Variations In Social Determinants Of Self-Rated Health And Self-Reported Illness
Correspondence Address :
Dr. Paul A. Bourne, 1Socio-Medical Research Institute (Formerly of Department of Community Health and Psychiatry, Faculty of Medical Sciences, University of the West Indies, Mona, Kingston, Jamaica)
Background There are extensive empirical studies which have examined the social and/or medical determinants of the health status, but none have evaluated whether those determinants vary by the definitions of health.Aims: This research seeks to elucidate the social determinants of health, based on the definition of health and the dichotomization of health, in order to establish whether variations exist in the social determinants, based on the definition and the measurement of health, as well as the correlation between the determinants.Design and setting: By using a representative probability sampling dataset, 2007 Jamaica Survey of Living Conditions, of some 6,782 respondents, and logistic regression analyses, the determinants of health were examined in these subjects.Methods and measure: Self-rated health was a five-point ordinal scale (very poor; poor; moderate; good; very good health) measure. It was dichotomized as good-to-very good health status, and moderate-to-very good health, and not reporting an illness, to measure health to explore the effect and the determinants of each definition of health.Results: When health was measured and the cut-off was good-to-very good, eight variables emerged as statistically significant factors of the self-rated health status of Jamaicans (Model, χ2 = 1187.67, P < 0.0001; -2 Log likelihood = 3374.2, R2 = 0.367). By using a cut-off of moderate-to-very good health, six variables emerged as statistically significant factors of the self-rated health status of Jamaicans. The health status was dichotomized as moderate-to-very good self-rated health and very poor-to-poor health (Model, χ2 = 498.41, P < 0.0001; -2 Log likelihood = 1491.30, R2 = 0.295). However, when self-reported illness was used to measure health, six factors emerged from a listing of social variables as explanations of self-rated health (using not reporting an illness) (Model, χ2 = 2012.57, P < 0.0001; -2 Log likelihood = 1726.05, R2 = 0.641).Conclusion: With the importance of correct information in policy making, health assessment and health modifications, the current findings provide pertinent materials that can be used by researchers and other health professionals to make correct conclusions, the choice of the dichotomization of health and the fact that the social determinants of health vary across the different subgroups of measure and the definitions of health.
Key words: Social determinants, self-rated health status, self-reported illness, dichotomization of self-rated health status
Social determinants, self-rated health status, self-reported illness, dichotomization of self-rated health status
Introduction
Many empirical studies have examined the social and/or medical determinants of health status (1)-(13), but none have evaluated whether those determinants vary by the definitions and the measurement of health. The majority of researches assume that there are no variations in the determinants of health, which is clearly not the case; because a study conducted by Bourne (14) revealed that differences exist between the social determinants of health by the sexes and the area of the residences. By using a cross-sectional survey, Bourne’s work highlighted that the “Length of time in a household and education were the social determinants which were synonymous with only urban areas; social class and gender were the social predictors of only rural areas, while age, self-reported illnesses and consumption were the correlates of all areas of residences”(14). Education and social class were the social determinants of health for females but not for males; and social assistance was a social determinant of health for males, but not for females (14). Clearly embedded in Bourne’s work, are the disparities which exist in the social determinants of health, based on particular variables and this has never been examined in the health literature within the context of the definitions and/or measurement of health.
Health models that have published provide a listing of the social and/or medical determinants of health (15), (16) as well as studies (1)-(14), but these fail to recognize the probability of the differences that are based on the conceptualization of health and the implications of such disparities for public policy and planning. Despite the fact that evidences show that differences do exist in the social determinants of health based on particular demographic characteristics (14), there is a paucity in health literature that have examined whether such differences are present, based on the definitions of health as well as the measurement of health.
While psychologists like Brannon and Feist (16) opined that using illness to measure health is a negative approach, they also posited that health is more than illness, which they referred to as the positive approach to health. The positive approach to health is in keeping with the broader definitions of health that include social, mental and physical wellbeing and more than the absence of illness. The aforementioned issues highlight the expanded definition of health, to include social, psychological and physical wellbeing and not merely the absence of diseases, as offered by the World Health Organization (WHO) in the preamble to its Constitution in 1946 (17). The WHO’s definition of health recognizes that health was and can be measured by using illness. Such a conceptualization required studies to operationalize it; to use it in health studies and to guide health policy formulation for the society.
Scholars like Grossman (1), Smith and Kington (2), Hambleton et al. (13) and Bourne (14) have used self-reported health status, which is keeping with the broader definition of health, to model the social and/or medical determinants of health. Studies conducted by the WHO and/or affiliated scholars (4)-(9) have utilized the health status to measure health, from which particular social or medical determinants emerged. On the other hand, by using self-reported illness, Bourne (11), (12) modelled the social determinants of health; and Hutchinson et al (3) used wellbeing, which is more in keeping with the positive approach to health as offered by the WHO.
Outside of the definition of health, be it illness, health status or wellbeing, how the health status is dichotomized, is another issue which has been omitted from the discourse of the variation in the social determinants of health. Many of the studies which have examined the social determinants of health have used multiple logistic regression analyses, which require the dichotomization of health. Self-reported (or rated, assessed and evaluated) health is measured by using a five-point ordinal scale variable (from very poor, poor, moderate, good and very good), and the question is about the cut-off point for the health dichotomization. According to Finnas et al. (18) “The dichotomization implies that some of the original information is lost”, and like Bourne (19), they found that care should be taken in interpreting the results, as well as the cut-off to use. This suggests that the dichotomization may affect the social determinants.
No studies emerged in a literature search that have examined whether the social determinants differ, based on not only the definition of health, but owing to the measurement of the health status (dichotomization). Clearly, there is a gap in health literature, which cannot be allowed to continue unresolved. Empirical findings have revealed that self-rated health is critical to understand biomedically determined health (18), (19), which emphasizes the importance of investigating the determinants of self-assessed health. With the importance of the social determinants of health to health policy formulation in a society (15),(16), the social determinants of health must be explored within the context of (1) the definition of health, (2) dichotomization of health, and (3) the variations of the determinants. The current research aims to evaluate the social determinants of health, based on the definition of health and the dichotomization of health, in order to establish whether variations exist in the social determinants, based on the definition and the measurement of health, as well as the correlation between the determinants.
Theoretical model
A model developed by Grossman (1), expanded upon by Smith and Kington (2), Hambleton et al. (13), and Bourne (11), (12), (14), was employed in this current study. This theoretical framework was used to guide this paper as it related to the measurement, the selection of variables and the appropriateness of statistical relationships. By using econometric analyses, Grossman modelled the social determinants of the self-rated health status of the people in the world. The use of the econometric analyses allowed Grossman to examine the influence of many independent variables on a single dependent variable, health status. Many scholars have used this approach since Michael Grossman (1), and Hambleton et al. (13), like Bourne (11), (12), (14), have used logistic regression analyses, which is among the econometric analytic techniques which are available for use in examining multiple variables on a single dependent variable.
Logistic regression is empirically accepted as a tool for interpreting the self-rated health status (11)-(14), (18), (19). This approach implies that health status must be dichotomized by the researcher, and this means that the choice of cut-off for the dichotomization might influence some difference in conclusion about the variable. Finnas et al. noted that “The odds ratios varied less when the cut-off point for dichotomization included a broader measure of bad health, than when it contained a less broad measure which could easily be interpreted, as the former cut-off point was more stable and reliable and thus, could be preferred in empirical analyses” (18). Bourne (19) concurred with Finnas et al, that self-rated health status is most reliable being cut-off at moderate-to-very good health than from good-to-very good health. Based on the recommendations by Bourne (19) and Finnas (18), examining the social determinants of health must take into account, not only the definition of health, but also its measurement (cut-off).
According to Smith and Kington (2), using Ht= f (Ht-1, Pm Go, Bt, MCt ED, Āt, ) to conceptualize a theoretical framework for “stock of health” noted that health in period t, Ht, is the result of health preceding this period (Ht-1), medical care is (MCt), good personal health is(Go), the price of medical care is (Pm), bad ones is (Bt), a vector of family education is(ED), and all sources of household income is(Āt).
In this paper, the researcher did not use a single definition of health or measurement (cut-off) like in other studies, but used two definitions of health (self-rated health status and self-reported illness) and two cut-offs in examining the social determinants of health. Thus, the current work used a modification of the multivariate model as utilized by aforementioned researchers to capture many independent variables on a single dependent variable, health.
Different independent variables can be used to examine self-rated health status and self-reported illness. However, the current research uses the same set of independent (explanatory) variables in the three hypotheses. This allows to ascertain the social determinants of each model and their effect and explanatory power of the measure of health. Thus, in this study, the researcher will test 3 hypotheses:
HG = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR, εi) ……………….. (1)
HM = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR, εi) ………..……… (2)
HI = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR,εi) ……..….……… (3)
Where HG is the self-rated health status measured from good-to-very good; HM is the self-rated health status measured from moderate-to-very good; HI is the self-rated health measure by using illness; Pmc is cost of medical care; ED is educational attainment; HH is household head; MS is marital status; HI is the health insurance coverage; SS is the social class; X is the gender of the respondents; CR is crowding; Ai is the age of the respondents; HSB is health care seeking behaviour; Y is income, AR is the area of residence and εi is the residual error.
Participants and questionnaire
The current research used the 2007 Jamaica Survey of Living Conditions (20). This sample was taken from a national cross-sectional survey of 6,718 respondents from Jamaica’s 14 parishes. The survey used a stratified random probability sampling technique to draw the original (20). The non-response rate for the survey was 29.7%, with 20.5% not responding to particular questions, 9% not participating in the survey, and another 0.2% being rejected due to ‘data cleaning’. The study used secondary cross-sectional data from the Jamaica Survey of Living Conditions (JSLC). The JSLC was commissioned by the PIOJ and the Statistical Institute of Jamaica (STATIN). These two organizations are responsible for the planning, data collection and the policy guidelines for Jamaica. Descriptive statistics provided background information on the demographic characteristics of the sub-sample population.
The JSLC is a self-administered questionnaire where the respondents are asked to recall detailed information on particular activities. The questionnaire covers demographic variables, health, the immunization of children between 0–59 months, education, daily expenses, non-food consumption expenditure, housing conditions, the inventory of durable goods and social assistance. The interviewers are trained to collect the data from the household members. The survey is conducted between April and July annually.
Statistical analyses used
Descriptive statistics such as mean, standard deviation (SD), frequency and percentage were used to analyze the demographic characteristics of the study population. Each metric variable (ie. age, income, medical expenditure, and crowding) was evaluated for skewness, as to where it exceeded 0.5 and whether the variable was logged to remove it (ie. lnincome, lnmedical expenditure). Bivariate analyses were carried out by using Pearson’s Product Moment Correlation, based on the definitions and the measurement of health. Four hypotheses were tested in this study, and they were based on the definitions and/or measurement of health. Stepwise logistic regression analyses which were used, examined the relationship between the dependent variable and some predisposed independent (explanatory) variables. The results were presented by using unstandardized B-coefficients, Wald statistics, Odds ratio and confidence interval (95% CI). The correlation matrix was examined in order to ascertain whether autocorrelation (or multicollinearity) existed between the variables. Wherever collinearity existed (r > 0.7), the variables were entered independently into the model to help determine which one must be retained during the final model construction (the decision was based on the variable’s contribution to the predictive power of the model and the goodness of fit).
Wald statistics were used to determine the magnitude (or contribution) of each statistically significant variable in comparison with the others, and the Odds Ratio (OR) for the interpreting of each significant variable.
Measures
Age is a continuous variable, which is the number of years alive since birth (by using the last birthday).
Self-reported illness (or self-reported dysfunction): The question was asked: “Is this a diagnosed recurring illness?” The answering options were: Yes, Cold; Yes, Diarrhoea; Yes, Asthma; Yes, Diabetes; Yes, Hypertension; Yes, Arthritis; Yes, Other; and No. A binary variable was later created from this construct (1= good health based on indicating no to all illnesses) in order to be applied in the logistic regression.
Self-rated health status: “How is your health in general?” And the options were very good; good; fair; poor and very poor. For this study, the construct was categorized into 3 groups – (i) good; (ii) fair, and (iii) poor. A binary variable was later created from this variable (1) (1=good-to-very good health, 0=otherwise), and (2) (1=moderate-to-very good health, 0=otherwise).
Social hierarchy: This variable was measured based on income quintiles: The upper classes were those in the wealthy quintiles (quintiles 4 and 5); the middle class was quintile 3 and the poor were those in the lower quintiles (quintiles 1 and 2).
Crowding was the total number of people in the household divided by the number of rooms (excluding kitchen, bathroom and verandah). Income was the measure by using the total expenditure (in Jamaican dollars).
Demographic characteristics of the study population
The sample size was 6 782 respondents: 48.7% males and 51.3% females (Table 1). Poverty was substantially a rural phenomenon (29.8%) compared to the peri-urban (11.5%) and the urban areas (9.3%). Eighty percent (82.2%) of the respondents indicated at least good health status (37.0% indicated excellent health status) as compared to 0.8% who claimed very poor health status. One percent (1.1%) of the sample was injured in the 4-week period of the survey, while 14.9% reported an illness and 43.2% indicated a chronic illness (i.e. Diabetes mellitus, 13.8%; hypertension, 23.1%; and arthritis, 6.2%), as compared to 30.4% who reported acute illness (influenza, 16.7%; diarrhoea, 3.0%; and asthma, 10.7%).
(Table/Fig 1): Demographic characteristics of study population, n = 6,782 †US$ 1.00 = Ja. $ 80.47
Almost 66% (i.e. 65.5%) of the sample visited a health care practitioner (i.e. doctor, nurse, healer, pharmacist) in the 4-week period of the survey; 29.6% were the heads of households; 23.3% were married; 69.2% were never married; 1.7% were divorced; 0.9% were separated; 4.9% were widowed; and the median number of persons per room was 4 (range = 1, 17). The median annual income was USD 7 050.66 (range = USD 261.56, USD 6 523.66) and the median per capita consumption was USD 1 523.88 (range = USD 179.57, USD 20 325.55).
Multivariate analyses
Hypothesis 1: HG = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR, εi)
(Table/Fig 2) presents information on the social determinants of self-rated health status. Health status is dichotomized as good-to-very good self-rated health status. Eight variables emerged as statistically significant factors of the self-rated health status of Jamaicans (Model, χ2 = 1187.67, P < 0.0001; -2 Log likelihood = 3374.2, R2 = 0.367). Overall, 84% of the data were correctly classified (95.3% of yes, and 42.0% of no).
(Table/Fig 2): Logistic regression: Explanatory variables of self-rated health status (good-to-very good health status) of Jamaicans
Model, χ2 = 1187.67, P < 0.0001; -2 Log likelihood = 3374.2, R2 = 0.367
Hypothesis 2: HM = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR, εi)
Six variables emerged as being statistically significant factors of the self-rated health status of Jamaicans. Health status was dichotomized as moderate-to-very good self-rated health and very poor-to-poor health (Model, χ2 = 498.41, P < 0.0001; -2 Log likelihood = 1491.30, R2 = 0.295). Overall, 94% of the data were correctly classified (99.2% of yes, and 15.6% of no) (Table 3).
(Table/Fig 3): Logistic regression: Explanatory variables of self-rated health status (moderate-to-very good health status)
Model, χ2 = 498.41, P < 0.0001; -2 Log likelihood = 1491.30, R2 = 0.295
Hypothesis 3: HI = f(Pmc, ED, HHt, MS, HI, SS, X, CR, Ai, HSB, Y, AR, εi)
(Table/Fig 4) shows that six factors emerged from a listing of social variables as explanations of self-rated health (using not reporting an illness) (Model, χ2 = 2012.57, P < 0.0001; -2 Log likelihood = 1726.05, R2 = 0.641). Overall, 94% of the data were correctly classified (99.6% of yes, and 64.3% of no).
(Table/Fig 4): Logistic regression: Explanatory variables of self-rated health (not reporting an illness)
Model, χ2 = 2012.57, P < 0.0001; -2 Log likelihood = 1726.05, R2 = 0.641
(Table/Fig 5) summarizes at a quick glance at the social determinants of health which influence a particular definition and the measurement of self-rated health.
(Table/Fig 5): Summary of the social determinants of self-rated health status of Jamaicans by good-to-very good, moderate-to-very good health status and not reporting an illness
N represents no; Y indicates yes. NA denotes not applicable.
Hypothesis 4: The strength and magnitude of the variables that were used to ascertain the social determinants of self-rated health status or self-reported illness.
[Tables/Figs 6]-(8) present information on the correlation among the different variables that were used to ascertain the social determinants of self-rated health status or self-reported illness.
(Table/Fig 6): Correlation Matrix of Good-to-very good health
(Table/Fig 7): Correlation Matrix of Moderate-to-very good health
(Table/Fig 8): Correlation Matrix of Not reporting an illness
In summary, variations exist in the social determinants of health as a result of the measure of health (ill-health to measure health rather than self-rated health) and cut-off point in the dichotomization choice of self-assessed health. Owing to the fact that self-rated health is an important tool in understanding the biomedically determined health, this implies that the cut-off point for health is important and that it cannot be made arbitrarily in health statistics. As empirical evidence shows that different definitions and the measure of health correlate with the social determinants of health differently, this must be taken into consideration in the social determinants of health. The results highlight the variations in the social determinants of health across the different cut-off points in dichotomizing self-rated health and self-reported illness, which indicate that researchers and policy makers must recognize these findings in health planning, as well as in the correlation among and between the social determinants.
No significant effects of health insurance and social class were detected on the measure of health, which was different from the social determinants of health which were identified by the WHO or other researchers. These findings more than highlight the variations in the social determinants of health, but bring into focus, differences which might occur between societies. Such variations in the social determinants of health, therefore, indicate that conditionality must be established in the determinants of health because of the measures of health, the cut-off approach in dichotomization and the locality which emerged from the current work.
Disclosures
The author does not report conflict of interest with this work.
Disclaimer
The researcher would like to note that while this study used secondary data from the Jamaica Survey of Living Conditions, none of the errors in this paper should be ascribed to the National Family Planning Board, but to the researcher.
The author thanks the Data Bank in the Sir Arthur Lewis Institute of Social and Economic Studies, the University of the West Indies, Mona, Jamaica for making the dataset available for use in this study.
- Emerging Sources Citation Index (Web of Science, thomsonreuters)
- Index Copernicus ICV 2017: 134.54
- Academic Search Complete Database
- Directory of Open Access Journals (DOAJ)
- Embase
- EBSCOhost
- Google Scholar
- HINARI Access to Research in Health Programme
- Indian Science Abstracts (ISA)
- Journal seek Database
- Popline (reproductive health literature)
- www.omnimedicalsearch.com