Page 20 - IJPS-4-2

Trends in healthy life expectancy among older adults in South Africa

self-rated health (good, fair, and not good), data from the United Kingdom showed that HLE among older men increased
more than older women from 1981 to 2001 (UK National Statistics, 2006). In China, improvements in self-rated health
based on two or three categories were comparable for men and women in the late 1980s (Saito, Qiao, and Jitapunkul,
2003) and the 1992–2002 (Gu, Dupre, Warner et al., 2009). No studies on gender difference have been conducted in South
Africa so far, which implies a need to examine the difference in self-rated health and HLE.
This study intends to explore the health state of older people in South Africa using a national HIV household dataset
collected by the Human Sciences Research Council of South Africa. To the best of our knowledge, this is the first study
to estimate health expectancies for South Africa using repeated cross-sectional surveys. The objective of this paper is to
examine trends and investigate gender differences in HLE for older people in South Africa for the period 2005–2012. The
research questions posed by this study are: Has the recent increase in total life expectancy (TLE) observed in South Africa
been accompanied by an increase or decrease in health status among older people? Is there any difference between women
and men? In answering these two questions, and based on the data on the elderly population of South Africa, we will also
be simultaneously testing the hypotheses of compression or expansion of morbidity.

2. Methods

2.1. Data sources

The study was based on 2005, 2008, and 2012 waves of the South African National HIV Incidence, Prevalence, Behaviour
and Communication Survey (SABSSM) conducted in South Africa by the Human Sciences Research Council (HSRC).
These are repeated cross-sectional surveys aiming at surveillance of HIV incidence and prevalence in South Africa. The
individual response rates for each survey were 96.0% in 2005 (Shisana, Rehle, Simbayi et al., 2005), 89.5% in 2008
(Shisana, Rehle, Simbayi et al., 2009), and 89.1% in 2012 (Shisana, Rehle, Simbayi et al., 2014). All three surveys
included persons residing in community dwellings and aged 2 years and above. Institutionalized individuals (i.e., those
in educational institutions, military barracks, old-age homes, or hospitals) were not included in the three surveys, and
hence as a result they were also excluded from this study. The surveys included a multistage cluster sample stratified by
province and settlement geography (genotype) with the predominant population group in each area used. In our analysis,
the design characteristics of the three surveys were adjusted using the weighted prevalence rates. Further details about
the sampling procedures were presented elsewhere for 2005 (Shisana, Rehle, Simbayi et al., 2005), 2008 (Shisana, Rehle,
Simbayi et al 2009), and 2012 (Shisana, Rehle, Simbayi et al., 2014). A questionnaire was administered through face-to-
face interviews conducted by trained fieldworkers. Sociodemographic and behavioral information were collected from
consenting individuals. The surveys were approved by the HSRC Research Ethics Committee (REC). We restrict our
analysis to older adults aged 50 years and older with a total valid sample of 14,344 respondents, consisting of 5333 men
and 9011 women, 3795 from the 2005 wave, 2702 from the 2008 wave, and 7847 from the 2012 wave.

2.2. Analysis
HLE were calculated using the Sullivan method (Sullivan, 1971). This method utilizes the age-specific prevalence of
different health states in a population at a certain point in time to calculate the person-years lived in the respective health
states among life table stable population. The next step is to derive the total person-years lived by summing up the person-
years lived from age x upward until the last age group in the life table. The total number of person-years lived is then
divided by the number of survivors to obtain the HLE at a given age.
In the present study, the HLE is calculated based on a self-rated health measure obtained from the 2005, 2008, and
2012 waves of the SABSSM described above. In the three SABSSM waves, self-rated health was measured by self-
assessed global health condition using a question “In general, would you say that your health is excellent, good, fair
or poor?” The same wording of the question was used in the 2005, 2008, and 2012 surveys, which makes it feasible to
evaluate trends in population health. Following the common practice in the field (Idler and Benyamini, 1997), a binary
variable was created by categorizing; excellent and good as “good health” and fair and poor as “poor health.” Multiplying
the age-sex-specific prevalence of “good health” to the corresponding number of person-years lived with the given age
range in the life table to obtain person-years in good health, and then accumulated them from a given age onward and
divide by total survivors at a given age to obtain its HLE.
As previous research evidenced that the categorization of self-rated health may yield different outcomes for HLE
(Doblhammer and Kytir, 2001; Gu, Dupre, Warner et al., 2009). A sensitivity analysis was performed by treating outcomes
as excellent/good/fair as a group, which produced a comparable result. The estimates of life expectancy used in this study

14 International Journal of Population Studies | 2018, Volume 4, Issue 2