Setting

This cross-sectional investigation encompassed the whole country. All COVID-19 vaccinations registered before 30 June 2021 were used in this study. The government organized a free-of-charge vaccination program using the Oxford/AstraZeneca COVID-19 vaccine, Janssen Ad26.COV2.S COVID-19 vaccine, Moderna COVID-19 (mRNA-1273) vaccine, Pfizer-BioNTech COVID-19 mRNA vaccine, Sinopharm COVID-19 vaccine, and the Sputnik V vaccine. The National Health Insurance Fund (NHIF) registered all vaccinations.

Mapping the segregated colonies

A governmental decree defines SCs as within settlement (within towns and within villages) clustering of residents 18–59 years old with not higher than primary level education and a lack of work-related income. The Hungarian Central Statistical Office determines the cluster or clusters as SCs and the complementary areas (CA) of the same settlements not belonging to any SC for all Hungarian settlements. Each Hungarian household is classified in this system as either an SC or a CA, in a mutually exclusive manner.

With the addresses of the adults the NHIF can define populations living in certain SCs or CAs. Therefore, all Hungarian adults can be classified as inhabitants living in an SC or living in a CA.

Data sources

Data were provided by the NHIF to which reporting COVID-19 vaccinations is compulsory. The NHIF updated their database day-by-day. In this study, vaccinations until 30 June 2021 were included.

The NHIF characterized all patients by their age, sex and eligibility for an exemption certificate (deprived patients with chronic disease are supported by an exemption certificate which is released by the local municipality based on the recommendation of the patient’s GP), and prepared the age group-, sex-, and eligibility for an exemption certificate-aggregated tables for each SC and CA based on the administration of the first and second COVID-19 vaccination doses.

The NHIF provided the target population data in the same structure. The age- (5-year bands), sex-, and possession of exemption certificate-specific numbers of vaccinations and populations were provided for each SC and CA.

Statistical analysis

People who have received both doses of the vaccine are defined as ‘vaccinated’ in this analysis. The crude vaccination coverage (cVC) was calculated for each SC and CA. The different vaccine types were not distinguished. Age, sex, and eligibility for exemption certificate-specific reference vaccination ratios for the whole population of Hungary were calculated and used in indirect standardization. Age, sex, and exemption certificate standardized vaccination coverage (sVC) was computed for each SC and CA.

The settlement-specific sVC was evaluated by comparing to the local reference value of settlement-specific sVC in CAs. Relative standardized vaccination coverage (RsVC), the ratio of sVCs for the SC to sVCs for the CA of the same settlement, along with the corresponding 95% confidence interval (95% CI), were calculated. (In the settlements where more than one SC was located, settlement-specific aggregated SC measures were computed by summarizing the observed number of vaccinations, populations, and the expected number of vaccinations.) The excess number of vaccinations was calculated as the difference between the observed and the expected number of vaccinations. The proportion of vaccinations attributable to segregation (attributable risk) was computed as the ratio of excess and observed number of vaccinations.

SPSS version 20 (IBM Corporation, New York, NY, USA) was used for the data analysis.

Ethical permission

Data provided by the Hungarian NHIF were used in this study. All data processed in our secondary analyses were geographically aggregated. Because individual information was not used, ethical approval and written informed consent were not required according to the Hungarian legal framework. The protocol to produce segregation-specific indicators was approved by the Office of the Commissioner for Fundamental Rights (AJB-3147/2013), the general director of the NHIF (E0101/215-3/2014), and the Hungarian National Authority for Data Protection and Freedom of Information (NAIH/2015/826/7N). Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the (S1 Checklist).

Main findings

Our study shows that COVID-19 vaccination coverage is significantly lower in SCs than in CAs according to both cVC (40.05% vs. 65.42%) and sVC (0.653 vs. 1.017). This demonstrates that the Hungarian COVID-19 vaccination program did not avoid geographic inequality, which has a strong ethnic dimension. This observation confirms the reported experiences from other countries about the inequality-generating characteristics of COVID-19 vaccination programs [17–23].

The cVC was much higher and vaccination inequality was much less pronounced among elderly individuals older than 60 years, compared to adults in the 18–59 age range. Organizational efforts were much more intensive and started earlier among elderly adults than among younger adults. The aim of public health authorities is not only to reduce of the epidemic’s intensity (reduction of susceptible population) but also to prevent serious complications, especially the infection of elderly individuals who were at higher risk. Personal motivation was also larger among elderly individuals for the same reasons.

By the time of our investigation, a total of 165,988 adults (9.24% of the sensitive population) were from the SCs and 1,629,216 from CAs. If the vaccination coverage in SCs were equal to that in CAs, then there would be only 95,758 susceptible inhabitants in SCs. It would be reflected in 3.91% decrease of the number of susceptible adults in the studied population, and in the 5.55% of the susceptible inhabitants would be from SCs. Consequently, although, the pandemic related future health loss could be prevented basically by the elevation of the vaccination in the CAs, a not negligible risk reduction could be achieved by the enforced/adapted vaccination project in the SCs.

Our investigation described the huge variability of settlement-specific RsVC. SC-related under-vaccination was generated by 437 SCs, where the number of adult inhabitants was 208,494 and the number of missing vaccinations was 47,494.5. The capacities required to make up the missing vaccinations seem to be small, given that in Hungary, 4,746,822 adults have been vaccinated at the time of our study.

It is probable that the vaccination inequality demonstrated by our study will increase the intensity of the fourth wave of the pandemic and result in health losses in Hungary, as it was observed in the second wave as area deprivation related COVID-19 mortality inequality [41].

Strengths and limitations

This investigation covered the entire country, which prevented selection bias. Because vaccination registration is compulsory in Hungary, misclassification of the vaccination status was also avoided.

The unique feature of the Hungarian health statistical system is that the segregated colonies populated mainly by Roma are defined in the census data, and health indicators can be computed for these geographical units. Since, the registration of ethnicity in Hungarian health care administration is prohibited as it is in other European countries as well; the main strength of our study design was that it could avoid the violation of personal rights by utilizing SC related opportunity.

The living place of adults could be misclassified if the registered residential address was in a SC but the real residential address was in a CA (or vice versa). The extent of this residential place misclassification is not known. Therefore, the potential dilution of the observed RsVC could not be quantified. Consequently, the real difference between SCs and CAs is underestimated by the observed RsVC = 0.643 (i.e. the real RsVC is less than 0.643). Furthermore, it must be acknowledged that this misclassification could be strong in certain SCs, where the segregation effect could be seriously underestimated. Therefore, there could be nonidentified high-risk SCs. Altogether, the segregation effect suggested by an RsVC of 0.643 is valid, and the identified high-risk SCs are real high-risk populations.

In this study, calculations for cVC, sVC and RsVC were made with the assumption that SC localization has not changed from the last Hungarian census collection. Since the most recent, census took place in 2011, both geographic boundaries and demographic composition of SCs and CAs may have changed due to internal migration. This effect can be evaluated having the data from the next census, which will be implemented in 2022.

Taking into consideration that SCs are dominated but not exclusively inhabited by Roma, and about three quarters of the Roma population does not reside in SCs, our findings of SC vaccination inequality cannot be extrapolated to characterize Roma vaccination inequality in general.

This study was not able to collect data on many factors that can affect vaccination coverage apart from age, sex, and exemption certificate eligibility [42]. A detailed exploration on the background of SC-specific vaccination’s variability needs further study. The social, cultural, economic and/or environmental differences between settlements with vaccination inequalities in SCs and those without inequality could be confounding variables in the rates of vaccination in these communities. Thus, there must be further understandings of the ethnic makeup and social determinants of health in Hungarian SCs.

Implications

The detected inequality demonstrates the necessity of special intervention programs targeting SCs in Hungary. Clearly, the general program was not as effective in SCs. The general program considered elderly individuals, patients with chronic disease, and health care personnel as high-risk groups. It appears that the population living in SCs, comprised dominantly of Roma, is also a high-risk group. Therefore, the population living in SCs should be considered high-risk target groups, where the general vaccination program should be adapted.

Extending the high-risk group definition in Hungary could help to mitigate the impacts of the epidemic on disadvantaged, mainly Roma minority of SCs. Although, there are obvious differences between Hungarian SC inhabitants and the ethnic minorities of the United States and the United Kingdom, this reconsideration is supported by published positive experiences and elaborated methodologies in these countries where vaccination organizations and monitoring are adapted to racial/ethnic minorities [6, 43–48].

It is also urgent to expand these methods in Hungary to reduce COVID-19 vaccination inequalities. If it is successful, then the approach could be used as a template for other inequality reductions. Otherwise, if a special intervention program is not implemented and applied, the lesson by the end of the pandemic will be that the COVID-19 epidemic (and other problems) can be managed without considering ethnic inequalities and accepting the associated additional health loss.

One of the seven objectives of the European Union is to “Increase effective equal access to adequate desegregated housing and essential services” showing that segregation is among the most important characteristics of European Roma. It is highly probable that segregated colonies of Roma in other European countries face similar problems demonstrated in Hungary [34]. Moreover, our observations could help to characterize inequities occurring in similarly segregated groups around the world. For example, in the United States, Native American populations on reservations closely resemble the geographic and ethnic isolation patters exhibited by the Roma in Hungary [49]. Thus, understanding vaccination discrimination in the Roma-populated SCs parallels the in access to healthcare and vaccination disparities of minority populations in the United States [50, 51].