The research, published in SSM – Population Health, was led by Hanyul Lee and colleagues, who applied an analytical framework known as MAIHDA—Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy—to a question that has long resisted conventional methods. Unmet medical need, or UMN, refers to situations in which people perceive a need for healthcare but cannot obtain appropriate or sufficient care. Previous studies have typically treated UMN as a single aggregate outcome, or examined risk factors one at a time using standard regression. The problem, the authors argue, is that additive approaches cannot capture what intersectionality theorists have described since Kimberlé Crenshaw’s foundational work: social categories do not simply stack. A woman with a parental role does not experience the sum of ‘female disadvantage’ plus ‘parental disadvantage.’ The intersection generates a qualitatively distinct experience that cannot be reconstructed from its component parts.

MAIHDA addresses this by treating each combination of social characteristics—144 unique strata in this study—as a group in a multilevel logistic regression model, with individuals nested within strata. The method models strata as random effects, which stabilizes estimates even for small groups through statistical shrinkage, a major advantage over stratified analyses that lose power or interaction-term models that become uninterpretable as axes multiply. The researchers fitted two models for each outcome: a null model capturing total between-stratum variation, and a main-effect model adding the average associations of each axis. The difference between the two reveals whether social positions interact in ways that additive averages cannot explain. The team also disaggregated UMN by cause, separating time-related barriers—such as inconvenient clinic hours, inability to leave work, or lack of childcare—from cost-related barriers, where the expense of care itself is prohibitive.

The main effects told a story of two different barriers. For time-related UMN, age, sex, and employment dominated: younger adults had more than three times the odds of older adults, women had 58 percent higher odds than men, and employed individuals had roughly three times the odds of those not employed. Living with children also raised the odds. For cost-related UMN, the picture inverted. Income was the decisive factor—the lowest-income group had more than four and a half times the odds of the highest—and living alone nearly quadrupled the odds. Employment, meanwhile, was protective: employed people had 45 percent lower odds of cost-related UMN than those not employed. Sex and age showed no significant association with cost-related barriers at all. The same social variable, in other words, can push in opposite directions depending on which barrier to care is being measured.

The variance decomposition underscored why disaggregation matters. Intersectional strata explained 18.6 percent of the variation in time-related UMN and 27.3 percent in cost-related UMN in the null models, but only 4.8 percent for overall UMN. When causes are collapsed into a single outcome, effects associated with different barriers partially offset one another, masking the true extent of social patterning. The authors note that even 4.8 percent is not trivial by MAIHDA standards, where values below five percent are common, particularly given that no health-condition variables were included. Discriminatory accuracy followed the same gradient: the area under the receiver operating characteristic curve reached 0.80 for cost-related UMN, compared with 0.60 for overall UMN, meaning that cause-specific models identify high-risk individuals far more effectively.

After accounting for average additive effects, the residual variation attributable to intersectional interaction effects shrank substantially—the proportional change in variance exceeded 80 percent for all outcomes. But the residual effects, though small in aggregate, were concentrated in a handful of strata that proved to be the study’s most revealing finding. For time-related UMN, the three highest interaction effects all belonged to a single social location: younger women who were not employed and living with children. Their interaction effects reached odds ratios of 2.10, 2.06, and 1.59 across the middle-, lowest-, and highest-income tertiles respectively. These women’s predicted probabilities of time-related UMN were comparable to those of employed young adults—an anomaly, since non-employment would ordinarily be expected to free up time. No comparable pattern appeared among non-employed young men.

The authors offer a plausible explanation rooted in the sociology of care. The ideology of intensive mothering prescribes continuous maternal availability and prioritization of children’s needs, which may render non-employment not as relief from time pressure but as an expectation of full-time caregiving. Under such norms, mothers who prioritize their own health needs can experience guilt, leading to deferred care. The pattern is consistent with stark gender asymmetries in unpaid labor: in South Korea, non-employed women in single-earner households with children spend roughly seven hours per day on unpaid household work, about double the time of non-employed men in the reverse configuration. International evidence similarly shows that male unemployment does not translate into equivalent increases in men’s unpaid care work. Gendered role expectations, in other words, operate in both directions—directing women toward child-centered domestic routines while discouraging men from assuming caregiving responsibilities.

For cost-related UMN, the most deviant strata told an unexpected story about men. The highest interaction effect of all belonged to middle-aged men who were not employed and living alone in the highest-income tertile—an odds ratio of 2.60, despite their objective affluence. Two other middle-aged male strata living alone, one non-employed and low-income and one employed and low-income, ranked among the top four. The authors suggest that cost-related unmet need in this group may reflect perceived affordability or the prioritization of healthcare spending rather than genuine financial constraint. Gendered divisions of household labor have historically positioned women as managers of family health, monitoring symptoms and facilitating medical visits; men without a co-residing partner may be less likely to act on health needs at all. In South Korea, where breadwinner norms are particularly strong, non-employment in midlife may also represent a disruptive role failure that further deprioritizes healthcare spending.

The policy implications are pointed. Interventions targeting formal workplaces—flexible hours, sick leave—would reach employed adults but miss the non-employed young mothers whose time-related unmet need was among the highest of any group. Conversely, means-tested subsidies for healthcare costs would capture low-income households but overlook affluent middle-aged men living alone who forgo care for reasons that may be social and psychological rather than financial. The authors caution that their study is cross-sectional and descriptive, that respondents could report only one reason for unmet need, and that binary sex categories cannot fully capture gender diversity. Yet the sensitivity analyses—adjusting for hypertension and diabetes, and excluding those with no perceived need—left the key findings intact. The broader message is that equitable access under universal coverage requires attention not only to entitlement and affordability, but to the gendered organization of work, care, and household responsibility that determines who, in practice, can walk through the clinic door.

Subject of Research: Intersectional gendered disparities in cause-specific unmet medical need in South Korea

Article Title: Gendered experiences of cause-specific unmet medical need: An intersectional analysis using MAIHDA

Article References: Lee, H., Choi, W., Lee, W., & Kim, H. (2026). Gendered experiences of cause-specific unmet medical need: An intersectional analysis using MAIHDA. SSM – Population Health, Article 101980. https://doi.org/10.1016/j.ssmph.2026.101980

Image Credits: AI Generated

DOI: 10.1016/j.ssmph.2026.101980

Keywords: unmet medical need, intersectionality, MAIHDA, gender inequality, South Korea, universal health coverage, healthcare access, time poverty, caregiving, household composition, health equity, multilevel analysis

The research team, led by Jesper Löve of the University of Gothenburg together with colleagues including Gunilla Priebe, Bo Burström, Ailiana Santosa, and Nawi Ng, drew on the SCIFI-PEARL database, a nationwide register platform that links national and regional records on notifiable diseases, hospitalisations, causes of death, and socioeconomic circumstances. Their dynamic cohort covered the entire adult population of Sweden, some eight million people aged 18 and older, followed from March 2020 to March 2022. Over that period the registers recorded 67,997 COVID-19 hospitalisations and 15,895 deaths, corresponding to 84.3 hospitalisations and 19.7 deaths per 10,000 population.

What distinguishes the study is its method. Rather than treating each social factor separately, the researchers used an approach known as Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy, or MAIHDA, which is grounded in intersectionality theory. The framework, first articulated by legal scholar Kimberlé Crenshaw in 1989, holds that social categories such as gender, class, and migrant status do not operate in isolation but mutually shape one another within systems of privilege and oppression. The team combined five dimensions — age group, sex, income relative to the at-risk-of-poverty threshold, educational attainment, and birthplace classified by the World Bank income level of the country of origin — into 80 distinct intersectional strata, and modelled individuals nested within those strata.

The statistical logic is elegant. A first, unadjusted model estimates how much of the total variation in outcomes is attributable to the intersectional strata, expressed as the intraclass correlation coefficient. A second model adds all five defining dimensions as fixed main effects; if no interactions existed between the dimensions, all between-stratum variance would be explained and the random effects would vanish. Whatever variance remains points to intersectional interactions beyond the simple sum of each characteristic’s effect. This structure also partially pools information across strata, stabilising estimates for small or sparsely populated combinations — a decisive advantage over saturated interaction models that become unstable when many dimensions are crossed at once.

The results were striking. Intersectional strata explained a substantial share of the variance in outcomes, particularly for mortality among adults under 70, where intraclass correlations reached 25 percent in the first wave, 15 percent in the second, 17 percent in the third, and 22 percent in the fourth. For hospitalisation, the corresponding figures ranged between 7 and 14 percent depending on age group. Among people aged 70 and older, by contrast, the strata explained less of the variation — typically at or below 10 percent — suggesting that biological susceptibility and frailty dominated severe outcomes in old age, whereas in younger adults the patterns more clearly reflected socially structured differences in exposure, living conditions, and access to protective resources.

The most severe risks clustered in specific and sometimes unexpected combinations. During the first wave, hospitalisation among older migrants born in low-income countries was roughly four times higher than among their counterparts born in high-income countries. Older men born in low-income countries with low income and low education recorded hospitalisation rates as high as 3,550 per 10,000 in wave one. Yet the study repeatedly found that broad labels concealed enormous heterogeneity. Among older men born in low-income countries, first-wave mortality ranged nearly fourfold, from 346 to 1,377 per 10,000, depending on income and education. In some waves, migrants in advantaged strata fared better than certain Swedish-born men with low income and low education, who themselves appeared among the ten highest-risk strata in three of the four waves.

Equally important was the finding that vulnerability was temporally fluid. No single stratum occupied the highest-risk position across all four waves. In the first two waves, the ten highest hospitalisation strata among younger adults consisted exclusively of migrant men, but by wave three upper-middle-income-country-born women with low income and low education had joined them, and by wave four migrant women filled five of the ten highest positions. Mortality showed even greater fluctuation: among low-income, low-educated men born in low-income countries, rates fell from 44 per 10,000 in the first wave to 6, 18, and 7 per 10,000 in subsequent waves, a trajectory the authors link to shifting viral variants, the vaccination rollout, and evolving policy conditions.

One particularly revealing interaction emerged among younger men born in low-income countries who had high educational attainment but low income. The usually protective effect of education reversed in these strata — a pattern the authors attribute to the well-documented occupation–education mismatch experienced by many migrants in Sweden, whereby highly educated immigrants are over-represented in essential, high-exposure occupations. Overall, however, most inequalities followed an additive pattern, best described as ‘double jeopardy’, in which each marginalised position adds to disadvantage, rather than the compounded ‘multiple jeopardy’ of effects beyond additivity.

The authors are careful to stress that intersectional strata are analytical constructs, not inherent risk groups. As they and other scholars note, identifying categories as stand-alone risk factors risks diverting attention from the structural processes — crowded housing, high-risk work, inadequate health information, institutional distrust — that actually produce vulnerability. The pandemic, they argue, was a syndemic in which viral exposure interacted with pre-existing medical and social inequalities, and Sweden’s early reliance on voluntary recommendations and individual responsibility may have interacted differently with the social and occupational conditions of different groups.

The study’s implications reach well beyond Sweden. Because vulnerability shifted across pandemic phases and social positions, the authors conclude that pandemic preparedness cannot rely on static assumptions about ‘high-risk’ groups. Surveillance and interventions, they argue, should be responsive to changing patterns of inequality across intersecting social positions, and reducing the inequitable consequences of future pandemics will require not only intersectionality-informed monitoring but political commitment to addressing the structural conditions — living circumstances, social protection, and pre-existing health burdens — that make some people far more exposed than others when a new pathogen arrives.

Subject of Research: Intersectional inequalities in COVID-19 hospitalisation and mortality across four pandemic waves in Sweden, analysed with multilevel intersectional methods.

Article Title: Intersectional Inequalities in COVID-19 Hospitalisation and Mortality: A Nationwide MAIHDA Study of Swedish Adults Across Four Pandemic Waves (2020–2022)

Article References: Löve, J., Priebe, G., Burström, B., Santosa, A., & Ng, N. (2026). Intersectional Inequalities in COVID-19 Hospitalisation and Mortality: A Nationwide MAIHDA Study of Swedish Adults Across Four Pandemic Waves (2020–2022). SSM – Population Health, Article 101968. https://doi.org/10.1016/j.ssmph.2026.101968

Image Credits: AI Generated

DOI: 10.1016/j.ssmph.2026.101968

Keywords: COVID-19, intersectionality, MAIHDA, health inequalities, Sweden, hospitalisation, mortality, pandemic waves, migration background, socioeconomic status, public health, pandemic preparedness