Self-rated health of the educated and uneducated classes in Jamaica.

BACKGROUND: Education provides choices, opportunities, access to resources and it is associated with an increased likelihood of higher income. Does this holds true in developing nations like Jamaica, and does the educated class experience greater self-rated health status than the uneducated classes? AIMS: The current study will identify the socio-demographic correlates of self-rated health status of Jamaicans, examine the effects of these variables, explore self-rated health status and self-reported diagnosed recurring illness among the educated and uneducated classes, compute mean income among the different educational types, and determine whether a significant statistical correlation exists between the different educational cohorts.
MATERIALS AND METHODS: The current study utilised the data set of Jamaica Survey Living Conditions which is a cross-sectional survey. It is a national probability survey, and data were collected across the 14 parishes of the island. Stratified random sampling techniques were used to draw the sample.
RESULTS: Self-rated health statuses of respondents are correlated with age, income, crowding, sex, marital status, area of residence, and self-reported illness (es) - χ(2) = 1,568.4, P < 0.001. Respondents with tertiary level educations were most likely to be classified in the wealthiest 20% (53.4%) and there was no significant statistical difference between their health status and the lower educated classes.
CONCLUSION: There is a need for a public health care campaign that is specifically geared towards the educated classes as their educational achievement is not translating itself into better health care-seeking behaviour and health status than the uneducated classes.

Introduction

Health is imperative for socio-economic and political development of people, a society and a nation. It is within this context that a study of health is critical as it relates to the wider society. Traditionally, the concept of health is measured using life expectancy, mortality, and diagnosed illness. In the social sciences, researchers have used self-rated health status[1-9], and self-reported illness[10-17] to measure health. Apart from those terminologies, other synonyms such as self-assessed health, self-reported health, perceived health, self assessment of health, global health status, and health status have all been used to speak about health. It follows from the aforementioned perspective that all those terms imply the same measurement of health or health status. Self-rated health status is among the subjective indexes used to measure health, and some scholars argue that they are not a good assessment of health when it comes to life expectancy, per capita income, or mortality[18-20].

The subjective/objective indexes of measuring health emerged as scholars sought to ensure that the measurement of health was a reliable and valid one. Some scholars opined that the self-assessment of one's health status was more comprehensive than objective assessment[3521] as it included one's health and general life satisfaction. Studies have shown that subjective indexes are a good measurement for mortality[222-24] and life expectancy[25]. Concurringly, a recently conducted study by Bourne[25] found that self-assessed illness was not a good measure of mortality; however, it was very useful when it came to the subject of life expectancy in Jamaica.

The subjective indexes in measuring health open themselves up to systematic and unsystematic biases[26]. People's perception can be biased as they may inflate or deflate their status in an interview or on a self-administered instrument (i.e., questionnaire). Another aspect of bias in subjective evaluation of health is the matter of recall. It is well established in research literature that as people age, their mental faculties decline[27-32], suggesting that some people will have difficulties recalling experiences which happened in the past. Within the context of the time recollection, bias can occur in subjective indexes. Kahneman[33] devised a procedure of integrating and reducing the subjective biases when he found that instantaneous subjective evaluations are more reliable than assessments of recollection of experiences. Contrary to Kahmeman's work, Bourne's[25] results show that self-assessed health for a 4-week period is a good measure of life expectancy (objective index). In spite of the fact that subjective indexes are a good measure of objective health, the former still contains biases, which Diener[34] opines still have valid variance.

It is well established in health research that there is a correlation between or among different socio-demographic, psychological and economic variables[46-1720] and self-rated health status. The correlates include education, marital status, area of residence, education, income, psychological conditions (i.e., positive and negative psychological affective conditions), and other variables. Freedman & Martin[35], using data from 1984 and 1993's panel survey of Income and Program Participation, noted that there was an association between educational level and physical functioning of people over 65 years. Another study by Koo, Rie & Park[36], using multivariate regression, concluded that education was a predictor of increased subjective wellbeing (t ( 2523) = 7.83, P<0.001, which means that education was more than associated with health. Concomitantly, another research found that the number of years of school (i.e., the Quantity Theory) was a crucial predictor of health status of an individual[37] which indicates that tertiary level graduates are more likely to be healthier than non-tertiary level educated people. While education provides choices, opportunities, access to resources and is associated with increased likelihood of achieving a higher income, does it hold true in developing nations like Jamaica that the educated class has greater self-rated health status than the uneducated classes? A paucity of information (research literature) exists in Jamaica on the educated and uneducated classes and their self-rated health status, self-reported illness(es), the areas in which the educated and uneducated classes reside, health care-seeking behaviour among the different educational classes and the self-rated health status of Jamaicans and its correlates.

The current study is important, as it uses a statistical technique which accommodates all items in self-rated health status categories as opposed to dichotomising self-rated health. Dichotomising self-rated health status in good and poor health means that some of the original information will be lost; and this explains why some researchers argue for the maintenance of the Likert nature of the measuring tool over dichotomisation[38-40]. Secondly, the study is significant as it included more variables: (1) educational levels and area of residence, (2) educational levels and health care-seeking behaviour, (3) health insurance coverage and educational levels, (4) self-reported illness(es) and educational levels, (5) social standing and educational levels. The objectives of the current study therefore are to (1) identify the socio-demographic and economic correlates of self-rated health status of Jamaicans, (2) examine the effects of these variables, (3) explore self-rated health status and self-reported diagnosed recurring illness among the educated and uneducated classes, (4) calculate the mean age of respondents in the different educational categories, (5) compute mean income among the different educational types, and (6) determine whether a significant statistical correlation exists between the different educational cohorts.

Materials and Methods

Data

A joint survey on the living conditions of Jamaicans was conducted between May and August of 2007 by the Planning Institute of Jamaica (PIOJ) and the Statistical Institute of Jamaica (STATIN)[41]. The survey is called the Jamaica Survey of Living Conditions (JSLC) which began in 1988 and is now conducted annually. The JSLC is a modification of the World Bank's Living Standards Measurement Study (LSMS) which is a household survey[42]. The current study used the JSLC's data set for 2007 in order to carry out the analyses of the data[43]. It had a sample size of 6,783 respondents, with a non-response rate of 26.2%.

The JSLC is a cross-sectional survey which used stratified random sampling techniques to draw the sample. It is a national probability survey, and data was collected across the 14 parishes of the island. The design for the JSLC was a two-stage stratified random sampling design where there was a Primary Sampling Unit (PSU) and a selection of dwellings from the primary units. The PSU is an Enumeration District (ED), which constitutes a minimum of 100 residences in rural areas and 150 in urban areas. An ED is an independent geographic unit that shares a common boundary. This means that the country was grouped into strata of equal size based on dwellings (EDs). Based on the PSUs, a listing of all the dwellings was made, and this became the sampling frame from which a Master Sample of dwellings was compiled. This, in turn, provided the sampling frame for the labour force. One third of the Labour Force Survey (i.e. LFS) was selected for the JSLC. The sample was weighted to reflect the population of the nation.

Instrument

A self-administered instrument (i.e., questionnaire) was used to collect the data from respondents. The questionnaire covers socio-demographic variables such as education, age, and consumption, as well as other variables like social security, self-rated health status, self-reported health conditions, medical care, inventory of durable goods, living arrangements, immunisation of children 0-59 months, and other issues. Many survey teams were sent to each parish according to the sample size. The teams consisted of trained supervisors and field workers from the Statistical Institute of Jamaica.

Statistical Analyses

The Statistical Packages for the Social Sciences – SPSS-PC for Windows version 16.0 (SPSS Inc; Chicago, IL, USA) – was used to store, retrieve and analyze the data. Descriptive statistics such as median, mean, percentages, and standard deviation were used to provide background information on the sample. Cross tabulations were used to examine non-metric dependent and independent variables. Analysis of variance was used to evaluate a metric and a non-dichotomous variable. Ordinal logistic regression was used to determine socio-demographic, economic and biological correlates of health status of Jamaicans, and identify whether the educated have a greater self-rated health status than uneducated respondents. A 95% confidence interval was used to examine whether a variable is statistically significant or not.

There was no selection criterion used for the current study. On the other hand, for the model, the selection criteria were based on 1) the literature; 2) low correlations, and 3) non-response rate. The correlation matrix was examined in order to ascertain if autocorrelation and/or multicollinearity existed between variables. Based on Cohen and Holliday[44] and Cohen and Cohen[45], low (weak) correlation ranges from 0.0 to 0.39, moderate – 0.4-0.69, and strong – 0.7-1.0. This was used to exclude (or allow) a variable in the model. Any correlation that had at least a moderate value was excluded from the model in order to reduce multicollinearity and/or autocorrelation between or among the independent variables[46-51]. Another approach in addressing and/or reducing autocorrelation was to include in the model all variables that were identified from the literature review with the exception of those where the percentage of missing cases were in excess of 30%.

The current study used the ordinal nature of the dependent variable (self-rated health status or self-rated health) which denotes that none of the original data will be lost as is the case in dichotomising self-rated health. Ordered regression model is written as:

Where x is the vector of covariates with coefficient to be estimated, k is the number of cut-points for the dependent variable, and αs, αl stand for the intercepts in the regression models. Anderson[52] opined that ø1 =1 and øk, and that other constraints are possible. In the current study, the researcher set ø1 =1 and 0= ø1 < ø2 < …< øk =1 to correspond to the levels from very good to very poor, and other levels of health are relative to “very good”. Based on Anderson's arguments, the monotone increase of ‘ø's are dealt with by varying the sign for β. Within this context, a positive estimation of coefficient denotes that those with this characteristic would be negatively associated with good health status and those without would positively associated with good health status (or self-rated health status). Simply put, positive estimation of coefficients means poor health and negative estimation of coefficients denotes better self-reported health status.

Measurement of variables

Dependent variable

Self-rated health status (i.e., self-rated health) was derived from the question, “Generally, how is your health?” with the options being very good, good, fair (or moderate), poor, or very poor. The ordinal nature of this variable was used as was the case in the literature[38-40].

Independent variables

Information on self-reported illness was derived from the question, “Have you had any illnesses other than injury?” The examples given include cold, diarrhoea, asthma attack, hypertension, arthritis, diabetes mellitus or other illness. A further question about illness asked, “(Have you been ill) in the past four weeks?” The options were yes and no. This variable was re-coded as binary value, 1 = yes and 0 = otherwise.

Information about self-reported diagnosed recurring illness was derived from the question, “Is this a diagnosed recurring illness?” The options were: (1) yes, cold; (2) yes, diarrhoea; (3) yes, asthma; (4) yes, diabetes mellitus; (5) yes, hypertension; (6) yes, arthritis; (7) yes, other; (8) no.

Information on medical care-seeking behaviour was taken from the question, “Has a health care practitioner, healer, or pharmacist been visited in the last 4 weeks?” The options were yes or no. Medical care-seeking behaviour therefore was coded as a binary measure where 1 = yes and 0 = otherwise.

The term crowding refers to the average number of person(s) per room excluding the kitchen, bathroom, and veranda (i.e., total number of people in household divided by the total number of rooms excluding kitchen, bathroom and veranda).

Total annual expenditure was used to measure income. Income quintile was used to measure social standing. The income quintiles ranged from poorest 20% to wealthiest 20%.

Results

Demographic characteristic of sample and bivariate analyses

The sample was 6,783 respondents: 48.7% males and 51.3% females. Eighty-two percent of respondents rated their health status as at least good compared to 4.9% who rated it as poor. Fifteen percent of respondents reported some form of illness within the last 4 weeks. Of those who recorded an ailment, 89% reported that the dysfunction was a diagnosed recurring one. The most frequently recurring illness was unspecified conditions (23.4%) followed by hypertension (20.6%), cold (14.9%), diabetes mellitus (12.3%), and others (Table 1).

Table 1

Demographic characteristic of sample (n=6,783)

The median age of the sample was 29.9 years (range = 99 years). The median annual income was US $7,050.66 ( rate in 2007: 1US$ = JA$80.47; range = US $4,406.20), and median crowding was 4.0 persons per room (range = 16 persons).

A cross-tabulation between educational level and area of residence revealed a significant statistical correlation – χ2 (DF = 40 = 78.02, P < 0. 001 (Table 2). Based on Table 2, 0.8% of rural respondents had tertiary level education and 5.4 times more urban residents had tertiary level education compared to rural respondents.

Table 2

Educational level by area of residence (n = 6,592)

No significant statistical correlation existed between educational level and sex of respondents – χ2 (DF = 2) = 5.61, P> 0.05 (Table 3). Similarly, no significant statistical association was found between purchased prescribed medication and educational levels of respondents – χ2 (df = 10) = 11.9, P > 0.05.

Table 3

Education level by sex of respondents (n = 6,592)

A significant statistical difference was found between mean age of respondents who are at different educational levels – F statistic (2, 6589) = 214.64, P <0.001. The mean age of respondents with primary level of education and below was 32.0 years (SD = 22.6, 95% CI = 31.4-32.6) compared to 14.6 years (SD = 1.7, 95% CI = 14.5-14.8) for those with secondary education level and 26.4 years (SD = 10.6, 95% CI = 24.6-28.2) for those with tertiary education level.

A cross-tabulation between self-reported illness and educational level revealed a significant statistical association - χ2 (DF = 2) = 61.33, P < 0.001. Respondents with primary education level and below recorded the greatest percent of people with illness(es) (16.2%) followed in descending order by tertiary level (9.2%) and secondary level respondents (5.4%). The statistical correlation was a weak one – correlation coefficient = 0.10.

A significant statistical correlation existed between self-reported diagnosed recurring illness and educational level – χ2 (DF = 14) = 42.56, P < 0 .001 (Table 4). Respondents with secondary level education (37.5%) had the highest percent of unspecified health conditions followed in descending order by tertiary (33.3%) and primary level respondents (22.7%). Hypertension was substantially a phenomenon occurring among those with primary education level and below: 21.6%, compared to 8.3% of tertiary level individuals. Similarly, diabetes mellitus (12.8%) was more prevalent among primary level respondents compared to 5.0% of secondary level respondents. On the other hand, asthma was the greatest among tertiary level respondents (33.3%) compared to secondary level (22.5%) and primary level respondents (8.7%).

Table 4

Self-reported diagnosed recurring illness and social standing by educational level

Respondents with tertiary level education were most likely to be classified in the wealthiest 20% (53.4%) compared to those with secondary education who were more likely to be in the middle class and those with primary level education were either in the poorest 20% (20.3%) or in the wealthiest 20% (20.3) (Table 4) – χ2 (df = 8) = 124.53, P < 0.001.

Of the 20.2% of respondents who had health insurance coverage, tertiary level people were more likely to have private coverage (35.9%) followed by primary or below (12.0%) and secondary level individuals (11.6%) – χ2 (DF = 4) = 76.95, P < 0.001 (Table 4).

Concurringly, a significant statistical difference existed between the mean age among the different educational levels in which respondents were categorised (Table 4) – F statistic (2, 6589) = 214.6, P < 0.001: mean age for those with at most primary level education was 32.0 years (SD = 22.6) compared to a mean age of 26.4 years (SD = 10.6) for those with tertiary level education. When educational level of respondents was disaggregated into no formal, basic, and primary to tertiary, the mean age of respondents with no formal education was 42.7 years (SD = 18.0), 2.7 years (SD = 1.9) for basic school level respondents, and 9.0 years (SD = 2.2) for those who have primary level education – F statistic (4,6587) = 2207.9, P < 0.001.

Multivariate analysis

Self-rated health statuses of respondents are correlated with (1) age, (2) income, (3) crowding, (4) sex, (5) marital status, (6) area of residence, and (7) self-reported illness(es) – χ2 = 1,568.4, P < 0.001; and that the data is a good fit for the model – LL = 9,218.0. The 7 socio-demographic and economic correlates accounted for 33% of the variability in self-rated health status (Table 5). Based on the Table 5, the older the respondents get, the more likely they are to rate their health status as poor and this was the same for crowding and for those who report an illness (health condition). Urban residents are more likely to report poor self-rated health status than rural residents. However, there was no statistical difference between self-rated health status for rural and semi-urban residents. Married people are more likely to report better self-rated health status than widowed people, people with more income are more likely to report better health status, and males are more likely than females to report better health status. However, no significant statistical difference was found between self-rated health status among the educated and uneducated cohorts.

Table 5

Ordinal logistic regression: Socio-demographic and biological differentials of self-rated health status of Jamaicans

Discussion

The current study concurs with the literature in that self-reported illness has the most influence on self-rated health status of people[8]. In a study of elderly Barbadians (ages 60+ years), Hambleton et al.[8] found that current illness accounted for 87.7% of the variance in self-rated health status. In another study on married people in Jamaica, Bourne and Francis[53] found that 73% of self-reported illnesses explain the variability in self-reported health status. Embedded in the current finding is whether self-rated health is examined on elderly or married people. Current self-reported illnesses accounted for a critical proportion of self-rated health and can be used to measure health. Within this context, self-reported illness is a good measure of self-rated health, and this has been established by other studies[10-1725]. A recently conducted research found that self-reported illness accounted for 54% (r-square) of the variance in life expectancy of Jamaicans[25], and this increased to 63% for males. Subjective indexes such as self-rated health and self-reported illness can be used to measure health, but the latter is a better measure and this must be taken into consideration in the interpretation of findings using this measurement.

The challenges noted by some researchers in using self-rated health are: (1) bias and (2) the dichotomisation of the measure. While bias is synonymous with subjective assessment or evaluation of any construct, the validity of using the measure is high. Diener[34] noted in 1984 that there are still some valid variances, which was validated in a recent study by Bourne[25]. Health literature has long established that subjective indexes such as self-rated health, happiness, and life satisfaction are good measures of health as they are more comprehensive (including social activities and relationships, psychological conditions, emotions, spirituality, life satisfaction) while still incorporating the objective component[32134]. This is justified by studies that found strong statistical correlations between subjective health and objective indexes such as life expectancy[25] and mortality[222-24]. It should be noted here that subjective indexes (e.g., self-reported illness) and mortality are lowly correlated in Jamaica[25], which suggests that health literature among regions has revealed different findings. This denotes that the wholesale use of what is obtained in one nation cannot be applied to another without understanding socio-demographic characteristics. However, Jamaica, like other nations, can use subjective indexes to assess health status of its people and by extension its entire population.

The issue of the dichotomisation of self-rated health, because some of the original values will be lost, is now resolved by this study as self-rated health was dichotomised and findings were similar to those who had dichotomised the dependent variable (i.e., self-rated health status). What are the similarities and dissimilarities between the two statistical approaches in operationalising subjective health?

Studies in the Caribbean found that age, marital status, crowding, sex of respondents, area of residence, income and illnesses were statistically correlated with subjective health[810-1753], which is validated by the current study. Even some non-Caribbean studies have found the aforementioned variables to be statistically associated with subjective health[79], indicating that dichotomising self-rated health status does not fundamentally change most of the socio-demographic, economic, and biological variables.

Examining data on married people by way of dichotomising self-rated health status, Bourne[25] found that men had a greater self-reported health status than women, and in the current study (non-dichotomisation of self-rated health status), males had a higher health status than females. On the other hand, in Bourne's work[25], he found in descending order self-reported illnesses, age, income and sex to be the only factors of self-reported good health while in the non-dichotomised study more variables accounted for health status. Nevertheless, ranking of the correlates were similar in both studies as in the current. The factors in descending order were self-reported illness, age, crowding, income, sex and the others, indicating the closeness of the statistical approaches. Married people are a component of the general populace and they have socio-demographic and economic experiences which differ from some unmarried people.

The literature showed that income is strongly correlated with self-rated health. However, in Jamaica this is clearly not the case. In Jamaica, income plays a secondary role to illness and age and when self-rated health is non-dichotomised, it becomes an even weaker variable. Although income affords one particular choices (or lack thereof), the educated class in Jamaica received more income than uneducated classes, yet the former class is not healthier than the latter. This finding is contrary to the literature that showed the association between higher education and health[7-9]. Education influences social standing and income, but it does not directly influence good health status in Jamaica. Concurringly, the current work found that education is positively correlated with more health insurance coverage. However, health insurance coverage is not significantly associated with better health status. Embedded here is the fact that health insurance coverage in Jamaica is not an indicator of health care-seeking behaviour but a product that is purchased for the eventuality of the onset of illness, as it will lower out-of-pocket medical care expenditure.

Education provides its recipients with knowledge, access to knowledge, access to income and other empowerment, but it does not mean that the educated classes are more concerned about their health, and this can be measured using health care-seeking behaviour and knowledge about the illnesses that are affecting the individual. The current paper found that 25 out of every 100 educated Jamaicans are aware of their health condition(s), and this is greater than that for uneducated classes. Jamaicans with the least level of education were most cognizant of their ailments and sought medical care just as much as did educated Jamaicans. Education, therefore, does not denote empowerment to seek medical care, which is embedded in the culture, in particular for men. Education is still unable to break the bondages of the perceptions of society which purport that health is weakness, and that to display weakness as a man removes his masculinity. This continues to shackle Jamaicans, particularly men, who still subscribe to the traditional notion that illness is correlated to weakness and that men should not display weakness. It is this cultural perspective that bars many men from visiting health care facilities, except in cases of severe illness or if they are married[25]. Hence, mortality being greater for men is not surprising[54] as many men will die prematurely because of the fact that they are reluctant to visit health care institutions. This reluctance to seek medical care is not limited to males. In 1988, when Jamaica began collecting data on the living conditions of its people, females sought more medical care than males, but the disparity ranged between -2 to 6%. In 2007, 68% of females sought medical care compared to 63% of males, which means that higher education, which is substantially a female phenomenon in Jamaica, is not fundamentally improving the health status of females or even males.

Educated Jamaicans are more likely to live in urban areas and those with primary education levels or below are more likely to live in semi-urban zones. The current findings found that semi-urban respondents were more likely to have better health status, although they are more likely to have at most primary level education. In 2007, statistics revealed that 15.3% of Jamaicans in rural areas were below the poverty line compared to 4% of semi-urban and 6.2% of urban Jamaicans[41], indicating that poverty is more synonymous with rural areas, yet there is no significant statistical difference between the self-rated health status of rural and urban Jamaicans. Income makes a difference in health, as those with more means can access more and greater resources including health care, but clearly income beyond a certain amount is retarding the health status of Jamaicans.

This study cannot stipulate a baseline income that people should receive in order to prevent a decline in health status. However, there is clearly a state of contentment among the poor and very poor who were equally as healthy as the wealthy. The health disparity between them and the educated showed no significant statistical difference and this emphasises that wealth does not automatically transfer itself into health. Another issue which is evident in the data is the variability in the measurement of health among the social classes, as the poorest 20% reported less illness than the wealthiest 20%[41], yet the former group still dwells in slums, inner-city neighbourhoods, and violent communities and they have lower levels of education. Despite Diener's findings[34] that the variance is minimal, Bourne's work showed a strong association between subjective health (i.e., self-reported illness) and life expectancy – a correlation coefficient between 50 and 60% for a single variable is strong. However, this highlights that there are still some challenges embedded in the use of self-rated health status.

Conclusion

While the dichotomisation of self-rated health status loses some of the original data, when self-rated health is non-dichotomised, socio-demographic and biological variables accounted for 33% of the explanation of the variance and this was 44% using dichotomisation for married Jamaica, suggesting dichotomisation of health status still holds some validity. Another critical finding that emerged from the current work is that education is not improving the health status of Jamaicans. However, it is correlated with better social standing and higher income. Income is significantly associated with better health status and it played a secondary role to self-reported illness and age of respondents. Education is associated with more health insurance coverage, but that health insurance coverage cannot be used to measure health care-seeking behaviour or measure better health status of Jamaicans. In summary, there is a need for a public health care campaign that is specifically geared towards the educated classes as their educational achievement is not translating itself into better health care-seeking behaviour and health status than the uneducated which suggests that societal pressures are barring Jamaicans from better health status choices.