This is the first study to report on COVID-19 outcomes between Kuwaiti and non-Kuwaiti patients. We found non-Kuwaitis were two times more likely to be admitted to the ICU or die from COVID-19. With adjustment to age, smoking and certain co-morbidities, non-Kuwaitis had two- and three-fold increase in odds ratios for ICU admission and death, respectively. While in many instances our underpowered results did not show statistical significance, the effect estimates for all outcomes were consistently in the same direction and magnitude especially when we considered a composite outcome of ICU admission or death.
Ethnicity is a complex entity. It represents social constructs, cultural identities, behaviors, as well as a large pool of genetic make-up.[18] The non-Kuwaiti subpopulation is a mixture of different ethnicities and races. We therefore interpret our findings in terms of socioeconomic and environmental rather than biological explanations. In Kuwait, migrant workers make up the majority of non-Kuwaitis. These workers usually have poor working and housing conditions and often performing unskilled labor jobs with minimal formal education. Most migrant workers are employed in unskilled labor jobs with low-paying wages and a background of low socioeconomic status (SES).[19, 20] Hence, the non-Kuwaiti subpopulation can serve as a proxy for SES.
Healthcare access and use may have a role in explaining the health disparities among marginalized subpopulations.4 For one, non-Kuwaitis might have underlying medical conditions that were left undiagnosed and untreated. Such conditions might be key factors into making their condition with COVID-19 worse. In general, migrant workers may have difficulty communicating with medical professionals, do not have access to interpreters and have limited knowledge of the health insurance systems.[20, 21] Kuwait has passed laws and regulations ensuring that all COVID-19 related care is provided for free regardless of citizenship status. However, with regards to non-COVID-19 care, all public-sector healthcare is provided to Kuwaiti citizens for free, but there may be limitations on coverage for non-Kuwaiti expatriates, with basic healthcare services being covered by an annual health insurance fee. Similarly, ineffective public health messaging with migrant workers who grapple with linguistic and cultural barriers may delay diagnosis and treatment of COVID-19.[9, 22] That is, non-Kuwaiti migrant workers may be presenting late and with more severe symptoms.
It is noteworthy that in studies that adjusted for socioeconomic factors, variations in hospitalization and mortality in marginalized subpopulations were not fully explained.[22, 23] Therefore, while our socioeconomic explanations are tentative, other differences including the environment warrant careful examination. Migrant workers in particular tend to live in over-crowded housing with unsanitary, shared bathrooms and kitchens. The segregated unmaintained residential neighborhoods that they live in may be another contributor to poor health.[24, 25]Many are living in food deserts, and their meager wages - most of which are sent overseas to support their families - may leave unhealthy food outlets as the only affordable option. Limited local transportation in such neighborhoods, or an inability for individuals to afford access to personal transportation may further frustrate their access to healthcare. Migrant workers in Kuwait were especially vulnerable to air pollution and extreme heat owing to a significant high exposure heterogeneity among the population in Kuwait.[14, 15] Emerging research is now showing that people living in areas with high outdoor ambient air pollution levels were at higher risk of dying from COVID-19 compared to those living in less polluted neighborhoods.[26] It is possible that the long-term exposure to air pollution may play a role in exacerbating the severity COVID-19 infections among non-Kuwaitis.
Suggestions of genetic susceptibility among ethnic minorities and COVID-19 outcomes are at best inconclusive and at worst could lead to victim-blaming. For example, different people may have different ACE2 (the receptor for SARS-CoV-2) expressions leading to greater susceptibility to COVID-19 in some subgroups.[27] One study used single-cell sequencing, reported that expression of ACE2 was more predominant in Asian men.[28] An observational study looked at patients in two Spanish hospitals and found an increased prevalence of androgenetic alopecia amongst those infected with COVID-19 in which no hormonal measures were taken.[29] Taken together, the link is not yet established, and further studies need to be conducted before such conclusions can be taken.
This study has a number of limitations. Firstly, being non-Kuwaiti (our exposure of interest) is not necessarily a valid proxy of low SES for every non-Kuwaiti individual. Since we did not have SES data, we may have included individuals with high SES resources under the non-national variable. However, if we assume that high SES is associated with a lower probability of adverse COVID-19 outcomes, then the misclassification bias from this proxy may have attenuated the relationship we observed. Secondly, control for smoking was not optimal. The prevalence of smoking in our study population was very low and we did not have information on long-term use (e.g. pack-years, or years of smoking). Although the evidence is still not conclusive, if smoking is associated with a higher probability of adverse COVID-19 outcomes and is more prevalent amongst the non-Kuwaiti sub-population, then we cannot rule out a residual confounding that could overestimate our observed relationship. Thirdly, many individuals had missing BMI data. In our complete-case-only analysis (model 3), the low sample size produced wide variability around our estimates when we adjusted for BMI. Fourth, we did not have data on neighborhood characteristics including air pollution, which is likely to be worse in areas populated by non-Kuwaitis compared to Kuwaitis. If, for example, long-term exposure to air pollution is associated with worse COVID-19 outcomes, then we may have overestimated our observed estimates since we did not control for it. Finally, our sample size was not powered enough to detect interaction by gender.