income are unlikely to cause differences in youth medication use and the observed differences by income are likely to reflect other, unobserved, family factors such as differential family patterns of treatment-seeking and adherence.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by Statistics Finland Board of Statistical Ethics (TK/638/07.03.00/2023) Social and Health Data Authority Findata (THL/5674/14.06.00/2023). This study is based on secondary data collected for administrative and statistical purposes. The study complies with the national legal framework for accessing pseudonymised personal data for scientific research carried out in public interest. The legal basis is stated in the Finnish Personal Data Act (523/1999), Act on Secondary Use of Social and Healthcare Data (552/2019), Finnish Statistics Act (280/2004) and the EU General Data Protection Regulation (GDPR). The GDPR permits processing this type of data for research without using the GDPR consent (Art. 9 of the GDPR). Informed consent is thereby waived.