This study was a part of the “Public Health Focused Model Program for Organising Primary Care Services Backed by a Virtual Care Service Center” program in Hungary (Ádány, 2013). The main aim of this program was to reorganize preventive service delivery at PHCs by the GP team and to provide improved care without discriminating against the Roma people (Sándor et al., 2016). This program was conducted in 21 general medical practices from Borsod-Abauj-Zemplen, Heves, Jasz-Nagykun-Szolnok, and Hajdu-Bihar counties located in northeastern Hungary in the period of 2012–2017.

Applying the definition of SRC as a segregated part of the settlement with at least four disadvantaged residential units (apartment/house/block of rooms/hut), with a lower housing condition and poorer environmental quality than the CA (Kósa et al., 2009; Sándor et al., 2018), a survey was carried out to identify addresses belonging to the SRCs as a part of our presented study by epidemiologist of the intervention program. Because, GP knew addresses of each patient, they could classify their clients as SRC inhabitant or CA inhabitant. In the intervention area, there were 67,017 residents. Among these residents, the numbers of inhabitants living in SRC and in the CA were 4,943 and 62,074, respectively.

Our investigation was carried out in 2012 before the interventions had been launched.

Data were provided by the NIHIFM. The unit of analysis was the prescription written by the health care provider and redeemed by the patient during the calendar year of 2012 before the interventions of the program had been launched (Sándor et al., 2016). The age- (applied age groups: 0–17, 18–24, 25–44, 45–64, and 65 years and above) and sex-specific nominators and denominators of the prescription redemptions were computed for the SRC and the CA. The prescribed medications were classified according to the WHO, 2012 version of the Anatomic Therapeutic Chemical (ATC) first level classification of drugs (WHO, 2011). Prescriptions written for all age groups were included in the analysis.

The crude redemption ratios were calculated by age, sex, and ATC group for the SRC, the CA, and Hungary. Indirect standardization was applied to ensure the comparability of SRC- and CA-specific indicators. Age- and sex-specific expected proportions of redeemed prescriptions were calculated by age- and sex-specific numbers of written prescriptions and the national reference redeemed-to-written ratio for all strata (Harsha et al., 2019a; Harsha et al., 2019b). To control for the confounding effect of the demographic factors, standardized redemption ratios (SRRs) were calculated by dividing accumulated age- and sex-specific numbers of observed redemptions by accumulated age- and sex-specific numbers of expected redemptions of prescribed drugs for both the SRC and the CA.

Relative redemption (RR), the ratio of SRRs for the SRC to SRRs for the CA, along with the corresponding 95% confidence interval (95% CI), was calculated for each ATC group and for the total practice. Attributable numbers of redemptions were calculated as the product of the difference in SRRs for the SRC and CA and the expected number of redemptions in the SRC.

This is the first study to investigate primary medication adherence among Roma people. We could evaluate redemption in the most vulnerable subgroup of Roma people who live in SRC by standardizing the redemption according to patient age and sex using Hungarian stratum-specific redemption ratios as a reference.

In general, the Hungarian crude redemption ratio (66.8%) was higher than those reported in the United States (60.7%) (Fernández et al., 2017) and Netherlands (48.5%) (van Geffen et al., 2009), but lower than the global estimate reported in a recent systematic review and meta-analysis (83%) (Cheen et al., 2019) and the crude redemption ratio reported in Portugal (77.2%) (da Costa et al., 2015) and Denmark (90.7%) (Pottegård et al., 2014). The discordance with a systematic review and meta-analysis might be due to firm follow up and good patient-physician relationship at comprehensive chronic care unit for chronic diseases. The positive effect of female sex on redemption in Hungary was found to be weak but statistically significant, which is consistent with observations from the United Kingdom, the United States and South Korea (Hagström et al., 2004; Farris et al., 2018; Lee et al., 2018). The relatively low redemption ratio among young and middle-aged Hungarian adults also fits the reported age dependency from other countries (Cooper et al., 1982; Mansur et al., 2009; Wenger et al., 2017).

Altogether, the observed primary medication adherence was better in the SRC than in the CA. The observed 2.8% excess corresponds to 920 redemptions a year in the population of 4,943 Roma people living in the SRC. Higher redemption rates in disadvantaged populations, as indicated by low levels of education and high level of deprivation, was observed previously in Hungarian studies (Boruzs et al., 2016; Boruzs et al., 2018; Harsha et al., 2019a; Harsha et al., 2019b; Juhász et al., 2020) and in other countries as well (Mauskop and Borden, 2011; Jackson et al., 2014; Rosicova et al., 2015; Ferdinand et al., 2017; Pandya and Bajorek, 2017; Lee et al., 2018; Gast and Mathes, 2019).

Cardiovascular drugs and alimentary tract and metabolism-related medications constituted the highest share of the prescriptions in both investigated populations. This reflects the prevalence of these chronic diseases in Hungary, as has been reported in previous studies (Balogh et al., 2010; Szigethy et al., 2012). Roma people living in the SRC exhibited better redemption of cardiovascular drugs (11.5%) and alimentary tract and metabolism-related drugs (3.5%) than those living in the CA, a difference that was confirmed by the standardized analysis.

Considering the number of prescriptions, the third most important ATC class is the group of anti-infectives for systemic use. The redemption inequality in this ATC group showed a reverse pattern: the redemption ratio was relatively high in the CA but lowest in the SRC, resulting in a significantly lower redemption ratio for the SRC. We found lower general redemption ratios in the SRC among children and the young adult age groups (Table 1), and a low relative importance of cardiovascular (0.7% of prescriptions, Supplementary Table 3C) and alimentary tract drugs (10.6% of prescriptions, Supplementary Table 3A) due to the low prevalence of chronic diseases among children and young adults. In addition, we observed a high importance of infections and anti-infective drugs among these groups (72.5% of prescriptions, Supplementary Table 3J). This lower redemption ratio among children could be the explanatory factor behind the observed deviation of the anti-infective redemption habit in the SRC from the general pattern. However, further investigation is needed to determine whether the causal mechanism behind this pattern is the impact of a Roma culture-related attitude toward children or the avoidance of antibiotic overuse among the Roma (Porco et al., 2012; Basu and Garg, 2018). The role of the poor living conditions determined high rate of infective disorders (reflected in the higher proportion of both prescription and redemption of antibiotics) in SRCs (Table 1) needs clarification as well.

Our study demonstrated that blaming the Roma for their poor health (beyond the fact that it is ethically unacceptable and it is a factor that contributes to the maintenance of their poor health) (European Union Agency for Fundamental Rights, 2009b; FRA/UNDP, 2013; Miranda et al., 2019) cannot be supported by data on primary medication adherence. Moreover, responsible health policy should consider the relatively good primary medication adherence of this population as an opportunity to organize culturally adapted interventions. When the poor healthcare utilization of this population is targeted by an intervention (Miranda et al., 2019), this factor should be added to the resources of the program. Our observations suggest that GPs should prefer medication based care instead of the non-medication based (life style modification, physiotherapy etc.) approaches for inhabitants of SRCs.

This study used data from the NHIFM, which covers the whole country. Selection bias is likely very low, as we included all the residents’ prescriptions in the study area. Because, this database has no data on the use of dispensed medications, only the primary non-adherence could be evaluated. Further, our study due to the SRC vs CA comparison design could avoid all the uncertainties which are related to the person level identification of Roma participants.

However, our study has some limitations. Although our study recruited a large sample, this does not mean that it shows representative national Roma figures. The main limitation is that our results cannot be interpreted as an indicator of Roma vs non-Roma differences. On one hand, 6% of people living in SRC are non-Roma (Kósa et al., 2013). On the other hand, three-quarters of the Roma are living in the CA in Hungary (Teller, 2010). Consequently, our SRC-based analysis underestimated the real Roma-to-non-Roma differences.

A weakness of this study is that its dataset is from 2012. But the primary medication adherence of SRCs’ inhabitants had been not investigated before. It was not known whether the primary medication adherence is one of the risk factors for Roma people. The intervention program could build up the necessary database in 2012. Taking into consideration the significance of our results further studies are needed, obviously, to monitor the primary medication adherence among Roma people.

A prescription is proof that the medicine is needed for the patient. The confounding effect from the obvious influence of clinical status on the motivation of the patient to purchase the medicine is controlled by the prescription itself, as the prescription does fit the clinical status of the patient.

Although, price of drugs is one of the key factors that affect redemption, economic explanations for the observations could not be formulated in this investigation, because data on participants’ income were not collected to avoid the jeopardisation of trust between participants and implementation staff of the intervention program and data of the NHIF on co-payment for drug purchasing was not allowed to use in our study. Therefore, the complex issue of economic risk factors of primary medication adherence could not be investigated in our project. Their role needs further studies. It restricts the interpretation, but not the validity of our reported findings.

The availability of PHC in Hungary is among the best in Europe. Moreover, it has been reported that the Roma people living in SRC visit the GP more frequently than does the general population (Sándor et al., 2018). It appears improbable that limited access to primary care could somehow influence redemption practices.