Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The association between socioeconomic status and pandemic influenza: Systematic review and meta-analysis

  • Svenn-Erik Mamelund ,

    Contributed equally to this work with: Svenn-Erik Mamelund, Clare Shelley-Egan, Ole Rogeberg

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Centre for Research on Pandemics & Society, Oslo Metropolitan University, Oslo, Norway

  • Clare Shelley-Egan ,

    Contributed equally to this work with: Svenn-Erik Mamelund, Clare Shelley-Egan, Ole Rogeberg

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Resources, Validation, Writing – original draft, Writing – review & editing

    Affiliation Work Research Institute, Oslo Metropolitan University, Oslo, Norway

  • Ole Rogeberg

    Contributed equally to this work with: Svenn-Erik Mamelund, Clare Shelley-Egan, Ole Rogeberg

    Roles Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Frisch Centre, University of Oslo, Oslo, Norway



The objective of this study is to document whether and to what extent there is an association between socioeconomic status (SES) and disease outcomes in the last five influenza pandemics.

Methods/principle findings

The review included studies published in English, Danish, Norwegian and Swedish. Records were identified through systematic literature searches in six databases. We summarized results narratively and through meta-analytic strategies. Only studies for the 1918 and 2009 pandemics were identified. Of 14 studies on the 2009 pandemic including data on both medical and social risk factors, after controlling for medical risk factors 8 demonstrated independent impact of SES. In the random effect analysis of 46 estimates from 35 studies we found a pooled mean odds ratio of 1.4 (95% CI: 1.2–1.7, p < 0.001), comparing the lowest to the highest SES, but with substantial effect heterogeneity across studies,–reflecting differences in outcome measures and definitions of case and control samples. Analyses by pandemic period (1918 or 2009) and by level of SES measure (individual or ecological) indicated no differences along these dimensions. Studies using healthy controls tended to document that low SES was associated with worse influenza outcome, and studies using infected controls find low SES associated with more severe outcomes. A few studies compared severe outcomes (ICU or death) to hospital admissions but these did not find significant SES associations in any direction. Studies with more unusual comparisons (e.g., pandemic vs seasonal influenza, seasonal influenza vs other patient groups) reported no or negative non-significant associations.


We found that SES was significantly associated with pandemic influenza outcomes with people of lower SES having the highest disease burden in both 1918 and 2009. To prepare for future pandemics, we must consider social vulnerability. The protocol for this study has been registered in PROSPERO (ref. no 87922) and has been published Mamelund et al. (2019).


It used to be believed that pandemic and infectious disease risks are the same for all, irrespective of socioeconomic status (SES). But when 61-year old superstar Madonna shared this belief on Instagram on the 23rd of March 2020, calling COVID-19 “the great equalizer” from a milky bath sprinkled with rose-petals [1], fans and others quickly pointed to the disproportionate pandemic burden and suffering of the poor. Indeed, their criticism is supported by a number of studies showing that certain indigenous people, people of colour, immigrants and the poor have experienced disproportionate harm as a result of COVID-19 as measured by infection rates, hospitalizations, the need for intensive care unit treatment, and death [25].

The idea that outcomes from infectious disease pandemics are socially neutral has a long history among lay people, researchers and policy makers responsible for pandemic preparedness plans. Literature on SES and 1918 influenza outcomes published by social historians between 1970 and 1990 argued that the disease was so highly transmissible that everybody was equally affected [610], pointing to anecdotal evidence such as the president and King of Spain falling ill and the Swedish Prince Erik dying at age 29 [11]. However, these studies used aggregate-level data, were mainly descriptive and did not use multivariate statistical models. Empirical studies appearing from the mid-2000s often reported evidence inconsistent with the socially neutral hypothesis: SES seemed to be linked to exposure, susceptibility and access to care, and SES indicators were statistically associated with mortality [1214]. Although several studies of the 2009 pandemic also found SES associated with various pandemic outcomes [1517], this social inequality in risk is still ignored in international influenza pandemic preparedness plans [18]. Apart from a systematic review and meta-analysis of the 2009 pandemic disease burden in low and low to middle income economies and differences in disease outcomes in that pandemic for ethnic minorities vs non-ethnic minorities [19], a systematic assessment of several influenza pandemics and of the evidence for disparities in pandemic outcomes by individual and/or area-level SES (e.g. education, income, household crowding and quality, unemployment, occupation-based social class, poverty status, share below poverty levels, deprivation indexes etc.) has been lacking.

In this paper, we present the first systematic review and meta-analysis on the association between SES and disease outcomes in the last 5 influenza pandemics. The objective is to document whether and to what extent there is an association between indicators of socioeconomic status (e.g. income, education) and pandemic outcomes (infection, hospitalizations, mortality) in the last five influenza pandemics (1889, 1918, 1957, 1968, 2009). In terms of PICOS criteria, the Population (P) consists of groups defined by socioeconomic status, the intervention (I) or exposure or risk factor is pandemic influenza, the comparison (C) or alternative interventions is not relevant, while the outcomes (O) are morbidity, hospitalization, or death associated with influenza pandemics. All types of study design were considered (S). As described in our pre-registered analysis plan, we hypothesized that the association between SES and pandemic outcomes would increase with outcome severity, as higher income and SES tend to be associated with access to resources and protective factors that reduce the risk of progression to more severe outcomes.

Materials and methods

Bibliographic database search

A systematic search of Medline, Embase, Cinahl, SocIndex, Scopus and Web of Science was performed to identify all relevant articles published on SES and pandemic influenza (morbidity, severe disease and mortality). SES was captured by keywords such as education, income, occupational social class etc. (see search strategy, S1 Table, for more examples). Morbidity was captured by keywords such as infection rates, transmission rates, lab confirmed influenza, flu like illness, and influenza like illness (ILI). Severe disease was captured by keywords such as disease severity, critical illness, critical disease, severe illness, severe disease, hospitalization, patient admission, hospital admission, intensive care unit (ICU) admission, and ICU treatment. Mortality was captured by keywords such as fatal outcome, fatal illness, fatal disease, fatality, lethal outcome, lethal illness, lethal disease, terminal outcome, terminal illness, terminal disease, lethality, death, death rate, and mortality rate. All of these keywords were used in both pilots and the final searches. The strategy for the literature search was developed by two information specialists in cooperation with the research group, starting 5 October 2017. Several pilot searches were conducted in Web of Science and Medline respectively, on 12 and 19 October 2017, to ensure a sensitive search. The search strategy combined relevant terms, both controlled vocabulary terms (i.e. MeSH) and text words. The main search strategy used in Medline is available in PROSPERO 87922 and in S1 Table. The final search was carried out on 17 November 2017. The strategy was modified to fit the other databases listed above. To generate manageable results, restrictions on language (English, Danish, Norwegian and Swedish) and publication type (article/research article) were added to the searches in the other databases. The searches in Medline and Embase were performed without publication type restrictions. The search strategy was peer-reviewed by a third information specialist using a structured tool based on the PRESS-framework [20]. Reference lists of relevant known studies were also screened and experts in the field consulted in order to identify other additional sources. Finally, we also contacted authors of published studies to ask for relevant data not presented in the papers or in appendices. However, we did not get any responses to these requests.

Inclusion criteria for title and abstract screening

After adding all identified records to an Endnote library and removing duplicates, the remaining results were imported to the program Covidence. Here, additional duplicates were removed. The title and abstract of each article was screened by two of the authors (SEM and CSE), according to the selection criteria. After screening of titles and abstracts, we added full-text versions of articles in Covidence. Divergences in the inclusion of studies were re-assessed by the same researchers until consensus was reached in terms of inclusion or exclusion. The criteria for inclusion were:

  1. The study period 1889–2009 includes the five pandemics in 1889, 1918, 1957, 1968 and 2009
  2. Studies investigating the association between SES and pandemic outcomes
  3. Studies of race, ethnicity, and indigenous people that reported data on SES controls
  4. Studies addressing both seasonal and pandemic influenza distinguishing between non-pandemic and pandemic years
  5. Studies addressing all regions/countries, type of studies (interventional, observational, etc.) and populations (age, gender, pregnant women, soldiers etc.)

Exclusion criteria for title and abstract screening

The following criteria excluded studies from the systematic literature review:

  1. Studies on pandemic diseases other than influenza
  2. Studies on seasonal influenza only
  3. Studies on both seasonal and pandemic influenza that did not distinguish between non-pandemic and pandemic years
  4. Studies on attitudes and compliance with (non)pharmaceutical pandemic interventions
  5. Qualitative studies on the associations between SES and pandemic outcomes
  6. Studies on social justice and pandemic influenza
  7. Studies of pandemic influenza preparedness plans
  8. Studies of race, ethnicity, and indigenous people that did not report data on SES controls

Data selection and extraction

We drafted a data abstraction form, pilot-tested it and modified it, when necessary. Two reviewers (SEM and CSE) independently extracted data from all included studies. Any disagreements were resolved via discussion or by involving a third reviewer for arbitration. 1–5 and 6 below were entered into separate spreadsheets for each article. The following information was extracted:

  1. Article info
    1. First author
    2. Year published
    3. Journal
  2. Data sample
    1. Country or region of analysis
    2. Pandemic years (1889, 1918, 1957, 1968, 2009)
    3. Sample inclusion criteria–i.e. characteristics of sample/population (civilian, military, gender, pregnant, age-group/median/average age, patient group etc).
    4. Sample size
    5. Unit of analysis (individuals, households, regions, hospitals etc)
    6. Data aggregation level (observations of individual units, aggregated units, etc.). e.g., if hospitals are the unit of analysis, does the data used occur at the hospital level or is it pooled across hospitals?
    7. Source of outcome data, e.g., census, routine notification data (e.g. influenza cases reported to a doctor), survey data, register data
      1. If survey or population data had incomplete coverage
        1. Response rate/coverage
        2. Representativity: Is the sample shown to be representative for the population? i.e. has a non-response analysis been carried out?
  3. Outcome variable—Pandemic outcome (a. morbidity, b. hospitalization, c. mortality)
    1. Definition of morbidity: influenza-like illness (ILI), Lab-confirmed Infection rates (PCR), transmission rates (reproduction number, R0), immunity/antibodies towards influenza (HI titer above a certain threshold) due to exposure to the disease and not vaccination
    2. Definition of hospitalization; Hospitalized inpatients with (PCR) or without confirmed influenza; patients admitted to intensive care unit (ICU) or not; mechanically ventilated patients (“lung machines”) or not; inpatients vs outpatients
    3. Definition of cause of mortality: Influenza and pneumonia (PI), excess mortality (PI, all causes of death etc.), respiratory diseases, pneumonia etc.
  4. Baseline outcomes (control type), i.e. what was the control group or baseline outcome comparison? (general healthy population, infected patients, the hospitalized, patients with lab-confirmed seasonal influenza)
  5. Independent variables of interest–relating to SES
    1. Type of SES indices (education, income, crowding, density, deprivation index, unemployment, occupational social class, poverty status, % below poverty level)
    2. Definition or brief descriptive text on SES indices (e.g., if based on a specific type of poverty index etc.)
  6. Statistical methodology
    1. Design of study (cross sectional, longitudinal, case-control, cohort studies)
    2. Estimation technique (Cross tables, correlation analysis, OLS, Poisson regression, Logistic regression, Cox regressions, GEE regressions, GLMM models etc.)
    3. Control variables included (e.g. age, gender, marital status, pre-existing disease, health behavior etc.) in light of sample restrictions (e.g. for pregnant women, sex is not among the controls)
    4. Reference categories with which all point estimates are compared
  7. Results reported (separate spreadsheet)

Data synthesis.

Our narrative review includes a table of the study characteristics of the included studies, such as study authors and year, pandemic years, study region (region/country/hospital), sample inclusion criteria, sample size, unit of outcomes, data aggregation level, data sources and type, outcomes, baseline outcomes, SES measure, design, statistical techniques, controls, whether the study estimates are used in the meta analysis, and whether SES is an independent predictor. The quantitative part of the study pools results across individual studies using meta-analytic methods.

In the simplest meta-analytic model (“fixed effect”) random sampling variation is assumed to be the only source of variation in estimates. This is implausible in our context, in which studies use different SES indicators and medical outcomes from different countries and time periods. A “random effect” model captures the resulting effect heterogeneity by estimating the distribution of these underlying associations. Systematic variation across study-level covariates (e.g., pandemic period, region, type of outcome) is assessed using sub-sample analyses as well as a Bayesian hierarchical model.

We searched the identified studies in our meta-review for quantitative estimates of associations between SES indicators and influenza related outcomes. The resulting estimates were assessed for inclusion in the meta-analysis, and included if they could be expressed as an odds-ratio or relative risk for low versus high socioeconomic status. This implied that estimates had to include an indicator of socioeconomic status at the individual or ecological level, and had to allow for an estimate of how the incidence or prevalence of some flu related outcome varied by levels of this indicator. Where studies included estimates for distinct data subsamples (different age groups, periods), single estimates pooling all data were preferred if available. If not, the separate estimates were all included. For some studies, multiple estimates were also extracted if they performed different comparisons (e.g., risk of infection, and risk of hospitalization given infection). We also collected study level factors indicating the pandemic period (1918 vs 2009), country/region, and data as to whether the study estimate involved an odds ratio or a relative risk or rate. The specific studies included and all judgments and adjustments concerning inclusion and adjustments of reported numbers are detailed in the S2 File.

We have been inspired by NOS [21] to assess the quality of the included studies. We have rated the following items: A) Selection of exposed population (Broad and representative sample/population of the exposed?, truly and somewhat = 1, selected sub-groups and no description of population = 0) and non-exposed populations (1 = drawn from same community as the exposed, 0 = drawn from different source or no description); B) Comparability (confounder controls (yes = 1, no = 0), biological controls (yes = 1, no = 0), SES measure significant beyond biological controls (yes = 1, no = 0); C) Data quality (Lab-confirmed outcomes (yes = 1, no = 0) and data aggregation level (individual = 1, aggregate = 0)); The average and median quality score in the 44 studies included in the narrative review is respectively 4.5 and 4, while min score is 2 and max 7. The quality scores were higher in the 35 studies included in the meta-analysis (average 4.9, median 5, min 3, max 7) than those 9 not included (average 3.1, median 3, min 2, max 4) (see the quality assessment scores per article in the S2 Table).

Relative to the pre-analysis plan, the ambitions of the quantitative analysis were scaled back given the large heterogeneity across the studies included (see Table 1). The pre-analysis plan specified three types of analysis [22]. The first, a standard random effect analysis with subsample analyses, was conducted as planned using the «REML» algorithm in the Metafor meta-analytic package for R [23]. The second, a PET-PEESE analysis testing and adjusting for publication bias, was found unsuitable given the large effect heterogeneity [24]. The third, a Bayesian model to assess “dose-response” effects and assess how estimates vary with study-level indicators and the type of comparisons made, is included in a simplified version without the “dose-response” element.

Table 1. Overview of 44 studies included in the systematic review by study characteristics.

The Bayesian model differs from the standard fixed and random effect models in two ways. It allows us to include study-level covariates to capture systematically different effects in specific regions, periods or for specific outcomes, using a hierarchical specification across the parameters to impose partial pooling and reduce the risk of large but spurious estimates [25]. If the evidence as a whole indicates that estimates vary no more across study level indicators than we would expect due to sampling variation, then this will pull the individual indicator coefficients towards zero.

Second, the Bayesian model requires a prior distribution for each model parameter that expresses reasonable (pre-analysis) beliefs regarding the parameter values. The estimation updates these beliefs in light of the data, resulting in a posterior distribution that blends the pre-existing knowledge encoded in the prior distribution with evidence from the observed data. To verify that the prior choices for the overall pooled effect and heterogeneity do not unduly influence the result, the Bayesian model is also estimated without study-level covariates to allow for comparison with the standard random effect model.


Narrative review

Flow of included studies.

Our database search identified 8,411 records. After leaving out duplicates, 4,203 studies were imported for screening. After removing another 75 duplicates, we screened the titles/abstracts of 4,128 records. Of these, 3952 studies were irrelevant, and 176 full text studies were then assessed for eligibility. In this phase, 117 studies were excluded, leaving us with 59 studies from which to extract data. In the data extraction phase, we removed an additional 15 studies. The final number of studies included in the narrative synthesis was therefore 44 (see PRISMA Flow Chart in S1 File).

Study characteristics.

The review identified a total of 44 studies, 9 studies of “Spanish flu of 1918–20” [1214, 2631] and 35 of the “Swine flu of 2009–2010” [15, 16, 3264] (Table 1). We identified no studies of the Russian flu of 1889–90, the Asian flu of 1957–58 or Hong-Kong flu of 1968–70. Most of the studies used data from North America, including 11 for USA [16, 27, 30, 33, 40, 41, 48, 50, 55, 56, 59] and 6 for Canada [38, 45, 49, 52, 60, 62]; Europe, including 6 for England [15, 26, 31, 32, 44, 64], 4 for Spain [39, 46, 51, 57], 2 for Norway [12, 13], and 1 for 30 EU/EFTA countries [53]; 4 for Australia [42, 43, 54, 61] and 3 for New Zealand [28, 29, 34]. While a few studies used data from Central America/South America including 1 for Mexico [37] and 1 for Brazil [47], and Asia, including 1 for Iran [35] and 1 for China [63], we identified no studies using data from Africa. Finally, 3 studies had a global approach studying several countries [14, 36, 58].

The sample inclusion criteria varied greatly from study to study. Two of the 44 studies studied military populations, one of these studied mortality in randomly selected records [28], the other studied mortality on one transport troop ship [29]. Of the 42 studies using civilian study populations, some studied particular patient populations/cohorts [46, 54, 61, 63], general patients at various hospitals and health centres [16, 32, 33, 35, 39, 40, 4749, 51, 52, 55, 57, 59, 60, 62], students at schools or students including their families [41, 45], or general populations living in various cities, states, counties or (several) countries [1215, 26, 30, 31, 34, 3638, 4244, 50, 53, 56, 58, 64].

The sample size in each study varied substantially and is reported in Table 1 whenever information was available for the pandemic events (for cases and controls) and the population at risk.

The unit of the outcome variables was either individual in 36 studies [13, 15, 16, 2740, 4249, 51, 52, 54, 55, 57, 5964] or aggregate in 8 studies [12, 14, 26, 41, 50, 53, 56, 58]. Some of the studies with individual-level outcome data nevertheless preformed analysis at aggregated levels. In 12 studies the data aggregation level was aggregate [12, 14, 15, 26, 30, 31, 36, 41, 50, 53, 56, 58]. 15 studies had individual-level outcome variables and control variables, but used area-level (and individual-level) SES variables [16, 27, 32, 33, 4244, 49, 52, 54, 55, 59, 61, 62, 64]. Studies using only ecological SES variables thus picked up a combination of individual-level and area-level SES effects on the outcome variables. Finally, in 17 of the studies, outcomes, explanatory variables and controls were all measured for individuals and the data aggregation level was thus the individual level [13, 28, 29, 34, 35, 3740, 4548, 51, 57, 60, 63].

There were generally three types of data source used in the 44 studies included in the narrative synthesis: 1) 28 studies used active surveillance of events coupled with SES and covariate data via questionnaires, face-to-face or telephone interviews or censuses [16, 3244, 4652, 54, 57, 5963]; 2) 14 studies used national vital registration systems on events coupled with SES and covariate data via censuses [1215, 2629, 31, 53, 55, 56, 58, 64]; 3) 2 studies used telephone survey or data collected via door-to-door survey to collect both event and population at risk data [30, 45].

The 3 broad categories of outcomes were studied (see details in Table 1): 1) people seeing doctors due to symptoms of influenza like illness (ILI)/influenza transmission(R0)/lab-confirmed influenza infection (using PCR tests)/immunity towards influenza (using blood serum samples to look for antibodies) [26, 27, 30, 32, 34, 35, 4145, 52, 54, 57, 60, 61, 63]; 2) lab-confirmed influenza hospitalizations/ICU treatment/mechanical ventilation [16, 33, 38, 39, 4651, 55, 57, 59, 62]; 3) lab-confirmed pandemic deaths/Influenza-Pneumonia (PI) deaths/excess deaths associated with pandemic influenza [1215, 2631, 36, 37, 40, 53, 56, 58, 59, 64].

The choice of baseline outcomes (or controls in case-control studies) partly depended on the outcomes studied, and included: 1) General population at risk [1216, 2633, 36, 50, 53, 56, 5860, 64]; 2) General population at risk without H1N1 Infection or ILI [4145]; 3) Patients with ILI, persons in quarantine for a suspected case and a close H1N1 contact or patients with ILI testing negative for influenza A H1N1 infection [30, 35, 52, 63]; 4) pre-pandemic immunity [34, 61]; 5) seasonal influenza A deaths [37]; 6) Non-hospitalized H1N1 positive patients or hospitalized H1N1 positive non-severe (not ICU or death) [38, 39, 55, 59, 62]; 7) Outpatients with H1N1 infection [40, 4649, 51, 57]; 8) Seronegative for H1N1 [54]; 9) Patients with other diseases than ILI [57].

The studies that used individual-level SES measures used one or several of the following; (household) income [40, 48, 60], economic status [30], education [35, 3740, 4649, 51, 52, 60, 63], occupation-based social class [13, 28, 29, 57], size of apartments, poor housing or crowding measures [13, 26, 34, 40, 45, 49, 51], and having health insurance [40]. Some used both individual-level and area-level measures of SES. The SES measures used at the area-level were often (but not always) indexes of economic, social and housing deprivation/development [12, 1416, 27, 3133, 36, 4144, 4850, 5256, 58, 59, 61, 62, 64].

The 44 studies included in the review used study designs that fall into four categories: 1) Systematic review and meta-analysis [36]; 2) Cross sectional univariate or control-variable design [12, 1416, 2634, 4245, 50, 5356, 59, 60, 64]; 2) Case-control design [35, 3741, 4649, 51, 52, 57, 6163]; 3) Longitudinal survival analysis [13]; 4) Time-series analysis [58].

The identified studies were descriptive or explanatory. The descriptive studies used statistical techniques to calculate pandemic disease burden estimates and univariate correlations between the outcomes and various variables as well as demographic standardization techniques to control for age and sex [15, 16, 28, 3033, 44, 56, 60]. The explanatory multivariate studies used modelling techniques such as OLS [12, 14, 50, 58], generalized linear mixed models [45], logistic regressions [26, 29, 34, 35, 3841, 4649, 52, 54, 55, 57, 59, 6163], propensity score logistic regressions [37], Poisson regressions [27, 53, 59, 64], Cox regressions [13, 51], random effect meta-regressions [36], and various types of Bayesian models [42, 43].

Study results.

The results in the 9 identified studies on the 1918 influenza and SES were mixed [1214, 2631]. After various controls were made, 6 studies found a significantly and expected higher mortality for lower SES groups [14] or higher mortality/transmission rates, but not a clear social gradient for all SES measures [12, 13, 27]; a significant higher mortality for lower SES groups, but only for 2 out of 3 pandemic waves [31]; or a significantly higher morbidity and mortality for lower SES groups [30], while 3 studies found no association between SES and mortality [28, 29] or mortality and transmission rates [26]. However, none of the 6 studies documenting significant associations with a higher pandemic disease burden for lower SES groups included data to control for medical risk factors. Hence, some or all of the identified associations between SES and the pandemic outcomes in the 6 above mentioned studies could potentially have been “explained away” by controlling for having latent tuberculosis [65] or other known comorbidities [66].

Fourteen of the 35 identified studies on the 2009 pandemic had data to adjust for both medical and social risk factors [34, 35, 38, 40, 4548, 51, 52, 54, 57, 59, 62]. After adjusting for medical risk factors, 7 of these studies documented independent and expected impact of SES (higher risks for lower SES) on either infection/immunity [34, 54], hospitalization [4648, 51] or both of these outcomes [57]; 1 study found both expected significant associations with SES (higher risk of hospitalization) and non-significant (ICU and death) impact of SES after medical risk factor were controlled for [59]; 5 studies found non-significant effects of SES on ILI/infection/immunity [45, 52], hospitalization/ICU [38, 62] and mortality [40]; and finally, 1 study found a significant but unexpected impact of SES on infection, that is higher infection rates for those with higher vs. lower education [35]. Although the findings in these 14 studies investigating both social and medical vulnerabilities were somewhat mixed, they show that medical risk factors are not simply 100% correlated with socioeconomic factors, and in 8 of these 14 studies social factors explained variation in the variation in the pandemic outcomes beyond that captured by medical factors.

21 of the 35 identified studies on the role of SES in the 2009 pandemic outcomes did not control for medical risk factors but found the following. First, 12 studies found significantly higher risks for the lowest socioeconomic status group, of which 5 studied ILI/infection/immunity [32, 43, 44, 63]; 4 investigated hospitalizations [16, 33, 39, 49]; and 4 studied mortality [15, 36, 56, 64]. Second, 7 studies found non-significant associations with SES, of which 2 studied ILI/infection/immunity [42, 60]; 2 studied hospitalizations [50, 62] or ICU treatment [62], and 3 studied mortality [37, 53, 58]. Finally, 2 studies unexpectedly found respectively a higher risk of a lab-confirmed case [41] or ICU treatment [55] in the highest SES groups. It is clear though, that most of the studies on SES and 2009 pandemic not controlling for medical at risk factors [13 of 21], showed that lower SES groups have the highest risks of the three considered pandemic outcomes.

Quantitative synthesis

The quantitative synthesis includes 46 estimates drawn from 35 of the 44 studies included in the narrative synthesis [1316, 2730, 3235, 3741, 4549, 5157, 5964], and a standard random effects analysis of all estimates pooled found a pooled effect mean odds ratio of 1.4 (95% CI: 1.2–1.7), comparing the low to the high SES groups. The pooled estimate was statistically significant at the 0.1 percent level, which means that we would have been highly unlikely to see an estimate of this or larger absolute magnitude if the true mean of the effect distribution was zero. As seen in the forest plot, the individual study estimates differed in both precision and location, with more variation in less precise estimates as we would expect (Fig 1). To test for the presence of publication bias, we used Egger’s test and Begg’s test as implemented in the regtest and ranktest commands of the Metafor R-package [23].As imprecise estimates have to be larger to be statistically significant, publication bias will tend to show up as a systematic relationship between point estimates and their standard errors. Neither a rank-rank correlation test (Begg’s test, p-value 0.68) nor a regression test (Egger’s test, p-value 0.81) indicated any such relationship.

Fig 1. Forest plot.

The plot shows the included estimates sorted by precision, along with their weights in the pooled effect estimate.

In the random effect analysis we found strong evidence of effect heterogeneity across studies, with an estimated 92% of the total variation across studies reflecting effect differences rather than sampling variation. The estimated standard deviation of the effect distribution (tau) has a point estimate of 0.45 on the log scale. If the underlying effects at the study level are normally distributed around their expectation, this tau is the estimated standard deviation of study effects. The estimates would then imply that there is a 50% chance that the true parameter value of a randomly selected study lies in the range of 1.1–1.9. The Cochran’s Q test strongly rejects a test of zero heterogeneity (p < 0.0001), confirming the choice of a random effects over a fixed effect model. Our subsample analyses indicate similar results in studies using individual level and aggregate SES indicators, case control and relative risk outcome measures, and studying the 1918 and 2009 pandemic period (Fig 2 and Table 2).

Fig 2. Subsample analyses.

The plot shows point estimates and 95% confidence intervals for different subsamples of studies, with a grey circle indicating the number of studies in each subsample.

Subsamples were also defined by specific combinations of case and control outcomes (Fig 3). These suggest that studies examining the risk of flu outcomes relative to a general healthy population (here defined as a control sample not selected on indicators of illness) tend to indicate a clear and substantial increased risk for lower SES groups. Studies comparing hospitalized to those infected also point to increased risks for lower SES groups. Studies assessing the risk of severe cases (e.g., treatment in ICU or death) conditional on hospitalization are fewer, but seem to report no clear SES associations in any direction. Finally, studies using “other” control samples (e.g., patients with flu symptoms who did not have flu, people with non-pandemic flu during a pandemic period, patients accessing or being treated by health care systems for other reasons) tend to find no (or reversed) associations with SES indicators.

Fig 3. Subsample analyses.

The plot shows point estimates and 95% confidence intervals for different subsamples of studies, with a grey circle indicating the number of studies in each subsample.

As all of these comparisons were based on different splits of the same study sample, they can be viewed as a series of univariate analyses. To assess the joint contribution of these study level features, and to include country/region indicators, we estimated two Bayesian models: One, without study level covariates, was closely analogous to the above meta-analysis, and was included to ensure that results from the two approaches were similar and comparable. This Bayesian model finds a pooled effect mean of 1.4 with a 95% credibility interval from 1.2–1.7, which is identical to the above estimate of 1.4 (95% CI: 1.2–1.7). The estimated standard deviation of the underlying study parameters, analogous to the parameter tau in the earlier analysis, is estimated at 0.46 (0.3–0.6), the same as the above estimated tau of 0.45 (See S2 File for model code and discussion of prior choices). The second Bayesian model includes all study level indicators (level of SES indicator, RR/OR indicator, period, case and control outcomes, and country/region), as well as an indicator for each unique combination of case and control outcome (as in Fig 3). Jointly, this reduces the estimated unexplained heterogeneity (tau) substantially, with the average value estimated dropping from 0.46 to 0.34 (See S2 File for model code and discussion of prior choices).

As shown in Figs 4 and 5, the Bayesian analysis finds similar results as the earlier subsample analyses, indicating that the patterns for the control and treatment outcome combinations are not “explained away” in an analysis when simultaneously accounting for other study level characteristics.

Fig 4. Differences across study level covariates.

The plot shows average estimates and 95% credibility intervals for different study level covariates. The parameters are constrained to sum to zero within each category (e.g., for each draw from the posterior distribution, the sum of country parameters will sum to zero, as will the sum of the period parameters, etc.) See S2 File for model details.

Fig 5. Differences across case and control outcome combinations.

The plot shows average estimates and 95% credibility intervals for all combinations of case and control outcomes observed in the data. See S2 File for model details.


Research on Covid-19 has shown that the disease burden differs by SES, race and ethnicity [25]. This is consistent with the results we report here from the first systematic literature review on the associations between SES and disease outcomes in the last 5 influenza pandemics. We identified nine studies of the “Spanish flu of 1918–20” and 35 of the “Swine flu of 2009–2010”, but no studies of the “Russian flu” pandemic of 1889–90, the “Asian flu” of 1957–58 or the “Hong-Kong flu” of 1968–70. Most of the studies included for the 1918 and 2009 influenza pandemics used data from western high-income countries. Out of 51 estimates from 35 studies, the overall pooled mean pandemic outcome odds ratio was 1.44 (95% CI: 1.23–1.68) comparing the lowest to the highest SES groups. As expected, the random effect model estimated substantial effect heterogeneity across studies, which means the pooled effect mean should not be taken as valid at the single-study level. Based on the model, we would expect about 50% of the underlying effects to be in the 1.1 to 1.9 range. There was no evidence suggesting differences by pandemic period (1918 or 2009), the level of SES measure (individual or ecological), or type of method (odds ratio or relative risk). Finally, studies using healthy controls tended to find low SES associated with worse influenza outcome, and studies using infected controls found low SES associated with more severe influenza outcomes. Studies comparing severe outcomes (ICU or death) to hospital admissions were few but indicated no clear association. Studies with more unusual comparisons (e.g., pandemic vs seasonal influenza, seasonal influenza vs other patient groups) reported no or negative associations. These patterns were similar in a multivariate Bayesian model accounting for all study level indicators simultaneously. The Bayesian model also included indicators for study region/country. Relative to the “across all country/regions” average, studies from Australia, UK and to a lesser extent the USA tended to report stronger associations in our sample, while New Zealand tended to report weaker associations. These country-level results should be viewed as exploratory: two of the three studies from New Zealand [28, 29], for instance, were studies of how pandemic influenza outcomes varied across pre-service occupational status amongst military personnel during the 1918 pandemic, which are unlikely to speak broadly to such associations in New Zealand more generally.

Our results provide strong evidence that social risk factors matter for pandemic influenza outcomes in addition to medical risk factors. We also documented that in the 2009 pandemic, social risk factors independently explained variation in disease outcomes even when medical risk factors were controlled for [34, 4648, 51, 54, 57, 59]. This is similar to the finding of a study of COVID-19 hospital deaths demonstrating that medical risk factors did little to explain the higher risks of the deprived and of immigrants in the UK [3]. Although we did not find support for our hypothesis that social disparities would be larger for more severe (e.g. ICU and death) than less severe outcomes (e.g. infection or hospitalization not requiring ICU), the similarity of results for the 1918 and 2009 pandemics show the persistence of individual- and ecological-level social risk factors, although the specific mechanisms and types of social vulnerabilities leading to social disparities in pandemic outcomes may differ between 1918 and 2009, or in 2020 during the COVID-19 pandemic. Results from this review on pandemic influenza and results from studies on the role of social and ethnic vulnerability in COVID-19 disease outcomes [25], support recent calls for the inclusion of social and ethnic vulnerabilities in addition to medical at risk factors in pandemic preparedness plans [18]. Examples are the prioritization of vaccines for medically vulnerable people living in socially vulnerable areas (urban slums or hard-to-reach groups in rural and remote areas), or SES groups with undiscovered medical vulnerabilities, and others who are at significantly higher risk of severe disease or death (various indigenous, ethnic, or racial groups, people living in extreme poverty, homeless and those living in informal settlements or; low-income migrant workers; refugees, internally displaced persons, asylum seekers, populations in conflict settings or those affected by humanitarian emergencies, vulnerable migrants in irregular situations and nomadic populations).

The studies reporting on social inequalities in influenza outcomes in 1918 and in 2009, identified in this review, and also early research on social disparities in COVID-19 outcomes, often lacked a discussion of the possible mechanisms for the estimated social disparities, a framework to discuss those mechanisms and/or the data to separate the distal (social and policy) and proximal (behavioral and biological factors) factors for unequal exposure, susceptibility and access to health care leading to socially unequal pandemic outcomes [68]. Socially unequal exposure may relate to hand washing behavior or mask use, cleaning of surfaces, cramped living conditions, multigenerational living, occupational exposure, ability to work from home or stay away from work to care for family members, and use of public transportation. Social disparities in susceptibility may relate to poor nutritional status or, concurrent illnesses (e.g. NCDs). Finally, socioeconomic inequalities in understanding of or access to health advice (e.g. hand hygiene, social distancing, travel advisories) and vaccination or other public recommendations due to poor reading and writing skills may also explain part of the variation in outcomes by SES [13, 18].

Two of the studies on the 2009 pandemic included in our review, on Iran [35] and USA [55], reported increased risks (of infection rates in Iran and ICU stay vs. hospitalized non-ICU patients in USA) for those with high socioeconomic status–contrary to the authors’ and our hypothesis. For the US study, the authors suggest that this may reflect social gradients in testing and demand for treatment and health care resources. In one matched case–control study of mortality among hospitalized H1N1 patients vs. H1N1 outpatients, no differences in any of the SES variables were found when controlling for health seeking behavior and barriers to health care access. However, it is not clear whether these or other controls (age, sex, race, urban-rural, vaccination status, health behaviors, pre-existing conditions) “explained away” the negative associations of having health care insurance or the positive association of poverty in the univariate models. If the studies with data generated from health care systems that found higher pandemic risks for lower SES groups had controlled for a social gradient in testing and demand for treatment and health care resources, we expect that the findings of a social pattern in disease burden would be reinforced.

An important strength of our study is the use of a pre-registered study protocol for data gathering and analysis, which was peer-reviewed and published prior to the gathering of study data [22]. This helped ensure that the process was specified in a reproducible way and followed a rigorous and systematic workflow to identify studies and describe and analyze results. The engagement of professional information specialists to design, test and improve the literature search strategies that were applied to a broad range of literature databases is particularly important, given the lack of any previous systematic reviews on this topic with which our list of included studies could be compared.

Our study also has some potential limitations. First, we carried out our library search on 17 November 2017, and potential studies published in the two pre-COVID-19 years of 2018–19 and during the COVID-19 pandemic (2020–21) are not included. Given the massive research on COVID-19 pandemic in 2020–21 and the fact that the identified 2009 pandemic studies in our review were published rapidly after 2009–10, with 2012 being the average publication year, it is likely that we have missed a few 2009 pandemic studies. Given the strength and consistency of the results, we do not expect that newer studies would alter our general conclusions, at least not for the 2009 pandemic that was the topic of 35 of the 44 included studies. Second, we would note that the generalizability of our results is necessarily limited by the geographic focus of the research we synthesize: no studies using data from Africa were found, and few from Asia and South America. It is therefore reasonable to ask whether our results are representative outside of high-income countries in North America, Europe and the Oceania region.


We have shown that influenza pandemic outcomes in 1918 and 2009 were associated with lower socioeconomic status and that pandemic outcomes in 2009 were not always socially neutral «great equalizers» once adjusting for medical risk factors [34, 4648, 51, 54, 57, 59]. This resembles the finding of a study of COVID-19 hospital deaths demonstrating that medical risk factors did little to explain the higher risks of the deprived and of immigrants in the UK [69]. The social lessons from historical influenza pandemics such as those in 1918 or 2009 have not yet been taken into account in influenza pandemic preparedness [18], and this blind spot has also been evident in the response to the COVID-19 pandemic. Such social and ethnic vulnerability factors should be explicitly included and addressed in current and future plans and responses in order to more effectively reduce pandemic burdens, reduce social disparities and ameliorate the social consequences of future pandemics [70]. The global health and economic crisis created by the COVID-19 pandemic has made us only too aware of the need for a more holistic and comprehensive approach towards pandemic preparedness.

Supporting information

S2 File. Specific studies included and all judgments and adjustments concerning inclusion and adjustments of reported numbers.



We are indebted to our librarians Bettina Grødem Knutsen, Ingjerd Legreid Ødemark and Elisabeth Karlsen at the Learning Center and Library, Oslo Metropolitan University. Without their expertise and assistance in doing library searches this research would never have been accomplished.


  1. 1. Smith E. Madonna calls coronavirus ‘the great equalizer’ in bizarre bathtub video. Page Six. 2020 22 March.
  2. 2. Steyn N, Binny R, Hannah K, Hendy S, James A, Kukutai T, et al. Estimated inequities in COVID-19 infection fatality rates by ethnicity for Aotearoa New Zealand. New Zealand Medical Journal. 2020;133(1520). pmid:32994635
  3. 3. Williamson EJ, Walker AJ, Bhaskaran K, Bacon S, Bates C, Morton CE, et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature. 2020;584(7821):430–6. pmid:32640463
  4. 4. Drefahl S, Wallace M, Mussino E, Aradhya S, Kolk M, Brandén M, et al. A population-based cohort study of socio-demographic risk factors for COVID-19 deaths in Sweden. Nature Communications. 2020;11(1):5097. pmid:33037218
  5. 5. Liu SH, Liu B, Li Y, Norbury A. Time courses of COVID-19 infection and local variation in socioeconomic and health disparities in England. medRxiv. 2020:2020.05.29.20116921.
  6. 6. Rice G. Black November. The 1918 influenza epidemic in New Zealand. Wellington, New Zealand: Allen and Unwin; 1988.
  7. 7. Phillips H. Black October: The impact of the Spanish influenza epidemic of 1918 on South Africa: University of Cape Town; 1984.
  8. 8. Tomkins SM. The Failure of Expertise: Public Health Policy in Britain during the 1918–19 Influenza Epidemic. Social History of Medicine. 1992;5(3):435–54. pmid:11623115
  9. 9. Van Hartesveldt FR. The 1918–1919 pandemic of influenza: the urban impact in the western world. New York: Edwin Mellen Press; 1992.
  10. 10. Crosby A. Epidemic and peace, 1918. Westport, Connecticut: Greenwood Press; 1976.
  11. 11. Mamelund SE. Profiling a Pandemic. Who were the victims of the Spanish flu? Natural History Magazine. 2017(September):6–10.
  12. 12. Mamelund S-E. Spanish Influenza Mortality of Ethnic Minorities in Norway 1918–1919. European Journal of Population / Revue européenne de Démographie. 2003;19(1):83–102.
  13. 13. Mamelund S-E. A socially neutral disease? Individual social class, household wealth and mortality from Spanish influenza in two socially contrasting parishes in Kristiania 1918–19. Social Science & Medicine. 2006;62(4):923–40. pmid:16084634
  14. 14. Murray CJ, Lopez AD, Chin B, Feehan D, Hill KH. Estimation of potential global pandemic influenza mortality on the basis of vital registry data from the 1918–20 pandemic: a quantitative analysis. The Lancet. 2007;368(9554):2211–8.
  15. 15. Rutter PD, Mytton OT, Mak M, Donaldson LJ. Socio-economic disparities in mortality due to pandemic influenza in England. International Journal Of Public Health. 2012;57(4):745–50. pmid:22297400
  16. 16. Tam K, Yousey-Hindes K, Hadler JL. Influenza-related hospitalization of adults associated with low census tract socioeconomic status and female sex in New Haven County, Connecticut, 2007–2011. Influenza and Other Respiratory Viruses. 2014;8(3):274–81. pmid:24382111
  17. 17. Chandrasekhar R, Sloan C, Mitchel E, Ndi D, Alden N, Thomas A, et al. Social determinants of influenza hospitalization in the United States. Influenza & Other Respiratory Viruses. 2017;05:05. pmid:28872776
  18. 18. Mamelund SE. Social Inequality–a Forgotten Factor in Pandemic Influenza Preparedness. Journal of the Norwegian Medical Association. 2017(12–13):911–3. pmid:28655260
  19. 19. Tricco AC, Lillie E, Soobiah C, Perrier L, Straus SE. Impact of H1N1 on Socially Disadvantaged Populations: Systematic Review. PLoS ONE. 2012;7(6):1–17. pmid:22761796
  20. 20. McGowan J, Sampson M, Salzwedel DM, Cogo E, Foerster V, Lefebvre C. PRESS Peer Review of Electronic Search Strategies: 2015 Guideline Statement. Journal of Clinical Epidemiology. 2016;75:40–6. pmid:27005575
  21. 21. Wells GA, Shea B, Peterson J, Welch V, Losos M, P T. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses.
  22. 22. Mamelund S-E, Shelley-Egan C, Rogeberg O. The association between socioeconomic status and pandemic influenza: protocol for a systematic review and meta-analysis. Systematic Reviews. 2019;8(1):5. pmid:30609940
  23. 23. Viechtbauer W. Conducting Meta-Analyses in R with the metafor Package. 2010. 2010;36(3):48.
  24. 24. Alinaghi N, Reed WR. Meta-analysis and publication bias: How well does the FAT-PET-PEESE procedure work? Res Synth Methods. 2018;9(2):285–311. pmid:29532634
  25. 25. Gelman A, Hill J, Yajima M. Why We (Usually) Don’t Have to Worry About Multiple Comparisons. Journal of Research on Educational Effectiveness. 2012;5(2):189–211.
  26. 26. Chowell G, Bettencourt LM, Johnson N, Alonso WJ, Viboud C. The 1918–1919 influenza pandemic in England and Wales: spatial patterns in transmissibility and mortality impact. Proceedings of the Royal Society of London—Series B: Biological Sciences. 2008;275(1634):501–9. pmid:18156123
  27. 27. Grantz KH, Cummings DAT, Glass GE, Rane MS, Salje H, Schachterle SE. Disparities in influenza mortality and transmission related to sociodemographic factors within Chicago in the pandemic of 1918. Proceedings of the National Academy of Sciences of the United States of America. 2016;113(48):13839–44. pmid:27872284
  28. 28. Summers JA, Shanks GD, Baker MG, Wilson N. Severe impact of the 1918–19 pandemic influenza in a national military force. The New Zealand Medical Journal. 2013;126(1378):36–47. pmid:24045314
  29. 29. Summers JA, Wilson N, Baker MG, Shanks GD. Mortality risk factors for pandemic influenza on New Zealand troop ship, 1918. Emerg Infect Dis. 2010;16(12):1931–7. pmid:21122224
  30. 30. Sydenstricker E. The Incidence of Influenza among Persons of Different Economic Status during the Epidemic of 1918. Public Health Reports (1896–1970). 1931;46(4):154–70.
  31. 31. Pearce DC, Pallaghy PK, McCaw JM, McVernon J, Mathews JD. Understanding mortality in the 1918–1919 influenza pandemic in England and Wales. Influenza & Other Respiratory Viruses. 2011;5(2):89–98. pmid:21306572
  32. 32. Balasegaram S, Ogilvie F, Glasswell A, Anderson C, Cleary V, Turbitt D, et al. Patterns of early transmission of pandemic influenza in London—link with deprivation. Influenza & Other Respiratory Viruses. 2012;6(3):e35–41. pmid:22236079
  33. 33. Balter S, Gupta LS, Lim S, Fu J, Perlman SE. Pandemic (H1N1) 2009 surveillance for severe illness and response, New York, New York, USA, April-July 2009. Emerging Infectious Diseases. 2010;16(8):1259–64. pmid:20678320
  34. 34. Bandaranayake D, Huang QS, Bissielo A, Wood T, Mackereth G, Baker MG, et al. Risk factors and immunity in a nationally representative population following the 2009 influenza A(H1N1) pandemic. PLoS ONE [Electronic Resource]. 2010;5(10):e13211. pmid:20976224
  35. 35. Cheraghi Z, Irani AD, Rezaiean S, Ahmadzadeh J, Poorolajal J, Erfani H, et al. Influenza A (H1N1) in Hamedan Province, Western Iran in 2009: A case-control study. Journal of Research in Health Sciences. 2010;10(1):15–21. pmid:22911912
  36. 36. Duggal A, Pinto R, Rubenfeld G, Fowler RA. Global Variability in Reported Mortality for Critical Illness during the 2009–10 Influenza A(H1N1) Pandemic: A Systematic Review and Meta-Regression to Guide Reporting of Outcomes during Disease Outbreaks. PLoS ONE. 2016;11(5):1–14. pmid:27170999
  37. 37. Fajardo-Dolci G, Gutierrez JP, Arboleya-Casanova H, Garcia-Saiso S. Comparing Deaths from Influenza H1N1 and Seasonal Influenza A:Main Sociodemographic and Clinical Differences between the Most Prevalent 2009 Viruses. Influenza Research & Treatment. 2012:1–5. pmid:23346393
  38. 38. Gilca R, de Serres G, Boulianne N, Ouhoummane N, Papenburg J, Douville-Fradet M, et al. Risk factors for hospitalization and severe outcomes of 2009 pandemic H1N1 influenza in Quebec, Canada. Influenza and other Respiratory Viruses. 2011;5(4):247–55. pmid:21651735
  39. 39. González-Candelas F, Astray J, Alonso J, Castro A, Cantón R, Galán JC, et al. Sociodemographic Factors and Clinical Conditions Associated to Hospitalization in Influenza A (H1N1) 2009 Virus Infected Patients in Spain, 2009–2010. PLoS ONE. 2012;7(3):1–8.
  40. 40. Hennessy TW, Bruden D, Castrodale L, Komatsu K, Erhart LM, Thompson D, et al. A case-control study of risk factors for death from 2009 pandemic influenza A(H1N1): is American Indian racial status an independent risk factor? Epidemiology And Infection. 2016;144(2):315–24. pmid:26118767
  41. 41. Hoen AG, Buckeridge DL, Chan EH, Freifeld CC, Keller M, Charland K, et al. Characteristics of US public schools with reported cases of novel influenza A (H1N1). International Journal of Infectious Diseases. 2010;14 Suppl 3:e6-8.
  42. 42. Hu W, Williams G, Phung H, Birrell F, Tong S, Mengersen K, et al. Did socio-ecological factors drive the spatiotemporal patterns of pandemic influenza A (H1N1)? Environment International. 2012;45(Supplement C):39–43. pmid:22572115
  43. 43. Huang X, Clements AC, Williams G, Mengersen K, Tong S, Hu W. Bayesian estimation of the dynamics of pandemic (H1N1) 2009 influenza transmission in Queensland: A space-time SIR-based model. Environmental Research. 2016;146:308–14. pmid:26799511
  44. 44. Inglis NJ, Bagnall H, Janmohamed K, Suleman S, Awofisayo A, De Souza V, et al. Measuring the effect of influenza A(H1N1)pdm09: the epidemiological experience in the West Midlands, England during the ’containment’ phase. Epidemiology & Infection. 2014;142(2):428–37.
  45. 45. Janjua NZ, Skowronski DM, Hottes TS, Osei W, Adams E, Petric M, et al. Transmission dynamics and risk factors for pandemic H1N1-related illness: outbreak investigation in a rural community of British Columbia, Canada. Influenza & Other Respiratory Viruses. 2012;6(3):e54–e62.
  46. 46. Launes C, García-García JJ, Martínez-Planas A, Moraga F, Astigarraga I, Arístegui J, et al. 2009 H1N1: risk factors for hospitalization in a matched case-control study. European Journal of Pediatrics. 2012;171(7):1127–31. pmid:22430351
  47. 47. Lenzi L, Mello AM, Silva LR, Grochocki MH, Pontarolo R. Pandemic influenza A (H1N1) 2009: risk factors for hospitalization. Jornal Brasileiro De Pneumologia: Publicacao Oficial Da Sociedade Brasileira De Pneumologia E Tisilogia. 2012;38(1):57–65. pmid:22407041
  48. 48. Levy NS, Nguyen TQ, Westheimer E, Layton M. Disparities in the Severity of Influenza Illness: A Descriptive Study of Hospitalized and Nonhospitalized Novel H1N1 Influenza-Positive Patients in New York City: 2009–2010 Influenza Season. Journal of Public Health Management & Practice. 2013;19(1):16–24. pmid:23169399
  49. 49. Lowcock EC, Rosella LC, Foisy J, McGeer A, Crowcroft N. The Social Determinants of Health and Pandemic H1N1 2009 Influenza Severity. American Journal of Public Health. 2012;102(8):e51–e8. pmid:22698024
  50. 50. Maliszewski PJ, Wei R. Ecological factors associated with pandemic influenza A (H1N1) hospitalization rates in California, USA: a geospatial analysis. Geospatial Health. 2011;6(1):95–105. pmid:22109867
  51. 51. Mayoral JM, Alonso J, Garín O, Herrador Z, Astray J, Baricot M, et al. Social factors related to the clinical severity of influenza cases in Spain during the A (H1N1) 2009 virus pandemic. BMC Public Health. 2013;13(1):1–7. pmid:23391376
  52. 52. Navaranjan D, Rosella LC, Kwong JC, Campitelli M, Crowcroft N. Ethnic disparities in acquiring 2009 pandemic H1N1 influenza: a case–control study. BMC Public Health. 2014;14(1):1–17. pmid:24580862
  53. 53. Nikolopoulos G, Bagos P, Lytras T, Bonovas S. An Ecological Study of the Determinants of Differences in 2009 Pandemic Influenza Mortality Rates between Countries in Europe. PLoS ONE. 2011;6(5):1–8. pmid:21589928
  54. 54. Pasco JA, Nicholson GC, Brennan SL, Bennett KE, Dobbins AG, Athan E. The epidemiology of the first wave of H1N1 influenza pandemic in Australia: A population-based study. Open Public Health Journal. 2012;5:80–5.
  55. 55. Placzek H, Madoff L. Effect of Race/Ethnicity and Socioeconomic Status on Pandemic H1N1-Related Outcomes in Massachusetts. American Journal of Public Health. 2014;104(1):e31–e8. pmid:24228651
  56. 56. Ponnambalam L, Samavedham L, Lee HR, Ho CS. Understanding the socioeconomic heterogeneity in healthcare in US counties: the effect of population density, education and poverty on H1N1 pandemic mortality. Epidemiology And Infection. 2012;140(5):803–13. pmid:21835071
  57. 57. Pujol J, Godoy P, Soldevila N, Castilla J, González-Candelas F, Mayoral JM, et al. Social class based on occupation is associated with hospitalization for A(H1N1)pdm09 infection. Comparison between hospitalized and ambulatory cases. Epidemiology And Infection. 2016;144(4):732–40. pmid:26271901
  58. 58. Simonsen L, Spreeuwenberg P, Lustig R, Taylor RJ, Fleming DM, Kroneman M, et al. Global mortality estimates for the 2009 Influenza Pandemic from the GLaMOR project: a modeling study. PLoS Medicine. 2013;10:e1001558–1277. pmid:24302890
  59. 59. Thompson DL, Jungk J, Hancock E, Smelser C, Landen M, Nichols M, et al. Risk Factors for 2009 Pandemic Influenza A (H1N1)-Related Hospitalization and Death Among Racial/Ethnic Groups in New Mexico. American Journal of Public Health. 2011;101(9):1776–84. pmid:21778495
  60. 60. Thompson LH, Mahmud SM, Keynan Y, Blanchard JF, Slater J, Dawood M, et al. Serological survey of the novel influenza A H1N1 in inner city Winnipeg, Manitoba, 2009. The Canadian Journal of Infectious Diseases & Medical Microbiology. 2012;23(2):65–70.
  61. 61. Trauer JM, Laurie KL, McDonnell J, Kelso A, Markey PG. Differential effects of pandemic (H1N1) 2009 on remote and indigenous groups, Northern Territory, Australia, 2009. Emerging Infectious Diseases. 2011;17(9):1615–23. pmid:21888786
  62. 62. Zarychanski R, Stuart TL, Kumar A, Doucette S, Elliott L, Kettner J, et al. Correlates of severe disease in patients with 2009 pandemic influenza (H1N1) virus infection. CMAJ Canadian Medical Association Journal. 2010;182(3):257–64. pmid:20093297
  63. 63. Zhang Y, Seale H, Yang P, MacIntyre CR, Blackwell B, Tang S, et al. Factors associated with the transmission of pandemic (H1N1) 2009 among hospital healthcare workers in Beijing, China. Influenza & Other Respiratory Viruses. 2013;7(3):466–71. pmid:23078163
  64. 64. Zhao H, Harris RJ, Ellis J, Pebody RG. Ethnicity, deprivation and mortality due to 2009 pandemic influenza A(H1N1) in England during the 2009/2010 pandemic and the first post-pandemic season. Epidemiology And Infection. 2015;143(16):3375–83. pmid:25850904
  65. 65. Mamelund S-E, Dimka J. Tuberculosis as a Risk Factor for 1918 Influenza Pandemic Outcomes. Tropical Medicine and Infectious Disease. 2019;4(2):74. pmid:31035651
  66. 66. Dimka J, Mamelund S-E. 1918 Influenza Outcomes among Institutionalized Norwegian Populations: Implications for Disability-Inclusive Pandemic Preparedness. Scandinavian Journal of Disability Research. 2020;22(1):175–86.
  67. 67. Galobardes B, Shaw M, Lawlor DA, Lynch JW. Indicators of socioeconomic position (part 1). Journal of Epidemiology and Community Health. 2006;60(1):7–12. pmid:16361448
  68. 68. Quinn SC, Kumar S. Health inequalities and infectious disease epidemics: a challenge for global health security. Biosecurity And Bioterrorism: Biodefense Strategy, Practice, And Science. 2014;12(5):263–73.
  69. 69. Williamson E, Walker AJ, Bhaskaran KJ, Bacon S, Bates C, Morton CE, et al. OpenSAFELY: factors associated with COVID-19-related hospital death in the linked electronic health records of 17 million adult NHS patients. medRxiv. 2020:2020.05.06.20092999.
  70. 70. Schmidt H, Pathak P, Sönmez T, Ünver MU. Covid-19: how to prioritize worse-off populations in allocating safe and effective vaccines. BMJ. 2020;371:m3795. pmid:33020072