A new measurement of health encompassing several dimensions of health

Get Complete Project Material File(s) Now! »

Measures of health status


Any empirical studies on health, especially those concerning inequalities, have to rely on a measurement of health. Survey data offer various health indicators all of which have different properties and so choosing which one to use is not straightforward.
Moreover, several recent studies have questioned the pertinence of the individual health information contained in health indicators.
Prior to the analyses that are proposed in this thesis, we intend to examine the measurements of health status which are involved in empirical studies.
This review is particularly relevant because economists tend to focus on the general rather than on the particular aspects of health. However, besides self-assessed health, there are no general health variables in most of the health surveys. Nevertheless self-assessed health is prone to dispute. Opinions have been divided on the use of self-assessed health for many years. Indeed, its validity for measuring health correctly has often been questioned in literature and evaluated by comparison to other health measures. As a result, several recent studies have dealt with individual variability in self-assessments of health and the existence of declaration biases related to individual characteristics.
This chapter is organised as follows: following a definition of what is a good indicator of general health, the first section presents from a content point of view the different health indicators which have been used in literature as proxies of “true” health. The term of “true” health is often used in literature to define the latent health. Then, as self-reported health is a widely used measure for general health, we question the use of this variable and point out its advantages and limitations in literature, as well as its correlations and discrepancies with other health indicators. The last section considers better ways to measure health, which consist of either improving available indicators or providing new indicators.

Health indicators: content, correlations and discrepan- cies

We briefly list some health indicators used as proxies of “true” health in literature. The aim of this subsection is to underline the wide range of health indicators used in empirical studies.

What is a good indicator of general health?

While many facets of health can be identified, such as functional abilities, biomedical status or emotional characteristics, the assessment or measurement of individual health must take all of these into account. However, there is no single measure or one-dimensional measurement scale for the health of an individual. At best, public health professionals rely on crude health indicators coming from data collection. The term health indicator refers to a single summary measure, expressed in quantitative or qualitative terms, and which represents a key dimension of health status, health care or other related factors.
As we cannot expect a health indicator to be objectively measured (as a temperature, for example), researchers postulate the existence of a latent health variable. In literature, several variables have been regarded as drawing closely on “true” health status, which is the latent health variable. Therefore, the key to measuring health is to be able to access the relevant health information (Knauper¨ & Turner, 2003). It is common to think that mortality or life expectancy are measures of health. However, surveys offer many other health indicators at individual level that have also been used as good proxies for “true” health. This is the case for indicators based on questions related to ability to carry out daily living activities, self-reported ailments, as well as height and body mass index, which are relatively easy to obtain, or physician-assessed health status or clinical interviews, which are more costly to obtain. Furthermore, mortality fails in the evaluation of the influence of recent social changes but are sensitive to changes in prevention, diagnosis and undertaking to reimburse medical expenses of diseases (Chauvin & Lebas, 2007).

Various health indicators to measure “true” health Mortality or survival

Mortality is considered to be a measure of health which is not based on a personal assessment of health (Anderson & Burkhauser, 1985). In concrete terms, it is obtained by death counts and related rates, and is sometimes defined by age, sex or sub-groups of population. Survival in elderly cohorts has been found to be a very good health mea-sure. In the absence of such cohorts, in some longitudinal household surveys the death of respondents can be identified in national death registrations using first and last name, month/day/year of birth, sex, race or social security numbers and this is how a mortality measure is obtained for the sample (Franks et al., 2003). There are, however, no such longitudinal surveys in France.
Mortality has been widely used as a measure of health (Grossman, 1972; Parsons, 1980) because it is a measure available for all respondents and it is also independent of a respondent’s reporting biases. However, there is a debate on the validity of mortality as an appropriate health measure (Haveman & Wolfe, 1984). In fact, mortality is not a perfect measure of health as some deaths occur suddenly and independently of health status. Moreover, some health problems affecting ability can induce a particularly bad health status, but the individual life prognosis is not reduced.
Clinically observed morbidity According to Murray and Chen (1992), morbidity can be observed through four types of indicators: the first category refers to physical and vital signs, directly related to the presence of a particular disease; the second to physiological indicators (such as labora-tory exams and radiography); the third to functional tests (for instance, ability to carry out daily activities), and the fourth to clinical diagnosis. All four categories require a professional’s diagnosis or examination which is why few surveys contain them.
In a shorter health care professional’s assessment, Bartel and Taubman (1979) describe an ideal health indicator constructed from the presence or absence of a doctor’s diagnosis of particular diseases. Nevertheless, authors underline some issues in their definition of this “ideal” health measure. Firstly, severity, cures and remissions of the disease are not considered. Secondly, a person can be ill without being so diagnosed. Then, the diagnosis requires a health care utilisation. Finally, a diagnosis can subsequently be proved wrong.
Baker et al. (2004) also consider self-reports of chronic diseases as a good proxy of “true” health and rely on medical records of the incidence of thirteen ailments1 to construct a “true” health status. Therefore, they investigate the validity of self-reported ailments and self-reported global health by comparing with medical records. However, although it is possible to do so with their data, generally few surveys look at physiological or physical performances. There are several reasons for this: insufficient survey time, cost of obtaining information as it involves a health professional, logistical difficulties of interviewing in houses and difficulties of doing it again for subsequent groups. In any case, when these health indicators of medical reports are available they mainly concern developed countries.
In literature, due to a lack of an aggregate health measure or clinically-assessed health measures, self-reported morbidity variables are used in analyses. They are individual reports concerning functional characteristics, health-related behaviour and health-related characteristics. For instance, Groot (2000) considers as “true” health status responses to a question on health problems and disabilities in reference to a list2.
In labour market participation, reported functional limitations have been used as proxy for “true” health (Bound, 1990; Disney et al., 2004). They can be instrumental activities of daily living and disabilities in activities of daily living (respectively IADL and ADL).
As far as we know, few studies consider health-related behaviours and characteristics as a unique health indicator. There is, however, a study on Bangladesh (Kuhn et al., 2004), which proposes to use body mass index as a proxy of “true” health. The BMI is considered as an indication of poor nutritional health status (BM I < 16), and is strongly associated with increased mortality for women. However, it is not associated with any other health measures such as self-reported health, activities of daily living, physical disability or self-reported chronic morbidity.

Health care utilisation

As visits to the doctor and hospital data are factual and countable, they have been used as proxies for “true” health status. The impetus for using these indicators is that they would be less influenced by an individual’s perception. Indeed, considering the Ontario Health Insurance Plan (OHIP) records, Baker et al. (2004) rely on health status measure by numbers of care contacts and associated diagnosis.

Another health indicator: the paradoxical self-assessed health

Self-assessed health is widespread in surveys, widely used in health studies and is sometimes considered as a valid indicator of “true” health (Butler et al., 1987).
Compared to a medically-assessed health status, defined by health care professionals disregarding subjective self-assessments by patients, an individual’s self-reported health assessment is the result of a more complex aggregation process. This process is based on his observed morbidity (defined by the number of illnesses, his physical performance, disability, and treatments prescribed), his health expectations and his interactions with 2“Do you have any of the health problems or disabilities listed on this card ? (exclude temporary con-ditions)”. List of conditions on the card: problems or disabilities with arms, legs, hand, feet, back, neck, difficulty in seeing; difficulty in hearing; skin conditions/allergies; Chest/breathing problems, asthma, bron-chitis; Heart/blood pressure or blood circulation problems; Stomach/liver/kidneys or digestive problems; Diabetes; Anxiety, depression or bad nerves; Alcohol/drug related problems; Epilepsy; Migraine or frequent headaches; Other health problems.

Health indicators: content, correlations and discrepancies 

health care professionals, but it is also part of his social, cultural and health knowledge en-vironment. Nevertheless, self-assessed health is always compared to other health variables to confirm its health content. The following part aims to highlight results on this ability to proxy “true” health. As evidence is given that self-reports also contain information about respondents’ own characteristics (education, standard of living, interaction with the health system) and beliefs about what good health is, disentangling these elements from the “true” health status definition is not straightforward (Thomas & Frankenberg, 2002). This is the reason why the increasing use of self-assessed health in empirical studies has also given rise to a wide debate concerning individual bias that self-assessed health could suffer from.
The following two sections respectively highlight results in literature which uphold self-assessed health as a good proxy of health status and those which criticise its use.
Why is self-assessed health used as a proxy of “true” health?
In literature, self-assessed health is widely considered by comparison with health indicators in order to emphasise its properties in terms of health content. Although self-assessed health questions were at first introduced because medically-assessed health status measures were costly to obtain, they increased in popularity because they are strongly correlated with other health indicators used to proxy “true” health.

Self-assessed health predicts survival.

Self-assessed health is a good predictor for survival. For instance, by comparing two points in time, Burstrom¨ and Fredlund (2001) observe a good prediction of self-rated health on subsequent mortality among adults. Those who rate their health as poor are found three times as likely to die (Long & Marshall, 1999). Using the NHANES data set3, Seibt (1998) finds that respondents having many contacts with doctors and regular health check-ups provide more valid health evaluations as assessed by the correlation between self-reported health and length for survival. Moreover, most of the studies support these results irrespective of the socioeconomic conditions (Kaplan et al., 1988; Idler & Benyamini, 1997; Burstrom¨ & Fredlund, 2001).
2. Self-assessed health is correlated with reported morbidity and medically-assessed health status.
Self-assessed health reflects various reported morbidity conditions. Indeed, Groot (2000) confirms a strong relationship between self-assessed health and chronic health conditions, specifically concerning related disabilities. As for diseases, Baker et al.
3The National Health and Nutrition Examination Surveys is conducted by the National Centre for Health Statistics, Centres for Disease Control. These surveys are designed to assess the health and nutri-tional status of adults and children in the United States through interviews and direct physical examina-tions. They have been carried out six times between 1971 and 2004.
(2004) show that all the diseases considered, with the exception of cataracts, are significantly correlated with self-reporting worse health. According to Verbrugge (1984), the most plausible reasons for self-reporting a worse health status are peo-ple’s greater awareness of their diseases, due to earlier diagnoses. Mechanic (1986) underlines this idea and argues that there may be personal predispositions in the perception of illness and bad health status: people who are chronically ill or get used to their diseases, could have a better perception of the subsequent vital risk and could approach closely their medically-assessed health status. Functional disability, particularly the degree which affects activities of daily living, is also central in the formation of subjective health in cross-sectional studies, whatever the age (Idler & Kasl, 1995; Idler & Kasl, 1991). A comprehensive study of rheumatoid arthritis em-phasises that greater activity restriction is associated with lower self-assessed health and provides supports for the idea that a positive change in disability level and es-pecially psychological well-being would have a positive effect on self-assessed health (Nagyova et al., 2005).

READ  Theoretical foundation to the study: symbolic interactionism

Self-assessed health is correlated with health care utilisation.

Health care use as well as health care cost are found positively associated with those who rated their health as poor (Long & Marshall, 1999; Ware, 1986; van Doorslaer et al., 2002). Long and Marshall’s analysis goes even further by calculating perfor-mance indices based on the ratio of actual-to-expected cost within each category of self-assessed health. These indices then suggest a more aggressive treatment of those who rate their health as poor. Furthermore, an important health care utilisation can both improve and worsen the self-reported health status. Improvement is explained by rapid treatment whereas worsening can be due to a higher number of diseases diagnosed. Nevertheless, the more one uses the health care services and has contact with health care professionals, the more one knows about his health issues and per-ceives his morbidity, allowing him to approach his physician-assessed health status (Baker et al., 2004).

Self-assessed health is a better predictor of health than other health in-dicators.

Although in literature self-assessed health is questioned on its health content by comparison to other health indicators, it has also been found to be a better indica-tion of “true” health than other health indicators. Indeed, the analyses of Krueger (1957) on the 1953-1955 Baltimore Health Survey reveal large discrepancies between self-reported morbidity and clinical diagnoses. Results show that the variance in causes of morbidity went from 2% for syphilis to 100% for rheumatoid arthritis. Fur-thermore, Idler and Benyamini (1997) underline that self-rated health contributes more to supplementary health information than other health indicators, even those determining mortality. Similarly, self-assessed health is found to be a very inclusive measure of health reflecting health aspects relevant to survival which are not covered by other health indicators (Mackenbach, 2002). Moreover, self-assessed health ex-tends the information obtainable from morbidity indicators by describing the quality rather than merely the quantity of functional abilities. It gives insights into matters of human concern such as pain, suffering or depression that could not be deduced solely from medically-assessed health or laboratory tests.

Self-assessed health expresses individual preferences.

Self-assessed health is comparable to a quality of life indicator as it focuses on peo-ple’s feelings about their personal circumstances. Life satisfaction refers to individual subjective assessment, such as self-assessed health, which is individually evaluated compared with a normative reference or according to an individual’s own aspira-tions. Moreover, this subjective health variable gives information about individuals regardless whether they seek care or not, and can thus reflect positive aspects of good health. From this point of view, biases inherent to subjective reports do not threaten the validity of the measurement process; health status or quality of life are such as the individual perceives them. This way of considering self-assessed health raises the political question of how to balance needs against an individual’s subjective demand.
Why is self-assessed health called into question?
Self-assessed health seems to be a good predictor of morbidity and mortality under the assumption that individuals rely on mortality and morbidity relevant information and ignore irrelevant information in their judgments. Nevertheless, as subjective health does not focus on a specific dimension of health, it encompasses strong emotional dimensions and self-reports can be distorted at various levels.
1. Self-assessed underestimates or overestimates “true” health according to morbidity conditions.
Doctors and individuals have dissimilar perceptions of an individual’s health status. Self-assessed health cannot reasonably be strictly equal to medically-assessed health indicators. However, an individual is expected to use only morbidity-relevant infor-mation to evaluate his health status. As health conditions are appreciated differently according to the burden of pathology or to the variations in illness perception, self-perceived health status may be far from “true” health. For instance, respondents who have the flu at the time of interview are likely to assess a health status differing from a more valid health judgment such as length of survival (Seibt, 1998). However, it is rather the psychosocial well-being related to the disease than their acute aspect that seems to influence self-reports of health. Indeed, depression has a negative ef-fect on perceived health even if the mortality risk is not higher, which is particularly observed with men (Rodin & MacAvay, 1992). As for the nature of pathologies or limitations, Groot (2000) picked out three health conditions which significantly in-crease the gap between a poor and a very poor health: difficulties in hearing; skin conditions and allergies; and heart or blood pressure or blood circulation problems. People with one of these three problems are inclined to say that their health is very poor even if these diseases do not increase mortality risk and are widespread amongst the population. The report of self-assessed health is thus influenced by the discomfort in daily life due to health problems. Indeed, self-assessed health is more influenced by the disability related to the chronic disease than by the chronic disease itself. Self-assessed health can also over-estimate “true” health because of chronic diseases or disabilities from birth, which could increase an individual’s tolerance of health difficulties. Therefore, an individual who is used to suffering, would assess a higher health better than peers even if they share similar pathological or functional health statuses. Analogous results are obtained in Ghana (Belcher et al., 1976), as people who miss body parts rarely report it as morbidity. They behave as if the loss of a limb is no longer a discomfort.
2. Self-assessed health suffers from individual’s response effects, which are independent of “true” health.
The correlation between self-assessed health and any individual variable can be re-stricted to two elements: firstly, an existing correlation between ”true” health and individual characteristics variable and, secondly, an individual’s appreciation for his “true” health. International studies (Bound, 1990) have stressed the difficulties en-countered in comparing levels of self-assessed health across individuals, who differ in terms of socioeconomic, demographic, pathological or cultural characteristics. Evi-dence is given in literature on variations between health indicators and self-assessed health related to socioeconomic and cultural characteristics and not necessarily dif-ferences in “true” health. A classic example in this respect concerns a famous Aus-tralian study (Mathers & Douglas, 1998), which observes that the Aboriginal people describe their health status as being much better than that of the general population, whereas they also experience the highest incidence rates of major health problems and other health indicators. This shows that there are individual characteristics that influence self-assessed health and take it away from “true” health. Various individual characteristics have been underlined in this context but, from one study to another, the influence of these variables on self-assessed health is of different magnitude or even sign.

Table of contents :

I Measuring health 
1 Measures of health status 
1.1 Introduction
1.2 Health indicators: content, correlations and discrepancies
1.2.1 What is a good indicator of general health?
1.2.2 Various health indicators to measure “true” health
1.2.3 Another health indicator: the paradoxical self-assessed health
1.3 Health status measures: can we measure “true” health in a better manner?
1.3.1 Towards a correction of individual response effects
1.3.2 Why is the reporting correction advised?
1.3.3 Another solution: providing global health indicators
1.4 Conclusion
2 A new measurement of health encompassing several dimensions of health
2.1 Introduction
2.2 Aggregating several dimensions of health to measure a general and cardinal health status
2.2.1 Why is the continuous aspect desirable?
2.2.2 How can we measure a continuous health variable?
2.3 A health assessment model
2.3.1 The model specification
2.3.2 A set of demographic, socioeconomic and health-related behaviour variables
2.3.3 Using individual characteristics to correct the drawbacks of self- assessed health
2.3.4 Construction of the health index
2.4 Empirical results
2.4.1 Ordered logit models with or without cluster effects
2.4.2 Ordered logit model with cluster effects and varying thresholds
2.4.3 The continuous health indicator
2.5 Comparisons with other constructions
2.5.1 Getting continuity from an “arbitrary” distribution
2.5.2 Getting continuity by combining different health dimensions
2.5.3 Getting continuity using external information
2.5.4 Some elements of discussion
2.6 Conclusion
II Measuring inequalities in health 
3 Measures of health inequality 
3.1 Introduction
3.2 Measurement of inequality in health: orderings and rankings
3.2.1 Stochastic dominance: first and second order
3.2.2 Multidimensional welfare analysis: symmetrical attributes
3.2.3 Multidimensional welfare analysis: asymmetrical attributes
3.2.4 Orderings and rankings: some elements of conclusion
3.3 Measurement of inequalities in health in a unidimensional context: the health Gini index
3.4 Measurement of inequalities in health in a bidimensional context
3.4.1 The pseudo-Gini
3.4.2 The concentration index
3.5 Conclusion
4 Income-related inequalities in health in France
4.1 Introduction
4.2 Health in France in 2004
4.2.1 The French health care system over the last decade
4.2.2 Health in 2004: a social health gradient in France?
4.3 Measuring inequalities in health: which measurement of health should be used?
4.3.1 New approach to measurement of health in Europe: an application to French data
4.3.2 Cardinalisation of self-assessed health: a reliable health distribution in France?
4.3.3 Innovative health index: a first empirical utilisation
4.3.4 Comparisons of the alternative mappings
4.4 Measuring income-related inequality in health
4.4.1 Measurement method
4.4.2 Explaining health within a linear model
4.4.3 Global concentration indices: income-related inequality in health
4.5 Explaining income-related inequality in health
4.5.1 Measurement method
4.5.2 Concentration indices over income
4.5.3 Contribution to the income-related inequality in health
4.5.4 Legitimate or illegitimate income-related inequalities in health?
4.6 Income-related inequality in health in 1998: a comparison with 2004
4.6.1 Measurements of health in 1998
4.6.2 Data and variables
4.6.3 Explaining health within a linear model in 1998: comparisons with 2004
4.6.4 Income-related inequality in health in 1998: comparisons with 2004
4.6.5 Decomposition of inequalities in 1998: comparisons with 2004
4.7 Conclusion
5 From inequalities in health to inequalities of opportunity in health 
5.1 Introduction
5.2 Equality of opportunity in health
5.3 The French part of SHARE: a relevant tool for empirical work
5.3.1 Data and sample
5.3.2 Variables measuring social conditions
5.3.3 Variables measuring health conditions
5.4 A first approach in terms of stochastic dominance
5.4.1 Dominance according to parents’ relative longevity
5.4.2 Dominance according to social background
5.4.3 Dominance according to current socioeconomic status
5.5 A second approach using regression analyses
5.5.1 Influence of social background and parents’ relative longevity
5.5.2 Influence of social background, parents’ relative longevity and cur- rent socioeconomic status
5.5.3 Endogeneity test of the social status in adulthood
5.6 A third approach using concentration indices
5.6.1 Measurement of health: cardinalisation of self-assessed health with SF6D
5.6.2 Inequalities in health related to parents’ relative longevity
5.6.3 A pseudo health concentration index according to parents’ socioeco- nomic status
5.7 Conclusion
Conclusion g´en´erale


Related Posts