Table 1. Overview of 44 studies included in the systematic review by study characteristics.
| * | Study authors and year | Study region | Pandemic period | Sample inclusion criteria | Sample size | Unit of outcomes | Data aggregation level | Data sources | Outcomes | Baseline outcomes | SES measures | Design | Statistical technique | Controls | Estimates used in meta analysis and is SES an independent predictor? |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2 | [32] | London, England | 20 April- 28 June 2009 | People of all ages seeing a doctor for influenza at hospitals and community clinics in London | 2,819 H1N1 patients (confirmed, presumed and probable) with valid LSOA postcodes | Individuals | Individual cases, but SES of cases based on the IMD of area post-codes | Data on cases and contacts were from the London Flu Response Center database and where coupled to IMD 2007 | Influenza cases per 100,000 | Population at risk in each LSOA area | Area Index of multiple Deprivation (IMD) 2007 quintiles (economic, social and housing issues) | Cross-sectional univariate design | Bivariate rate ratios with 95% CI | Age and weekly interactions with IMD | Meta analysis: Yes (all ages and whole period) SES measure significant |
| 3 | [33] | New York, USA | 24 April-7 July 2009 | Active hospitalized-based surveillance and passive collection of on demographics, risk conditions, and clinical severity | 996 H1N1 patients (929 Confirmed and 67 probable) | Individuals | Individual cases, but SES of cases based on United Hospital Fund Poverty neighborhoods | Active hospitalized-based surveillance and passive collection of on demographic, risk conditions, and clinical severity | Hospitalizations per 100,000 | Population at risk in high, medium or low poverty areas | Tertiles of percentage of residents living <200% of the federal poverty level according to the 2000 US Census | Cross-sectional univariate design | Bivariate Rate ratios with 95% CI | Age | Meta analysis: Yes SES measure significant |
| 4 | [34] | New Zealand | Nov 2009- March 2010 | Randomly selected serum samples from GPs countrywide and in the Auckland region 3 months after the pandemic | 1,687 serum samples | Individuals | Individual observations | seroprevalence data coupled with questionnaires evaluating demographics and potential risk factors. | H1N1 Infection rates (Seroprevalence; Antibody titer >1:40) | Baseline immunity was measured from 521 sera collected during 2004 to April-2009 | Damp housing (poor housing conditions is an often used measure of SES, see [67]) | Multi-stage random cross-sectional design | Multivariate logistic regressions | Age, ethnicity, gender, vaccination history, chronic illness | Meta analysis: Yes SES measure ns. |
| 6 | [35] | Eight cities in Hamedan Province, western Iran | July-December 2009 | Subjects (cases and controls) were selected from patients with signs and symptoms of respiratory tract infection who were referred to health centers | 245 cases and 388 controls | Individuals | Individual observations | Data are from health centers on H1N1 infection status coupled with covariate data from interviewers using predetermined questionnaires | Cases were identified by pharyngeal soap specimens positive for influenza A virus using PCR | Controls were testing negative for influenza A virus using PCR | Education 1. low education: illiterate, primary school and middle school. 2. High education: high school and academic | Unmatched case-control study | Multivariate logistic regressions | Age, sex, pregnancy, suspected close contact with influenza patients, smoking, region (urban rural), trip during last week, chronic disease, influenza vaccination, and BMI | Meta analysis: Yes SES measure significant, but unexpectedly higher risk for the high education group. |
| 7 | [26] | England & Wales | 12 Oct 1918–5 April 1919 | Influenza deaths in all parts of E & W | - | Aggregate: 305 adm. units & 62 counties | Aggregate | Deaths from National vital registration systems and demographic data from the 1921 census | Influenza death rates and reproduction number R (the average number of secondary cases generated by an index case) | Population at risk | People per acres, dwellings and rooms | Cross- Cross-sectional control-variable design | Spearman correlations, using a Bonferroni correction for multiple comparisons (transmissibility and death rates) and multivariate logistic regressions (death rates) | Population size, fall and winter waves, urban-rural | Meta analysis: No There were no association between transmissibility, death rates and indicators of population density and residential crowding |
| 10 | [36] | Global (226 studies from 50 countries met the inclusion criteria) | 2009–2010 | Described confirmed, probable or suspected cases of 2009–2010 influenza A (H1N1) infection; and (2) described patient(s) who were critically ill | 10695 | Individuals | Aggregate, Global | Medline, Embase, LiLACs and African Index Medicus to June 2009-March 2016 | Mortality associated with H1N1-related critical illness | Population at risk | World Bank economic development status of countries (High, upper middle, lower middle income) | Systematic review and meta analysis | Random effects meta regressions | No controls | Meta analysis: No SES measure significant |
| 11 | [37] | Mexico | 10 April to 13 July 2009 | Data from clinical files from all influenza A deaths | 239 H1N1 cases and 85 influenza A controls | Individuals | Individual observations | Patients’ clinical records and reporting forms from health facilities | Lab-confirmed A/H1N1 deaths (rt-PCR-test) | Seasonal influenza A deaths | Education (Primary school or less, Junior high school, High school or higher level) | Case-control | Propensity score multivariate logistic regressions | Sex, age, have a partner, smoking, employment status | Meta analysis: Yes SES measure ns. |
| 12 | [38] | Canada (Quebec) | 16 April-1 July 2009 | Lab-confirmed H1N1 hospitalizations or ICU admission/ deaths | 321 hospitalized incl. 47 ICU and 15 deaths (cases) and 395 non-hospitalized N1H1 infection patients (controls) | Individuals | Individual observations | Suspected H1N1 case at primary care clinics or hospital coupled with other data from standardized questionaries’ | Lab-confirmed influenza associated hospitalizations (24 hrs or more) and ICU/death | Non-hospitalized H1N1 patients (vs. hospitalized) or hospitalized non-severe (vs. ICU/death) | Education (high school not competed, non-University certificate, university degree) | Case-control | Multivariate logistic regressions | Age, sex, HCW, smoking, flu jab in 2008–09, consultation, days after onset, antiviral use, pregnancy, underlying condition, obesity | Meta analysis: Yes (both outcomes included) SES measure ns. |
| 13 | [39] | Spain (Andalusia, Basque Country, Catalonia, Castile and Leon, Madrid, Navarra and Valencia) | July 2009-Febr. 2010 | Lab-confirmed hospitalization (RT-PCR) | 699 hospitalized and 703 non-hospitalized cases of a(H1N1) infection | Individuals | Individual observations | Data from 36 hospitals and primary care centers in 7 spanish regions | Lab-confirmed hospitalizations (patient admitted to hospital for > 24 hours with RT-PCR confirmed H1N1 infection) | Non-hospitalized people with RT-PCR confirmed infection with the same pandemic virus | Education Secondary or higher | Case-control | Multivariate logistic regressions | Age, sex, ethnic group | Meta analysis: Yes SES measure significant. However, data on underlying health collected but not controlled for |
| 14 | [27] | USA (Chicago) | 29 Sep-16 Nov 1918 | Influenza and pneumonia (PI) mortality | 7971 PI deaths | Individuals | Individual deaths, but SES measured at the level of 496 Census tracts | Historical maps of point-level mortality incidence, spatial data and near contemporaneous census data | Influenza and pneumonia mortality and reproduction number (R0) | Population at risk | Census tract-based SES (% illiteracy, unemployment, homeownership, population density) | Cross-sectional control-variable design | Poisson regressions with GEE and Spearman correlations | Age | Meta analysis: Yes % illiterate sig. predictor of mortality controlling for age and all other SES variables. Sig. ass btw. R0 and population density, illiteracy, and unemployment but not homeownership. |
| 15 | [40] | USA (Alaska, Arizona, New Mexico, Oklahoma, Wyoming) | 15 April 2009–31 Jan 2010 | Lab-confirmed A (H1N1) fatalities; state residents who died relating to infection with lab-confirmed influenza A | 145 fatal cases and 236 controls | Individuals | Individual observations | Medical records (notifiable disease reports), death certificates, interviews with cases and controls | Lab-confirmed A(H1N1) fatalities using RT-PCR test | Outpatients with lab-confirmed H1N1 | Healthcare insurance, >1,5 persons per room, graduated high school, poverty (<US$ 25000/year) | Matched case-control | Logistic regressions | Age, sex, race, barriers to health care access, urban-rural, health seeking behavior, vaccination status, health behaviors, pre-existing conditions. | Meta analysis: Yes (poverty) None of the SES variables were significant. |
| 16 | [41] | USA (23 counties) | 23 April-8 June 2009 | English language media reports of A (H1N1) cases | 32 public primary & secondary schools with at least one confirmed H1N1 case and 6815 control schools located in the same 23 counties as the case schools | Aggregate, Schools | Aggregate | Health Map | Media reports of A (H1N1) cases | Schools located in the same 23 counties as the case schools without N1N1 cases | Title 1 school (Whether or not schools qualifies for a federal funding to support economically disadvantaged students. | Matched case-control | Logistic regression | Highest grade at school and size | Meta analysis: Yes SES measure significant |
| 17 | [42] | Australia (Brisbane) | Jan-Dec 2009 | Lab-confirmed daily A (H1N1) cases | 11,979 cases | Individuals | Individual cases, but SES measured for postcode areas (SLA) | Queensland Health, SEIFA data from Australian Bureau of Statistics (ABS) & daily rainfall & temperature data from the Australian Bureau of Meteorology | Lab-confirmed daily A (H1N1) cases | Rest of the population with no lab-confirmed case | SEIFA: socioeconomic index for areas, incl. education, occupation and wealth | Cross-sectional control-variable design | Bayesian spatial conditional autoregressive poisson models | Rainfall (mm) and temperature (degrees Celsius) | Meta analysis: No SES measure ns. |
| 18 | [43] | Australia (Queensland) | 7 May-31 Dec 2009 | Lab-confirmed A (H1N1) cases | - | Individuals | Individual cases, but SES measured for postcode areas (SLA) | Queensland Health, SEIFA data from Australian Bureau of Statistics (ABS) & daily rainfall & temperature data from the Australian Bureau of Meteorology | Lab-confirmed daily A (H1N1) cases | Rest of the population with no lab-confirmed case | SEIFA: socioeconomic index for areas, incl. education, occupation and wealth | Cross-sectional control-variable design | Flexible Bayesian, space-time. SIR models | Rainfall (mm) and temperature (degrees Celsius) | Meta analysis: No SES measure significant |
| 19 | [44] | England (West Midlands) | 16 April-6 July 2009 | Lab-confirmed A (H1N1) cases | 3063 cases | Individuals | Individual cases, but SES measured for postcode areas | FluZone, a national surveillance database with case reporting. SES data from IMD 2007 | Lab-confirmed A (H1N1) cases | Rest of the population with no lab-confirmed case | Index of Multiple Deprivation of an area and postcodes (IMD 2007). It includes seven dimensions: income, employment, health deprivation and disability, skills and training, barriers to housing and services, crime and disorder, living environment SES indexes IMD 2007: Index of Multiple Deprivation | Cross-sectional | Descriptive analysis | Age, sex, ethnicity, exposure and illness severity, but no controls were made | Meta analysis: No SES measure significant |
| 21 | [45] | Canada (Rural community of British Columbia; local town and surrounding First Nation reserves | Late April/early May 2009 | One elementary school and on-reserve aboriginal participants; | 83 ILI cases and 281 non-ILI cases | Individuals | Individual observations | Phone survey of households with at least one child enrolled in any of the community schools | Influenza-like illness (ILI) | Non-ILI cases | Household density | Cross-sectional control-variable design | Generalised linear mixed models (GLMM) | Age, chronic conditions, aboriginal status, received vaccination 2008–09 | Meta analysis: Yes SES measure ns. |
| 23 | [46] | Spain (Andalusia, the Basque Country, Castile and Leon, Catalonia, Madrid, Navarre, and Valencia | July 2009- Feb. 2010 | Patients aged 6 months to 18 years with confirmed H1N1 at 32 Hospitals of the Spanish National Health survey | 195 confirmed H1N1 hospitalized cases and 184 outpatient controls with confirmed H1N1 | Individuals | Individual observations | Spanish National Health Service | Lab-confirmed A (H1N1) inpatient (hospitalized) cases | Outpatient (non-hospitalized) controls with confirmed H1N1 | Parents education (Primary or lower vs. secondary or higher) | Matched case control, prospective, observational study | Logistic regressions | Age, pulmonary, disease, neurological disease, diabetes mellitus, cardiovascular disease, and non-Caucasian ethnicity | Meta analysis: Yes SES measure significant |
| 24 | [47] | Brazil (Paraná) | 2009 | Patients (in- and outpatients) with lab-confirmed H1N1 infection | 1911 Inpatient cases and 2829 outpatients controlls | Individuals | Individual observations | Brazilian Ministry of Health National Case Registry Database | Lab-confirmed A (H1N1) inpatient cases and outpatient controlls | Lab-confirmed H1N1 outpatients controlls | Level of education (Literate vs. illiterate) | Retrospective observational case-control study | Logistic regressions | age, gender, ethnicity, having a comorbiditiy, number of comorbiditis, 8 types of underlying health conditions, smoking, clinical manifestations, treatment (Oseltamivir), time to treatment initiation in days | Meta analysis: Yes SES measure significant |
| 26 | [48] | USA (New York) | 1 Oct 2009–28 Feb 2010 | Lab-confirmed illness among adults and children | 128 inpatients with lab-confirmed flu cases matched by age and month of diagnosis with 246 non-hospitalized lab-confirmed influenza A controls (assumed to be H1N1) | Individuals | Individual observations | Sentinel surveillance system used by NYC Department of Health and Mental Hygiene; telephone interview to collect clinical and demographic data | Lab-confirmed A (H1N1) inpatient cases and outpatient influenza A controls | Non-hospitalized lab-confirmed influenza A controls (assumed to be H1N1) | Education (Some college or more, not a high school graduate, high school graduate), annual household income and neighbourhood poverty (% Persons living below the federal poverty line) | 1:2 case-control study design, matching by age group and month of diagnosis | Conditional multivariate logistic regressions | Access to care (primary physician, insurance) and at least one underlying condition (various diseases, pregnancy and obesity) | Meta analysis: Yes Education among adults and neighbourhood poverty among children and adults were significant |
| 28 | [49] | Canada (Ontario) | Two waves in 2009 (April 23-July 20 and August 1 Nov 6) | Residents of all ages who received nasopharyngeal swabs and tested positive for H1N1 | 401 self-reported hospitalization cases and 624 non-hospitalized controls (150 hospitalized and 184 non-hospitalized in wave 1, 251 hospitalized and 440 non-hospitalized in wave 2) | Individuals | Individual hospitalizations by individual-level education and contextual level SES variables | Surveillance data and standardised phone interviews | Lab-confirmed A (H1N1) inpatients (hospitalized patients) | Non-hospitalized controls H1N1 positives | individual level education level (of adult participants aged 18 years or older & of parents respondents for children younger than 16 years), household density (individuals per sleeping rooms) and several contextual level SES variables (employment, education, income, social and material deprivation) | Case-control study | Binomial or multinomial logistic regression, using generalized estimating equations to account for clustering/dependence in the data | Age and gender | Meta analysis: Yes (Total deprivation and individual and parental education for both waves). First wave: High school education or less and living in a neighborhood with high material or total deprivation sign. Second wave: High school education or less sign. Moreover, a mediation analysis showed that clinical risk factors explain only a portion of the ass. btw SES & hospitalization. |
| 29 | [50] | USA (California) | 3 April-15 Sep 2009 | Reported counts of H1N1 hospitalizations, not lab-confirmed | 2010 hospitalizations | 58 counties | Aggregate | California Department of Public Health surveillance data | Reported H1N1 Hospitalizations | Population at risk in each 58 counties | Education (% of persons aged > 25 years with a high school diploma); Poverty (% of pop under poverty line); Income (median HH income in dollars) | Cross-sectional control-variable design | OLS | Sex, race/ethnicity, age, climate, agricultural and transportation variables | Meta analysis: No The 3 SES variables were ns. but results not shown |
| 30 | [12] | Norway | 1918–1919 | PI deaths covering the whole of Norway | 16,005 deaths | Aggregate, 351 medical districts | Aggregate | Regional district physician reports and census data | PI mortality reported to a doctor | Population at risk | % receiving public support due to poverty; Wealth per person (in 1000 Nok); Average number of persons per room | Cross-sectional control-variable design | OLS | age, sex, ethnicity, % in fishing, coast-inland, summer wave exposure | Meta analysis: No Poverty and wealth, but not crowding was sign. |
| 31 | [13] | Norway (Frogner and Grønland/Wexels parishes in Oslo) | 1 Feb 1918–1 feb 1919 | PI deaths in the two selected parishes | 250 PI deaths | Individuals | Individuals | Death certificates coupled with census data | PI mortality reported on death certificates | Population at risk | Occupational based social class, apartment size (1–8 rooms +) and parish | Longitudinal multivariate survival analysis | Cox regressions | Age, sex, marital status | Meta analysis: Yes (occupation based social class) Apartment size and parish but not occupation-based social class was sign. |
| 34 | [51] | Spain (Andalusia, the Basque Country, Castile and Leon, Catalonia, Madrid, Navarre, and Valencia) | July 2009-Feb 2010 | Patients recruited from hospitals & primary health care clinics & emergency units during the peak of the influenza A 2009 pandemic in | 699 hospitalized and 699 non-hospitalized with Lab-confirmed cases A(H1N1) cases using (RT-PCR) | Individuals | Individuals | Cases filled in a questionaries’ at the health centre or by phone to obtain covariate information | Hospitalized lab-confirmed A (H1N1) cases | Non-hospitalized (family physician visits at primary health care clinics and emergency units) cases of A(H1N1) infection | Education (Secondary or higher vs no formal education or primary education) and overcrowding (below the fifth percentile of the distribution of square metres available per person in the normal residence of all study participants) | Multicenter Matched case-control (according to age, date of hospitalisation in of the case (+/- 21 days) & province of the residence of the case) | Binomial logistic regression using Cox conditional logistic regressions | Sex, ethnicity, prior preventive information, prior pandemic vaccination, previous outpatient care or emergency care and unfavourable medical factors (smoking, morbid obesity (BMI >40), hypertension, lung disease, cardiovascular disease, kidney failure, diabetes, chronic liver disease, immunodeficiency, disabling neurological disease, malignancy, transplantation, cognitive dysfunction, seizure disorders and rheumatic diseases) | Meta analysis: Yes (education) Education decreases & Overcrowding increases outcome significantly |
| 35 | [14] | Global study covering 27 countries with high-quality vital registration data for the 1918–1920 pandemic | 1918–20 | Data for populations where vital registrations are believed to be more than 80% complete, supplemented with subnational data for US states & provinces of "pre-partition" India | 27 countries for 1918–1920, 24 US states with data available for the period, and nine Indian provinces | Countries and states | Aggregate | Human mortality database, B.R. Mitchels International Historical Statistics Series, subnational data from US states and provinces of prepartition India | Excess mortality by comparison of annual death rates during the pandemic to the average of annual death rates before and after the pandemic | Population at risk | Income (Per-head income in real international dollars (corrected for price changes) | Cross-sectional control-variable design | OLS with log of pandemic mortality and log income and absolute value of latitude | Latitude, to control for diurnal temperature fluctuation | Meta analysis: Yes Log per-head income in 1918 sign. |
| 36 | [52] | Canada (Ontario) | 13 April-20 July 2009 | Residents (children and adults) tested for A(H1N1) using RT-PCR | 240 cases and 112 controls among children (< 18 years) and 173 cases and 229 controls among adults (>18 years) | Individuals | Individuals H1N1 status by individual education and several ecological SES variables | Clinic-based sample from Ontario, individuals presented to clinics for medical care + standardised telephone interviews | Lab-confirmed 2009 pandemic cases | RT-PCR negative H1N1 cases | Individual Education (high school or less and post-secondary school completion) Area measures: Material, social, total, low employment rate, low income. | Test-negative case-control study | Logistic regressions | age, gender, bmi, ethnicity, current smoker, underlying medical conditions, household density, children in household, receipt of 2008 seasonal vaccine, tested prior to 11th June 2009, healthcare provider, Toronto residence, immigrant category | Meta analysis: Yes (Total deprivation, one for adults and one for children). None of the SES variables were sign. in univariate models and were therefore not entered in the multivariate models. |
| 37 | [53] | Europa (30 EU/EFTA countries) | May 2009-May 2010 | Confirmed and notified fatal pandemic influenza A(H1N1) deaths in EU/EFTA region | 2896 fatal cases | Aggregate, Countries | Aggregate | ECDC and Eurostat | Lab- confirmed and notified deaths | Population at risk | GDP per capita | Cross-sectional control-variable design | Random effect Poisson regressions | greenhouse gas emissions, concertation of particular matter, latitude, hospital beds per 100,000 inhabitants, per capita government expenditure on health, unmet need for medical examination/treatment, Gini coefficient, employment rate, proportion of population aged 65+, old age dependency ratio, women per 100 men | Meta analysis: Yes GDP per capita was sign. in univariate model, but not in multivariate model. |
| 38 | [54] | Australia (Barwon statistical division in Southeastern Australia) | Sep 2009-May 2010 | Adult subjects in Geelong Osteoporosis Study, a group randomly selected from electoral rolls, were invited to participate in this sub-study to provide blood samples and complete a questionnaire. Sample of seropositive adults prior to the availability of a vaccine | 1184 individuals (129 seropositives and 1055 seronegatives) | Individuals | Individual seropostive status by ecological SES variables | Blood samples and self-report questionnaire | Haem agglutination inhibition test, seroposotivity was defined as a titre > 1:40 | Seronegative persons | Australian Bureau of Statistics’ Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD) Area-level measure of education, occupation, income, unemployment and household structure (quintiles 1–5) | Cross-sectional control-variable design | Multivariate logistic regressions | age, bmi, obese, current smoker, healthcare worker, childcare worker/teacher, employment status, highest level of education, lives alone, lives with children aged <12 years, chronic respiratory disease, pregnancy, chronic heart disease, diabetes | Meta analysis: Yes The SES variable was significant in multivariate models |
| 39 | [31] | England and Wales (62 of 82 counties) | Week ending 29 June 1918 to 10 May 1919 | Counties with SES info from 2000 which could be linked to counties in 1918 | Sample covers 333 units and 62 out of 82 counties | Individual deaths | Aggregate | Weekly influenza deaths & annualised rates/1000 population, collated by the Registrar General’s Office in 1920 | Influenza mortality | Population at risk | The average of Ward Scores from the Indices of Deprivation 2000: District level Presentations for England It combines a number of indicators which cover a range of domains (Income, Employment, Health Deprivation and Disability, Education, Skills and Training, Housing and Geographical Access to Services) into a single deprivation score for each area. | Cross-sectional control-variable design | non-parametric Spearman correlation coefficient | Pre-pandemic mortality, age, population size (persons/acre) | Meta analysis: No SES measure sign. in waves 1 and 3, but not wave 2 |
| 40 | [55] | USA (state of Massachusetts) | 26 April-30 Sep 2009 (before the vaccine became available) | Patients met the following inclusion criteria: 1) Patients were discharged from acute care hospital. 2) assigned 1 or more diagnosis codes corresponding to a grouping of ICD-9. 3) younger than 65 years | 4874 hospitalizations of which 526 admitted to ICU | Individuals | Individual hospitalizations, but area-level SES variables | Linked hospital discharge and American Community Survey and US Census data | Lab-confirmed H1N1 ICU stays | Hospitalized non-ICU patients | % of pop below poverty level 2006–2010 for zip code areas | Cross-sectional control-variable design | Logistic regressions | Racial/ethnic groups, gender, age, admission though EP/OP | Meta analysis: Yes Unexpectedly, those in less affluent SES groups had sign. lower risk of ICU stay than the most affluent SES group |
| 41 | [56] | USA (341 US counties in 14 states) | July 2009-June 2010 | Only states with consistent reporting and updating of H1N1 statistics, that is reporting standards met by the CDC | Sample size not given. | Aggregate,341 counties | Aggregate | County-level H1N1 deaths are from CDC and SES variables from US census and CDC 11% of US counties covered, SES measures are representative to similar characteristics to USA as a whole | H1N1 deaths according to CDC | Population at risk | Per capita personal income; median household income; educational attainment (persons aged >/ = to 25 years), percent high school graduate or higher, educational attainment (persons aged >/ = 25 years), percent bachelor’s degree or higher; people of all ages in poverty (%) | Univariate and cross-sectional design | Correlations | No controls | Meta analysis: Yes In univariate models poverty positively predicted mortality while income and education variables negatively predicted mortality. Multivariate modelling was not carried out. |
| 42 | [57] | Spain (Andalusia, the Basque Country, Castile and Leon, Catalonia, Madrid, Navarre, and Valencia) | July 2009-Feb 2011 | Cases and controls were aged > 18 years and picked from 36 hospitals and 22 primary-care centres | 715 primary care centre H1N1 cases, 715 other diseases than ILI primary centre controls, and 406 hospitalized H1N1 cases | Individuals | Individuals | Hospital and primary care data | Lab-confirmed H1N1 cases and hospitalizations (RT-PCR) | Infection model: Controls were primary care patients with other disease than ILI Hospitalization model: cases were primary care centre H1N1 cases | Occupational based social class (Manual vs. non-manual workers) | Matched case-control study | Logistic regressions | In model for infection: age, pregnancy, diabetes and influenza vaccination. In hospitalization model: age, pregnancy, COPD, cardiovascular disease, diabetes, and influenza vaccination | Meta analysis: Yes SES variable sign. in multivariate models for both infection and hospitalization risks |
| 44 | [17] | England | 1 June 2009–18 April 2010 | All deaths reported due to pandemic flu | 349 out of 365 deaths (95,6%) in England | Individual deaths | Aggregate: Individuals aggregated up to five approximately equal population groups to create area deprivation quintiles | National Health Service; basic set of demographic information | Pandemic deaths, no info whether these were lab-confirmed or not, but they were probably lab-confirmed | Population at risk | Index of Multiple Deprivation of an area and postcodes (pooled measure based on income, education, housing, health and crime) (1–5, where 5 is least deprived and 1 most deprived) | Cross-sectional table analysis | Direct age-sex standardization of mortality rates using mid-point 2009 pop estimates for England | Age, sex, and Urban and rural areas | Meta analysis: Yes SES variable significant with and without urban-rural interactions |
| 45 | [58] | Global: 20 countries covering 35% of the world population | 2009 pandemic mortality | Weekly virology and underlying cause-of-death mortality time series for 2005–2009 | 123,000–203,000 deaths in the last 9 months of 2009 | Aggregate | Aggregate | Weekly virology data from the WHO FluNet and national mortality time series | Excess mortality associated with the 2009 pandemic | Population at risk | Gross national income (GNI) per capita (US dollars | Univariate cross-sectional time-series analysis | Multivariate OLS regressions | - | Meta analysis: No. Coefficients not given in the paper or in online appendix Estimates between Gross national income and mortality was ns. |
| 48 | [28] | New Zealand | 27 Aug 1918-March 1919 | Male soldiers (New Zealand Expeditionary Forces (NZEF) in both hemispheres in 1918–1919 pandemic period) | 930 deaths, taken from 1000 randomly selected records | Individuals | Individuals | Death certificates | Influenza, pneumonia, and bronchitis deaths | NZEF population at risk | Pre-enlistment occupational based social class (1–3 (most privileged), 4–6 and 7–9 (least privileged) | Univariate cross-sectional design | Univariate Rate ratios | No controls | Meta analysis: Yes SES measure not significant |
| 49 | [29] | New Zealand | 20 July-13 Oct 1918 | Male navy soldiers (military personnel in HM New Zealand Transport troop ship Tahiti) | 77 deaths, 1117 military personnel plus 100 crew (total pop at risk 1217) | Individuals | Individuals | Death certificates | Influenza and pneumonia deaths | Population at risk at HM New Zealand Transport troop ship Tahiti | Occupation-based social class (1–6 and 7–9 (1 is company manager and 9 is labourer) | Cross-sectional control-variable design | Multivariate logistic regression | age, military rank, rurality score, military unit | Meta analysis: YES SES measure not significant |
| 50 | [30] | USA (New London, Connecticut, Baltimore, Maryland, Augusta, Georgia, Macon, Georgia., Des Moines, Iowa, Louisville, Kentucky, Little Rock, Arkansas, San Antonia, Texas, San Francisco, California | 1 Sep-Dec 1918 | Nine urban localities with a population of at least 25,000, randomly selected, only white populations | 94,678 individuals, 26,824 morbidity cases (influenza, pneumonia and "doubtful" cases), X deaths | Individuals | Aggregate | Survey data (e.g. Baltimore: sample 33,776 (5.68% of pop) | Self-reported pandemic morbidity, mortality and case fatality rates ( | Morbidity: Population at risk in canvassed areas and lethality: mortality among the sick | Economic status (Very poor; poor; moderate; well-to-do (based on the enumerators impression) | Cross-sectional control-variable design | Cross-tables and direct standardization techniques to control for age-differences etc. | age, sex, size of household | Meta analysis: Yes SES measure sign. related with both outcomes. |
| 51 | [16] | USA (New Haven County, Connecticut) | 2009–10 | Hospitalized, laboratory confirmed influenza among adults 18 years and older | 213 hospitalizations | Individuals | Individual lab-confirmed hospitalizations but neighbourhood level SES measures (185 Census tracts) | Surveillance data (Connecticut Emerging Infections Program’s influenza-associated hospitalisation surveillance system) + chart reviews & interviews with healthcare providers & with patients or their proxies. Census tract level data obtained from the US Census Bureau’s 2006–2010 American Community Survey (ACS) | H1N1 lab-based hospitalizations | Population at risk in New Haven | Below federal poverty, no high school diploma, median income | Cross-sectional design | Age-adjusted incidence of influenza-associated hospitalizations among adults by neighbourhood SES characteristics. | Age. | Meta analysis: Yes All 3 SES measures are sign. and display a clear social gradient |
| 52 | [59] | USA (state of New Mexico) | 14 Sep 2009–13 Jan 2010 | Hospitalized, positive influenza hospitalization, Mechanical ventilation and death among the hospitalized | 926 lab-confirmed H1N1 hosp. Patients, 106 mechanically ventilated and 35 deaths | Individuals | Individuals outcomes, but 33 counties divided into 4 quartiles by median household income | New Mexico Department of Health statewide surveillance of hospitalizations and deaths. Estimates from the US Census Bureau’s Small Area Income and Poverty Estimates programme. | H1N1 related hospitalisations, mechanical ventilation and death | Comparison group for hospitalizations: general statewide population; Comparison group for mechanical ventilation and death among those hospitalized were the hospitalized | Household Income (County median household annual income quartile) | Cross-sectional control-variable design | Poisson and logistic regressions | Hospitalization model: age, gender, and race/ethnicity. Mechanical ventilation model: age, gender, and race/ethnicity, obesity, high risk conditions, neuraminidase treatment, time from illness onset to seeking medical care. Mortality risk model: ns in unadjusted model, therefore no multivariate model | Meta analysis: Yes SES measure sign. in model for hospitalization risk but not in models for mechanical ventilation and death |
| 53 | [60] | Canada (Winnipeg, Manitoba) | Oct- Dec 2009 | Adults presenting to three inner city community clinics were recruited as study participants using convenience sampling. | 458 study participants (174 participantsOct-12 Nov, before the vaccine was available), 206 cases 13 Nov-Dec, which did not get take the vaccine; 78 participants enrolled on or after Nov 13 which did get the vaccine are not included in our meta-analysis) | Individuals | Individuals | Serological testing and questionnaire data | Seropositive cases | convenience sample population at risk | Education (High school or not) and annual household income | Univariate & cross-sectional analysis | Prevalence estimates with exact binomial 95% CI using Clopper Pearson intervals | no controls | Meta analysis: Yes The two SES measures ns. for both periods. |
| 55 | [61] | Australia (Northern Territory) | 2009 (June-August) | Antibody titers were determined by hemagglutination inhibition against reference virus A/California/7/2009 on serum samples collected opportunistically from outpatients | 1689 serologic specimen post pandemic (cases 3–30 September 2009) and 445 serological specimen prepandemic (controls January 10 to May 29, 2009) | Individuals | Individual seropositive status but SES measure is aggregate | Serological data, specimens from pathology lab, and computer matching of data to indigenous status and SEIFA measures | lab-confirmed seropositivesand attack rates (difference between post and pre-pandemic immunity) | serological specimen prepandemic (controls January 10 to May 29, 2009) | 2006 Statistical Local area (SLA) was linked to Australian Bureau of Statistics’ Socio-Economic Indexes for Ara (SEIFA). SEIFA measures (quintiles) use information from census data relating to material and social resources and ability to participate in society to obtain a broad level of relative socioeconomic status for each SLA | Case-control design | Logistic regressions | age, gender, aboriginal and Torres strait islanders, region | Meta analysis: Yes SES measure ns. |
| 57 | [62] | Canada (province of Manitoba) | 2 April-5 Sep 2009 | Confirmed H1N1 cases for whom the final location of treatment was known | 795,569 community cases, 181 hospitalized but not ICU, 45 admitted to ICU | Individuals | Individual H1N1 case status, but area income quintiles | Lab-confirmed H1N1 data, hospital data and data collection–form completion via interviews | lab-confirmed community cases, hospitaliations and ICU admissions | Two control groups. Community cases (vs. hospitalizations) and hospitalized, non ICU (vs. ICU). | Income based on postal codes (Top three quintiles vs the bottom two quintiles) | Cumulative case-control design | Logistic regressions | Age, gender, pregnancy, ethnicity, any comorbidity, Interval from symptom onset to antiviral treatment, rural vs urban | Meta analysis: Yes SES measure ns. in models for both hospitalizations and ICU admissions |
| 58 | [63] | China (Beijing) | 1 Aug-30 Sep 2009 | Households of hospital healthcare workers. Case households were: (1) has an index patient of H1N1. (2) index case was quarantined in household from onset of diagnosis to 7 days after onset of illness; (3) secondary case had potential contact with index patient; (4) symptoms onset of secondary case occurred within 7 days since last known contact with index case during infectious period of index case; (5) RT-PCT confirmation date of secondary case occurred within 7 days since last known contact with index case during infectious period of index case; (6) none of the household members previously received a vaccine against pandemic H1N1 2009 influenza | 54 case households (HH with a self-quarantined index patient and a secondary case), 108 control households (HH with a self-quarantined index patient and a close contact) | Individuals | Households | Household transmission data | Lab-confirmed secondary cases (RT-PCT) | Households with a self-quarantined index patient and a close contact | Education (High school and higher vs middle school and lower) | 1:2 matched case-control design | Conditional logistic regression | Sharing room with index case-patient; Ventilating room every day; and Frequency of hand washing | Meta analysis: Yes SES measure significant |
| 59 | [64] | England | 27–30 April 2009 | Lab-confirmed AH1N1 pandemic flu deaths | 337 of 389 lab-confirmed fatalities (86.6%) | Individuals | Individual lab-confirmed deaths, but SES is measured for 32378 super output areas (LSOA) | National Health Service | lab-confirmed deaths | Population at risk | Index of Multiple Deprivation of an area and postcodes (IMD 2007). It includes seven dimensions: income, employment, health deprivation and disability, skills and training, barriers to housing and services, crime and disorder, living environment | Cross-sectional control-variable design | Poisson regressions | Age, gender, rural vs urban | Meta analysis: Yes SES measure significant. |
* These numbers correspond to the 59 studies from which we extracted data. In the data extraction phase, we removed an additional 15 studies The final number of studies included in the narrative synthesis was therefore the 44 listed in this table, also see documentation in S2 File.