Loading...
Please wait, while we are loading the content...
Similar Documents
Organization ’ s Study on Global Ageing and Adult Health ( SAGE ) Wave 1 : Health states and objective health measures in six low and middle income countries
| Content Provider | Semantic Scholar |
|---|---|
| Author | Chatterji Somnath Kowal Paul Verdes |
| Copyright Year | 2011 |
| Abstract | Over 60% of the world's population aged 60 years and older are currently living in less developed countries (UN 2009), with an average annual growth rate almost three times that of more developed countries (2.5% versus 0.9%). However, data on health outcomes of older adults that can be directly compared in these countries is lacking. A key question, as these populations age, will be their health status over time and across cohorts. The implications for health policy planning and resource allocation decisions will be critical depending on their needs. WHO’s Study of Global Ageing and Adult Health (SAGE) provides an important data collection platform to monitor health and health-related outcomes and their determinants in these populations over time. We report data from SAGE Wave 1 and present results comparing self-reported and measured health status in six low and middle income countries. We also describe strategies to adjust for systematic reporting biases and evaluate the performance of these methods. There were identifiable reporting biases, and the methods used to correct for these biases adjusted mainly for country-reporting biases. Clear declines in health by age and differentials by socioeconomic status, gender and residence were evident. As these cohorts are followed, issues related to the compression of morbidity can be addressed more systematically. INTRODUCTION Over 60% of the world's population aged 60 years and older are currently living in less developed countries (UN 2009), with an average annual growth rate almost three times that of more developed countries (2.5% versus 0.9%). This growth in the older population is occurring in parallel with increasing income inequality, disparities in access to health care and social support systems, and widening health gaps as a result of complex disease burden patterns and globalization of health risks. In many less developed countries, these issues are compounded for individuals by a lifetime of accumulated health risks associated with poverty and inadequate access to health care. Yet, few countries have the age-specific health and functioning data necessary to determine basic health parameters, much less which morbidity trajectory their respective aging populations are following: expansion (Gruenberg 1977, Scheider and Brody 1983), compression (Fries 1980, Fries 2003) or dynamic equilibrium (Manton 1982, Manton 2006). The World Health Organization, with support from the US National Institute on Aging, has created a longitudinal data collection platform to generate cross-nationally comparable health data for the purposes of monitoring population health over time. WHO’s Study on Global Ageing and Adult Health (SAGE) is a nationally representative longitudinal cohort study in six low and middle income countries (LMICs), China, Ghana, India, Mexico, Russia and South Africa, to measure health and health-related outcomes, and their determinants, and to understand the relationship between health and well-being over time. One of the major challenges comparing self-reported health status across populations is that respondents often have systematic reporting biases. Typically, the measurement of health state relies on self-reported responses in surveys and the resulting data take the form of ordered categorical (ordinal) responses. However, the way people report their own health varies systematically with factors such as education, sex, age, or other cultural factors. Various people use different response category cut-points across cultures or population sub-groups, and this 'response shift' implies that self-report categorical data are not comparable across individuals (Murray 2003). Consequently, these responses on ordinal scales cannot be directly used to measure health states without adjustment. In order to ensure that data from self-reported interview surveys are truly comparable, four essential steps need to be followed: 1) agreement on a common conceptual framework for the measurement; 2) agreement on a measurement strategy that identifies a parsimonious set of domains and items for measurement; 3) ex-ante harmonization of questions, response scales and calibration strategies across languages and population groups; and, 4) ex-post harmonization of data using the calibrated responses. The first three of these approaches are most often employed in international efforts at data collection but, though necessary, are often not sufficient. The focus of this paper is strategies that can be used for ex-post harmonization. Statistical methods must be devised that detect systematic reporting biases, determine factors that are responsible for these systematic biases and correct for these in order to approximate the true underlying level in the latent quantity of interest. In recognition of this, the WHO World Health Surveys (WHS), used a set of questions across a core set of domains to measure health states, and employed vignettes to detect and correct for biases in self-report in order to adjust for response category cut-point shifts (Ustun 2003). An anchoring vignette is a brief description of a concrete level on a given health domain, with multiple vignettes used per health domain reflecting different levels on a latent scale. Respondents are asked to respond to a vignette using the same questions and response scales applied to self-assessments on that domain, as if the person described in the vignette is like the respondent herself / himself. Vignettes fix the level of ability on a domain so that variation in categorical responses is attributable to variation in response category cut-points. There are two key requirements for the use of anchoring vignettes: response consistency: this requires that an individual will have the same cognitive mind set when evaluating hypothetical scenarios as when providing a self-assessment; and vignette equivalence: the requirement that the underlying domain levels represented in each vignette are understood in approximately the same way by all respondents, irrespective of their age, sex, income, education, country of residence or other characteristics. We use data from SAGE to assess declines in health with age in six LMICs and the impact of systematic reporting biases. We compare results across age, socioeconomic variables and countries. DATA AND METHODS Data from SAGE Wave 1 on adults aged 50 years and older, from China, Ghana, India, Mexico, Russia and South Africa, were used in this analysis. SAGE is a household survey that interviewed respondents drawn from a nationally representative frame. All respondents had a known non-zero probability of selection. The survey instrument collected information on sociodemographic variables such as age, sex, place of residence, education, marital and economic status, and employment. Health status was assessed in eight domains of functioning: affect, cognition, interpersonal relationships, mobility, pain, self-care, sleep and energy, and vision. Respondents were asked the extent to which they have difficulty in, or experience problems with, carrying out a task or action. Two questions are asked about each health domain. In addition, respondents were also asked to respond to a set of vignettes about the health domains where they were asked to imagine they are the person described in the vignette, and then answer the same questions that were asked about their own health status for that particular domain. Responses were ordered in the same manner as for the self-report from no difficulty (or problem) to complete difficulty (or problem). In addition to self-reported health status, we also used data from the performance tests used in SAGE. Tests of cognition included a measure of verbal fluency assessed using a category naming task of animals, a word list for immediate and delayed recall, and digit span forwards and backwards for assessment of working memory. Normal and rapid walking speed and grip strength were also measured. A combined score on these tests was created using factor analysis to get an overall combined measure of performance. A composite health status score was created using a partial-credit model using graded response based on Item Response Theory. Data from the vignette responses were used to adjust for the self-report and to identify shifts in cut-points. A score on the latent trait was created using the Binormal Hierarchical Ordered Probit based on the Compound Hierarchical Probit model (Tandon 2003). RESULTS Background characteristics of the sample are shown in Table 1. Vignette results by age groups (18-49, 50-64, 65+) are presented in Figure 1, showing robust results with no systematic reporting differences by age. The expected trends in vignetteadjusted health scores by sex, age, education level and income level are evident across all countries. Performance measures in three domains (grip strength, timed walk and cognitive function) are presented. Initial analyses suggest a correlation between the vignette-adjusted health score and each performance test, but curiously not significantly different by age. A weak negative correlation was found between one health domain (mobility) and timed regular-paced walk at all ages, with a moderate positive correlation with grip strength (0.2-0.3) and cognitive function (0.3-0.4). These correlations were not significantly different by age group. The presence of depression, however, does create systematic reporting biases in health by age and education levels. More educated depressed respondents and younger respondents systematically underestimated their health, whereas, older depressed respondents systematically overestimated their health. Figure 2 shows the estimated median cut-points from the Binormal Hierarchical Probit model (BiHOPIT) using the vignettes data and controlling for a range of covariates such as age, sex, education, income, and other self-reported health variables such as stress, satisfaction with health, depression and c |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://paa2011.princeton.edu/papers/111612 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
