Key findings and recommendations
-
There is no standard method for assessing sex differentials in under-five mortality in order to identify unexpected sex ratios that may suggest the presence of gender bias.
-
The identification of unexpected sex ratios will vary depending on the assumptions and methods used to define expected values.
-
The historical demographic experiences of currently high-income countries provide the main reference parameters for the assessment of sex differentials, but different epidemiological profiles are present in current-day populations.
-
Most studies failed to consider the overall level of mortality, relied upon a single expected value for multiple countries, and did not report on statistical assessments of the differences between observed and expected estimates.
-
The use of external parameters for comparison that consider the level of overall mortality and apply appropriate statistical methods to compare observed and expected ratios are recommended.
-
Whereas gender bias against girls is the most frequent explanation for unexpected sex ratios, methods should also consider the possibility of higher-than-expected mortality of boys.
Background
Methods
Data sources and search strategy
Eligibility criteria
Article relevance screening
Studies characterization
Additional information
Results
Selection of studies
General characteristics of studies included
Methodological characteristics
Reference studies
Author, year | Observed data | Measure | Age group | Definition of expected values | Comparison method | Type of reference | Citations | ||
---|---|---|---|---|---|---|---|---|---|
Reference population | Assessment method | Expected value | |||||||
Clark, 1987 [77] | India, published estimates | Sex ratioa | 1–59 months | 25 ‘Third World’ countries, World Fertility Surveys | Average sex ratio | 99.6 | Magnitude of observed and expected sex ratios, significance assessment | Descriptive | 1 |
Johansson and Nygren, 1991 [93] | China, 1988 Two-per-thousand Fertility Survey | Sex ratio | IMR | Countries with information for at least four years 1976–1984, United Nations Demographic Yearbook | Average sex ratio | 130 | Magnitude of observed and expected sex ratios | Prescriptive | 155 |
Svedberg, 1991 [95] | 23 Sub-Saharan African countries, published estimates, 1953–1983 | Sex ratio, Excess mortality | IMR | Sweden official statistics, 1983–1987 | Sex ratio of the Swedish population | Ratio of ratios (observed/ expected) above or below the unity | Descriptive | 89 | |
CMR | |||||||||
Goodkind, 1995 [100] | Vietnam | Sex ratio | IMR | East Asian populations | Average sex ratio | 97 | Magnitude of observed and expected sex ratios | Descriptive | 35 |
CMR | 95 | ||||||||
Hill and Upchurch, 1995 [12] | 35 LMICs, DHS, 1986–1993 | Sex ratioa | IMR | England and Wales, France, Netherlands, New Zealand, and Sweden, Coale and Demeny (1983) West model, 1820–1964 | LOWESS curve for the association between sex ratios and male U5MR (from 25 to 300) | 130–118 | Difference between observed and expected sex ratio for each value of male mortality | Prescriptive | 128 |
CMR | 122–103 | ||||||||
U5MR | 129–111 | ||||||||
Klasen, 1996 [105] | 20 LMICs, Census and World Fertility Surveys | Sex ratio; Male mortality | IMR | Coale and Demeny (1983) West and North models and Sweden official statistics, 1983–1987 | Sex ratio of reference populations | Ratio of ratios (observed/ expected) above or below the unity | Descriptive | 42 | |
CMR | |||||||||
Chaudhuri, 2011 [18] | India, 2005–06 National Family and Health Survey | Female mortality | IMR | Kerala state, India (same survey) | Linear regression for sex ratio based on male U5MR and being from Kerala | Expected female mortality derived from expected sex ratio based on the regression coefficients | Excess female mortality if observed value greater than expected; probit model | Prescriptive | NA |
Srinivasan and Bedi, 2011 [42] | Tamil Nadu, India, Vital Events Survey, 1996–1999 | Female mortality | IMR | Values from Waldron (1983), Johansson and Nygren (1991), Hill and Upchurch (1995), United States and United Kingdom life tables | Female mortality in function of male mortality based on the values reported in previous studies | Expected female mortality = 80% of male mortality | Difference between observed and expected values | Prescriptive | CrossRef citations: 9 |
Chaudhuri, 2012 [19] | 13 Indian states, National Family Health Surveys, 1992, 1998 and 2005 | Female mortality | IMR | Kerala state, India (same surveys) | Multivariate logistic regression | Regression coefficients | Incidence of excess female mortality = difference between observed female IMR in each state and the benchmarking female IMR (Kerala) | Prescriptive | 1 |
Monden and Smits, 2013 [138] | 35 Sub-Saharan African and Southern Asian countries, DHS, 2000s | Sex ratio | CMR | Austria, Belgium, United Kingdom, France, Germany, Netherlands, Scandinavian countries, USA, Canada, New Zealand, and Australia, Human Mortality Database, since 1920 | Average sex ratio | 117 | Magnitude of observed and expected sex ratios | Prescriptive | 14 |
U5MR | 125 | ||||||||
Jamison et al., 2013 [136] | India and China, published estimates | Sex ratio | U5MR | Demographic and health surveys from LMICs | Average sex ratio | 118 | Magnitude of observed and expected sex ratios; estimates of excess of female mortality based on male mortality | Descriptive | 605 |
Alkema et al., 2014 [2] | 195 countries, areas, and territories, multiple sources, 990–2012 | Sex ratio | IMR | 195 countries, areas, and territories, multiple sources, 1990–2012 | Global relation between sex ratios and mortality levels; Bayesian model | 120, 126 and 115 for IMR of 5, 20 and 150 | Assessment of outlying values | Descriptive | 45 |
CMR | 121–101 for CMR of 5 to > 30 | ||||||||
U5MR | 125–109 for U5MR of 20–400 | ||||||||
Chaudhuri, 2015 [20] | 14 Indian states, National Family Health Surveys, 1992, 1998 and 2005 | Female mortality | IMR | Bihar vs 13 Indian states; Bihar vs 8 less gender bias Indian states (same surveys) | Multivariate logistic regression | Coefficient of interaction between being from Bihar and female sex | Descriptive | 1 | |
Guilmoto et al., 2018 [153] | India, 2011 Census | Female mortality | U5MR | 46 countries without known gender discrimination, World Population Prospects | Quadratic regression for the relation between female and male U5MRs | Expected female U5MR for each level of male U5MR | Difference between observed and expected female U5MR; absolute excess female mortality | Prescriptive | 19 |
Comparative studies
Author, year | Observed data | Measure | Age group | Definition of comparison parameters | ||
---|---|---|---|---|---|---|
Reference population | Reference value (when applicable) | Comparison method | ||||
Hammoud, 1977 [68] | Algeria, Democratic Yemen, Egypt, Iraq, Jordan, Kuwait, Libyan Arab Jamahiriya, Morocco, Syrian Arab Republic and Tunisia, multiple sources, 1951–1974 | Sex ratio | IMR | Mauritius, Canada, Chile, Mexico, Paraguay, United States, Hong Kong, Japan, Philippines, Thailand, Denmark, Hungary, Portugal, Yugoslavia, and Australia (United Nations and World Health Organization) | – | Magnitude of observed and expected sex ratios |
CMR | ||||||
Khosla, 1980 [69] | 17 states, India, Health statistics, 1971–1975 | Sex ratio | NMR | 46 countries from 1970 to 1974 (WHO annual statistics) | – | Magnitude of observed and expected sex ratios |
PNMR | ||||||
IMR | ||||||
Choe, 1987 [76] | Korea, 1974 National Fertility Survey, 1960–1974 | Sex ratio | IMR | Coale and Demeny (1983) West and North models, levels 19, 20 and 21 and life tables for 10 countries (Israel, Jordan, Kuwait, Hong Kong, Sarawak, Panama, Belize, Jamaica, Guyana, Portugal) | 122–133; 81–140 | Magnitude of observed and expected sex ratios; Hazard models for multivariate analysis |
CMR | 111–124; 71–118 | |||||
Das Gupta, 1987 [78] | Rural Punjab, India, Khanna Study, 1984 | Sex ratio | U5MRa | Khanna 1957–1959 and Matlab Thana 1974–1977 | – | Magnitude of sex ratios |
Karkal, 1987 [79] | India, Sample Registration System, 1970–1980 | Sex difference | 0, 1, 5 years | South Asia region | – | Magnitude of sex-specific mortality and sex differences |
Makinson, 1987 [80] | Egypt, 1980 World Fertility Survey | Female mortality | U5MRa | Coale and Demeny (1966) West model, level 13.7 | Observed female mortality | Magnitude of sex-specific mortality; multivariable logistic model |
Chowdhury et al., 1990 [88] | Bangladesh, Matlab Demographic Surveillance System, 1977–1985 | Female mortality | NMR | Comparison of 204 sex discordant twin pairs with a random sample of 2371 singletons | Odds ratio 98 | Logistic regression and McNemar’s test to assess sex differences and conditional survivorship |
IMR | Odds ratio 140 | |||||
Pebley and Amin, 1991 [39] | 26 rural villages in India, Narangwal Study | Sex ratio | Under-3a | Study comparison area | – | Expected mortality rates 1971–1973 without intervention (control villages) |
Tabutin, 1992 [97] | Algeria, Morocco, Tunisia and Egypt, multiple sources, 1965–1988 | Sex ratio | IMR | United Nations model life tables for developing countries | “general pattern” for the reference countries | Magnitude of observed and expected sex ratios |
CMR | ||||||
Choe et al., 1995 [99] | China, 1988 Two-per-Thousand Survey of Fertility and Birth Control, 1965–1987 | Sex ratio | IMR | Coale and Demeny (1983) West and North models, level 20, and Japan 1953–1960 | 119, 123, 129 | Magnitude of sex ratios; multivariate proportional hazard models |
CMR | 111, 113, 115 | |||||
Clark, 1995 [53] | Gwembe District, Zambia, Gwembe Study, 1956–1992 | Sex-specific mortality | IMR | Twin pairs and singletons | – | Comparison of sex-specific mortality rates |
Johansson, 1996 [104] | Meiji, Japan, Published estimates, 1908 | Sex ratio | IMR | Swedish estimates (1750–1900), Preston standard (1976) | – | Magnitude of observed and expected sex ratios |
CMR | ||||||
Muhuri and Menken, 1997 [108] | Matlab, Bangladesh | Sex ratio | 1–5 years | Study comparison area | – | Magnitude of sex ratios; logistic regression |
Goodkind, 1999 [113] | North Korea, 1993 Census | Sex ratio | IMR | Previous studies (Makinson 1994; UN 1998); South Korea, China, and Taiwan | 115–140 | Magnitude of observed and expected sex ratios |
CMR | 100–120 | |||||
Datta and Bairagi, 2000 [21] | Bangladesh, Matlab Demographic Surveillance System, 1977–1995 | Sex ratio | IMR | Coale and Demeny (1983) West model and study comparison area | Excess female mortality from the equation [(observed sex ratio) - (expected sex ratio)] / (observed sex ratio)] (× 100) | |
Yount, 2001 [47] | 14 Middle Eastern countries, United Nations, 1970s and 1980s | Sex ratio | IMR | Same datasets for expected and observed estimates | From Hill and Upchurch (1995) estimated from the same dataset | Magnitude of observed and expected sex ratios |
CMR | ||||||
U5MR | ||||||
Li et al., 2004 [124] | Chinese county in Shaanxi province, 1997 Household survey and community survey | Sex ratio | IMR | Published estimates of sex ratios from Li and Feldman (1996); Coale and Demeny (1983) West model | 120–140 | Magnitude of sex ratios; likelihood ratio test; t test; multivariate logistic regression and Cox survival |
CMR | 100–120 | |||||
U5MR | > 100 | |||||
Fuse and Crenshaw, 2006 [127] | 93 countries, United Nations Statistics Division, 2000 | Sex ratio | IMR | Published estimates Johansson and Nygren (1991); Hill and Upchurch (1995); Tabutin and Willems (1995) | 115 to 130 | Magnitude of sex ratios |
Jayaraj, 2009 [132] | India, Vital Registration System (1991 and 2001) and published estimates | Female mortality | U5MR | Coale and Demeny (1966) West model, levels 18 and 19 | Relative survival advantage of females (RSASF) | Magnitude of observed and expected RSAF |
Oster, 2009 [36] | India, National Family Health Surveys, 1992 and 1998 | Female mortality | Under-10a | Ethiopia, Kenya, Malawi, Namibia, Tanzania, and Zambia, DHS, 1992–2001 | Regression coefficients, allowing for the interaction between being from India and female sex | Difference-in-differences |
Costa et al., 2017 [14] | 60 LMICs, DHS, 2005–2014 | Female mortality | U5MR | Same DHS datasets for expected and observed estimates | From Hill and Upchurch (1995) and Alkema et al. (2014) estimated from the same dataset | Excess female mortality (%) = observed/ expected |