The authors have declared that no competing interests exist.
The links between two commonly used measures of health—self-rated health (SRH) and self-reported illness (SRI)–and socio-economic and contextual factors are poorly understood in Low and Middle Income Countries (LMICs) and more specifically among women in conflict areas. This study assesses the socioeconomic determinants of three self-reported measures of health among women in the occupied Palestinian territories; self-reported self-rated health (SRH) and two self-reported illness indicators (acute and chronic diseases). Data were obtained from the 2010 Palestinian Family Health Survey (PFHS), providing a sample of 14,819 women aged 15–54. Data were used to construct three binary dependent variable—SRH (poor or otherwise), and reporting two SRI indicators—general illness and chronic illness (yes or otherwise). Multilevel logistic regression models for each dependent variable were estimated, with individual level socioeconomic and sociodemographic predictors and random intercepts at the governorate and community level included, to explore the determinants of inequalities in health. Consistent socioeconomic inequalities in women’s reports of both SRH and SRI are found. Better educated, wealthier women are significantly less likely to report an SRI and poor SRH. However, intra-oPt regional disparities are not consistent across SRH and SRI. Women from the Gaza Strip are less likely to report poor SRH compared to women from all other regions in the West Bank. Geographic and residential factors, together with socioeconomic status, are key to understanding differences between women’s reports of SRI and SRH in the oPt. More evidence is needed on the health of women in the oPt beyond the ages currently included in surveys. The results for SRH show discrepancies which can often occur in conflict affected settings where a combination of ill-health and poor access to health services impact on women’s health. These results indicate that future policies should be developed in a holistic manner by targeting physical and mental health and well-being in programmes addressing the health needs of women, especially those in conflict affected zones.
Health research and policy efforts focused on women in Low and Middle Income Countries (LMICs) have concentrated on women’s reproductive lives, specifically antenatal care and the spacing and limiting of births. Women’s health beyond reproductive ages in LMICs is generally neglected [
The oPt is an LMIC with a fragmented, over-burdened and under-resourced health system. Life expectancy at birth is 73 years and there are high levels of poverty and poor nutrition (World Bank 2015). The on-going Israeli military occupation of the oPt has created two administratively separated geographic zones: the Gaza Strip (GS) and the West Bank (WB). The population of the GS bears a heavier burden of structural violence, with restriction of movement, a restricted economy and a resultant lack of access to goods and services alongside exposure to political violence and a fractured health care system [
Comparative research between the GS and WB has shown that subjective health is heavily influenced by local perceptions of health, both at the neighbourhood level and at the level of wider social networks [
Evidence from the oPt shows that women consider the (ill-)health of a family member to take precedence over their own; women’s health is impacted both by their multiple caring responsibilities and a normative understanding that women’s health is less important than the health of others [
Intra-oPt variations in health services and their use are also present with differences in availability, access and utilization of health care between the GS and the WB [
Self-reported measures are increasingly used to assess health status and needs in LMICs [
Whilst considered a more objective health measure compared to SRH, SRI is based on self- or proxy-reports rather than medical data or diagnosis by a clinical professional.
Both SRH and SRI are independently associated with a wide range of factors (demographic, socioeconomic, education, health behaviours, health knowledge, and context), and the relation between the two indicators provides an added dimension to understanding a population’s health. The substantial literature that explores self-reported perception-based measures such as SRI and SRH with diagnosed clinical data highlights the need to better understand and interpret SRH and SRI [
In studies analysing both SRH and SRI, SRI is generally limited to being used as an indicator of the ‘robustness’ of SRH. When there is discordance between the two measures, this is frequently used to disregard the results of an analysis of SRH [
Whilst political and economic insecurity among Palestinians is positively associated with poorer objective and subjective health outcomes [
Given the diverse socioeconomic and cultural conditions, disparate health systems in the oPt, it presents a unique opportunity to explore in greater depth aspects of the SRI-SRH relationship.
The aim of this study is to assess the socioeconomic determinants of three self-reported measures of health among women in the oPt—self-reported self-rated health (SRH) and two self-reported illness indicators (acute and chronic diseases) accounting for community and setting factors.
We used the 2010 Palestinian Family Health Survey (PFHS), conducted by the Palestinian Central Bureau of Statistics (PCBS). The PFHS employed a multi-stage stratified sampling design to provide nationally representative demographic, health, and socioeconomic data for the Palestinian population living in the occupied Palestinian territory in 2010 (PCBS 2013). The survey is collected in collaboration with UNICEF and is based on the international standards of the Multiple Indicator Cluster Surveys (MICS) wave 4. Fieldwork was completed in August 2010 for the WB and October 2010 for the GS, with response rates of 90.5% and 94.8%, respectively. The sample is stratified by 16 governorates (5 in GS, 11 in WB) and 644 clusters (Primary Sampling Unit, PSU) (PCBS 2013). Within each of the clusters, 24 households were selected for the survey, yielding a sample of 15,355 households from which 19,509 women aged 15–54 were eligible for interview, of which 15,734 completed their interviews with a response rate of 74.2% (PCBS 2013).
For women aged 15–54 years, irrespective of marital status, data on SRI (chronic and acute) and SRH were collected [
Self-Reported Health (SRH) is the most commonly used measure and it uses a 5 (or 6)-point Likert scale, with respondents rating their current health status from poor to excellent [
In the 2010 PFHS, respondents were asked to evaluate their self-rated health on a 6-point scale (bad—acceptable—moderate—good—very good—excellent). In our analyses, women who rated their health as bad or acceptable or moderate were grouped as having a relatively poorer health status (20.8%; n = 3,449) and women who rated their health good or very good or excellent were grouped as having relatively good health (79.2%). Using such a grouping approach has been found to be effective in the analyses of comparative data because the results are less affected by context-specific perceptions of health [
Determinants of these three binary dependent variables are analysed using logistic regression models. Prior to the estimation of the regression models, polychoric correlation coefficients, estimated using maximum likelihood estimates, are calculated to explore the bivariate nature between the three dependent variables which have shown to be predictive of each other in the literature.
Three separate logistic models for each health outcome have been estimated. These include demographic, socioeconomic and regional covariates alongside the remaining SRH/SRI variables to explore both the concordance between measures of SRI and SRH, controlling for the determinants of each, as well as the wider determinants of SRI and SRH the covariates describe.
Initially a single level logistic regression model is estimated, but given the multilevel nature of the determinants of health outcomes and the multi-stage sampling procedure employed in the PFHS two random intercepts are added sequentially to the model, an intercept for the sampling cluster and then an intercept for the governorate.
The sampling design of the PFHS is a multistage cluster sample, with households selected within geographic clusters. In order to maintain the independence of individual level observations (here, of the women) and prevent type I errors, a random intercept for Primary Sampling Units (PSU) was introduced in each model. For each model across the dependent variables (SRIa, pSRIc, SRH), a log-likelihood ratio test is conducted. A significant result of the log-likelihood ratio test shows that there is nesting if individual women within the intercept tested (here be it sampling cluster or governorate). Beyond the methodological need to introduce a random intercept for nested data to maintain the integrity of the assumptions of model, the intercept at the cluster level is a means to explore the importance of community effects on the three health measures, with each cluster identifying a neighbourhood in the oPt. Thus, in fitting the cluster random intercept we are able to not only control for nesting in the data, but also explore the amount of variation in each health outcome that is explained at the community level.
To further explore place effects in the data, an intercept for governorate is also introduced. Governorates are important as they reflect an administrative level at which health systems can differ within the oPt, both in policy and provision. Were the log-likelihood ratio test significant, an intercept for governorate will be included in the model. This intercept can maintain the assumptions of the model, accounting for the nesting of clusters within governorates but also enable the exploration of the amount of variation in the health measures explained at the community and governorate levels in the oPt. Data are weighted to control for the survey design and random intercept models run in Stata 13 [
In these models, covariates are introduced to analyse the determinants of SRH and SRI within the oPt, including demographic, socioeconomic and regional variables as well as data on pregnancy status, anemia and parity. These were included in addition to the socio-demographic variable as they could be factors of risk for specific health outcomes (e.g.: parity can have a positive effect on SRH). Demographic covariates included in the analysis are age and marital status. Age is included in the analysis as a categorical variable with 5 year age groups from 15–19 to 50–54. Marital status is introduced into the models as a categorical variable.
Socioeconomic status of women in the sample uses three variables: household wealth; a woman’s level of education; and, a woman’s employment status. Household wealth is taken as the status of the household in which the woman resides. The asset scores, based on Filmer and Pritchett (2001), are included in the model as a categorical variable of wealth quintiles determined from asset scores using Principle Component Analysis [
Anaemia prevalence, pregnancy status and parity, self-reported by the respondents or proxies, were controlled for in the models, based on evidence showing that each independently increases the odds of reporting SRI and/or poor SRH [
Region and locality variables were included in the models to examine the effect of place on the relationship between SRH and SRI. This variable is particularly important for a study of the oPt given the geographic variation in political violence and health systems.
For this analysis, four oPt administrative regions were defined: the GS, North WB, South WB and the Central WB (inclusive of the cities of Nablus and Bethlehem and the rural areas surrounding them). Within the oPt there are also three key types of locality—rural areas, urban areas and refugee camps. Both region and locality were included as fixed effects in the models. Routine checks for outliers, collinearity and leverage were performed.
The sample’s median age was 30 years. 63.1% of women in the sample are married, 34.6% are never-married, and 2.2% are divorced or widowed. As for place of residence 36.5% lives in GS, 17.6% in North WB, 14.8% in South WB and 31.1% in Centre WB. Intra-oPt differentials in socio-demographic and health measures are revealed. In general GS reports the largest percentage of poor households and the highest number of households with more than 8 children.
Moderate positive polychoric correlations (including chi squares) were found between reporting a chronic or reported health problem and a woman rating her health as ‘poor’ (
Self-Rated Health (%) | Chronic Illness (%) | n | ||||||
---|---|---|---|---|---|---|---|---|
Good | Poor | No | Yes | |||||
Chronic Illness(%) | No | 83.5 | 16.5 | 0.56 | ||||
Yes | 42.1 | 57.9 | ||||||
Acute health problems (%) | No | 85.8 | 14.2 | 0.50 | 93.5 | 6.5 | 0.4 | 11,223 |
Yes | 56.2 | 43.8 | 76.7 | 23.3 | 3,596 | |||
Total (%) | 79.2 | 20.8 | - | 89.3 | 10.3 | - | 100 | |
N | 11,370 | 3,449 | 13,017 | 1,802 | 14,819 |
Reported anaemia is highest in the GS and Central WB (36.8% and 31.6%, respectively), compared to the North and South WB (17.7% and 13.8% respectively). The largest proportion of women who do not know their anaemia status is evident in the South WB (38.6% of women), and the reported levels of anaemia are likely and underestimate of true levels. Self-reported anaemia also varies by location and education in the oPt, higher levels of anaemia reporting are present among urban women and women with only primary level education.
The modelling strategy estimated a single level model, followed by a model with intercept for community (PSU) then a model for governorates. Log-likelihood Ratio tests results showed that both community and governorate intercepts significantly improved the model (
Models Compared for Log-Likelihood Ratio Test | Acute | SRH | Chronic Illness |
---|---|---|---|
Community vs none | 66.84 |
95.68 |
17.05 |
Governorate vs none | 139.1 |
74.46 |
9.89 |
Community & Governorate vs none | 166.4 |
134.95 |
22.66 |
*** p<0.001
** 0.001<p<0.05
The multilevel logistic model regression results for each model underline the consistency between SRI and SRH in the oPt. Controlling for all other variables, a reported illness (chronic or acute) increases a woman’s odds of rating her health as poor (
Model 1 | Model 2 | Model 3 | % of sample |
||
---|---|---|---|---|---|
FIXED | Acute | SRH | Chronic | ||
Education | Up to completed primary | 1.09 | 1.65 |
1.62 |
60.92 |
Secondary | 1.08 | 1.33 |
1.23 | 16.26 | |
Some tertiary | 1 | 1 | 1 | 22.82 | |
Region | Gaza Strip | 1 | 1 | 1 | 36.46 |
North West Bank | 2.16 |
1.70 |
1.17 | 17.65 | |
South West Bank | 1.14 | 1.56 | 1.03 | 14.80 | |
Centre West Bank | 1.66 |
1.60 |
1.18 | 31.09 | |
Locality | Camp | 1 | 1 | 1 | 10.13 |
Rural | 0.84 | 0.79 |
0.62 |
17.05 | |
Urban | 0.84 | 0.91 | 0.74 |
72.81 | |
Household Wealth Quintile | Poorest | 1.32 |
1.86 |
1.31 |
18.10 |
Poorer | 1.16 |
1.73 |
1.18 | 19.55 | |
Middle | 1.11 | 1.44 |
1.21 | 20.99 | |
Richer | 1.03 | 1.24 |
1.11 | 20.64 | |
Richest | 1 | 1 | 1 | 20.73 | |
Parity | 0 | 1 | 1 | 1 | 36.21 |
1 to 3 | 0.99 | 1.63 |
1.82 | 24.66 | |
4 to 7 | 1.12 | 1.77 | 2.02 | 30.24 | |
More than 7 | 1.21 | 1.87 |
1.73 | 8.89 | |
Age | 15–19 | 1 | 1 | 1 | 23.36 |
20–24 | 1.24 | 1.46 |
1.14 | 18.64 | |
25–29 | 1.40 |
1.95 |
2.39 |
14.70 | |
30–34 | 1.41 |
2.37 |
3.97 |
12.57 | |
35–39 | 1.44 |
2.83 |
7.042 |
10.54 | |
40–44 | 1.43 |
3.749 |
13.845 |
08.62 | |
45–49 | 1.39 |
4.05 |
22.2.438 |
06.77 | |
50–54 | 1.18 | 5.19 |
41.251 |
04.80 | |
Employment Status | Unemployed | 1 | 1 | 1 | 65.60 |
Employed | 1.07 | 0.75 |
0.942 | 09.42 | |
Student | 1.05 | 0.94 | 0.810 | 24.98 | |
Marital Status | Never Married | 0.60 |
1.14 | 1.682 | 34.66 |
Divorced/Widowed | 1.804 | 3.258 |
1.78 | 02.20 | |
Married | 1 | 1 | 1 | 63.15 | |
Pregnancy status | Not | 1 | 1 | 1 | 88.79 |
Yes | 0.847 |
1.273 |
0.646 |
08.74 |
|
Do not know | 0.616 | 0.396 |
0.711 | ||
Anaemia | No | 1 | 1 | 1 | 89.81 |
Yes | 2.159 |
2.005 |
1.683 |
06.86 | |
Don’t Know | 1.168 | 1.809 |
0.870 | 03.33 | |
Chronic Illness | No |
1 |
1 |
89.71 |
|
Reported health problems | No |
1 |
1 |
77.70 |
|
SRH | Average to Good |
1 |
1 |
79.21 |
|
Constant | 0.098 |
0.017 |
0.004 |
- | |
RANDOM: | |||||
Cluster-level | 0.449 |
0.403 |
0.349 |
- | |
Governorate | 0.201 |
0.252 |
0.157 |
- | |
N | 14819 | 14819 | 14819 | 100.00 |
***p<0.001
**p<0.01
*p<0.05
Looking at the multilevel structure, whilst there is a concordance between both chronic and acute SRI measures with SRH, there is divergence in the explanatory patterns for reporting illness or poor SRH among women aged 15–54 in the oPt. Self-reports of anaemia are significantly associated with increased odds of both chronic and acute SRI, as well as with poor SRH. The intraclass correlation coefficients in
Dependent Variable | Acute |
SRH |
Chronic Illness |
---|---|---|---|
Governorate | 0.011 |
0.018 |
0.007 |
Community within Governorate | 0.068 |
0.064 |
0.043 |
Regionally, there are differences in the explanatory patterns for the different reported health measures. However, across all four regions, no difference is found in the odds of a woman reported as having a chronic illness, holding all other variables constant (
The community level variable has no significant effect on the odds of reporting illness in the last 2 weeks. However, locality (e.g.: camp/rural/urban) is significantly associated with both reporting a chronic illness and poor SRH. Women living in camps are more likely to report poor SRH and chronic illness compared to women in rural areas (OR 0.620, p<0.001 for rural women), and are also more likely to report a chronic illness compared to women living in urban areas (OR 0.745, p<0.01 for urban women). The results show pervasive regional differences in SRI and SRH in the oPt; some variation in health measures are explained at the governorate and community levels and there are clear regional and locality differences in SRI and SRH (
Household wealth has the most consistent effect across all three health measures; with women in the poor and poorest wealth quintiles being more likely to report illness or poor health than wealthier groups. For SRH, there is a consistent wealth gradient, with the poorest women 1.9 times more likely to rate their health as poor compared to women in the richest quintile. For chronic and reported acute health problems, women are 1.3 times more likely to self-report incidence of either illness in the poorest quintile compared to the richest quintile. However, for pSRI, being in the richest wealth quintile is only comparatively protective against illness compared to women from the poorest households. Women in the richest wealth quintile are no more likely to report not having an acute or chronic illness compared to all other women, suggesting an attenuated effect of wealth on women’s health in the oPt in all but the poorest households (
Employment status only has a significant effect for SRH, with employed women less likely to rate their health as poor compared to unemployed women (OR 0.752, p<0.001). For SRI, both acute and chronic, there is no difference between employment categories (
Education level has no effect on the odds of SRIa, controlling for all other variables. However, women who have complete primary education are more likely to report both poor SRH and pSRIc compared to those with some tertiary education. Women with some tertiary education are also less likely to report a chronic illness than women who have secondary education. In general, the results are consistent with higher socioeconomic status being protective against poor health measures, but this varies between SRI and SRH measures.
The results show an expected age gradient in SRI and SRH, with older women having higher odds of reporting poor health compared to young women aged 15–19. Never married women, are 40% less likely to have a SRIa than married women, although never married women reported no significant differences in chronic illnesses and SRH compared to married women. Women who are divorced are 3.2 times more likely to rate their health as poor compared to married women, holding all other variables constant. Marital status, for women, has no effect on the likelihood of a chronic illness being reported (
Controlling for all other variables, parity is only a statistically significant predictor of SRH. Women with either 1–3 children or more than 7 children are more likely to report poor SRH compared to their nulliparous peers.
Being pregnant increases the odds of a woman rating her health as poor, compared to those who are not pregnant (OR 1.273, p<0.01). However, being pregnant reduces the odds of reporting either SRIa or chronic illness (OR 0.847, p<0.05 and OR 0.646, p<0.01 respectively). Not knowing current pregnancy status reduces the odds of reporting SRH (OR 0.396, p<0.05). Whilst there are regional and socioeconomic differentials in the odds of reporting poor health measures, the explanatory patterns vary by health measure (
This study assessed the socioeconomic determinants of three self-reported measures of health among women in the oPt: self-rated health; self-reported illness for acute illness; and, self-reported illness for any chronic disease. Analyses of the socioeconomic determinants of these measures in one context contribute to an understanding of the relation between SRI and SRH, whilst accounting for context-specific factors.
Our analyses showed a concordance between SRI and SRH health measures with significant differences in explanatory patterns across the oPt by region and socioeconomic status. Not only were there regional differences in health in general across the oPt, but there were also regional differences in measures of subjective and objective health. Women from the GS reported lower levels of self-reported and self-rated poor health outcomes. Gazan women’s reports of better health (both objective and subjective) were at odds with their relatively poorer health infrastructure, living conditions, nutrition and socioeconomic status due to the severity of occupation violence in Gaza. This finding highlights the importance of understanding context for both objective and subjective measures of health. Relative to other regions, the Gaza Strip experienced the most extreme consequences of the ongoing conflict in the oPt. Severe restrictions of the movement of people and goods, the economic and physical ramifications of the conflict has left the region with the poorest health system infrastructure in the oPt [
Employed women, relative to unemployed women were less likely to report poor SRH. Women in employment tended to have better health compared to those not in the labour force, explained as a combination of a selection effect and higher levels of life satisfaction [
The results for SRI also show that access to improved health infrastructure remains important for women’s health in the oPt. In camps, women were more likely to report a chronic illness, compared to women from either urban or rural areas. There were two possible, non-exclusive, explanations for this finding. Firstly, it is possible that there is a higher prevalence of chronic illness in refugee camps compared to urban and rural areas. Living conditions in camps, particularly those with large population densities in areas of chronic exposure to political violence, can be conducive to increased levels of stress, poverty and poor nutrition that have been shown to increase the likelihood of developing a chronic disease [
The finding that women in camps reporting better health than elsewhere supports the argument that, with chronic illness reporting, we are likely seeing higher rates due to greater access to health care, better health knowledge and also a higher rate of diagnosis of chronic illness and not necessarily just a greater incidence in refugee camps. The importance of health infrastructure and health outcomes in the oPt is also supported by the significance of the random intercept at the governorate level. In particular, with respect to government-run health services, health system planning and management is mainly centralized by the Ministry of Health in the oPt, but there is some variability to local directorate control and management at the governorate level, giving rise to further variations in health provision and access at the governorate level (WHO 2006). More specifically the Palestinian Authority’s distribution of health services and personnel is unevenly distributed by governorate in relation to population and also skewed towards hospital care, which accounts for the bulk of the national health budget [
Women in camps were also more likely to rate their health as poor compared to rural women, which is at odds with relatively better health care access but might be explained by better lifestyle and nutrition in rural areas, with lower levels of psychosocial stress from living in areas with high population density, in addition to perhaps better health knowledge given better access to health care services in the camps. Again, this result highlights the concordance between SRIc and SRH. Across areas in the oPt, there is a diverse tapestry of factors that affect women’s health, and their importance varies across regions and localities.
When we considered both socio-cultural aspects and access to health care, the fact that Gazan women had lower odds of reporting health problems, contrary to what would be expected given the living conditions, poverty and population density in the Gaza Strip, seems understandable. In addition to perhaps Gaza’s women rating their health in relation to others in their community and the dire context in which everyone lives making health complaints perceived as insignificant, increased inequality has been found to be associated with higher rates of poor SRH, and Gaza has the lowest level of inequality in the oPt (PCBS 2015). Thus despite higher levels of psychosocial stressors, poor living conditions, risk of injury and disease in GS, lower levels of inequality may well lead to lower levels of SRI and poor SRH.
Our analyses show that in the oPt women’s health is determined by a diverse range of factors, including regional, socioeconomic, demographic and cultural factors. There is a concordance between SRI and SRH measures in the oPt, suggesting that both measures are capturing elements of an underlying concept of health. However, the differences between SRI and SRH highlight the importance of elucidating and understanding more subjective and objective measures of health.
Both subjective and objective health measures should continue to be included in future health surveys in order to better understand how local populations perceive and feel about their wellbeing. There is a need for more detailed data on reported health problems, with differentiation in terms of severity and whether acute or chronic. Current, routine data collection does not permit disaggregated analyses by severity of reported health problems. Finally, there is a need for more qualitative research to better understand how health is understood in diverse contexts.
Moving forward policies at national and international level in this area will need to focus more on the needs of women in a more holistic approach which includes pre and post reproductive life. This will need to include the mental and physical wellbeing in the aftermath of post-childbearing, including peri-menopausal healthcare. Mental health as the SRH results show, is key in understanding the stigma and feeling of worthiness that follows the move of the centre of attention from being the child bearer to simply a wife in the household once the children have left. Access to diagnostics as well as to overall health care needs to be approached in a lifecourse manner in particular in conflict settings where barriers to access are further heightened. Community health workers for example could be directed in this respect would be of great use both in terms of outreach and in terms of local understandings of current needs.
This study highlights the need for a greater understanding of context when collecting health-related data in settings such the oPt affected by political violence. It also supports the understanding that, among Palestinian women at least, it is both place and subjective perspective that are important for not only the conceptualization of health but also its reporting. Finally there is a need for a greater understanding of women’s health needs over the lifecourse moving beyond narrow foci on ‘reproductive health’ and ‘health of mothers’. Without a deeper understanding of this we are not able to move forward in trying to meet peoples’ health needs.