Disability patterns among Italian adult population: a multistate - - PDF document

Disability patterns among Italian adult population: a multistate approach Cristina Giudici(Sapienza University of Rome), Nicolas Brouard (INED), Lucia Coppola (ISTAT), Eleonora Trappolini (Sapienza University of Rome). Italian population living with severe functional limitations was about 3.2 million in 2015, most of whom were aged 60 and over. As age increases, so does dependency: about 9 % among people aged 70 to 74 years, 43% over age 80. Women are over-represented among people with functional limitations, impairments or severe chronic diseases, mainly because they are characterized by a higher prevalence of diseases that have a major impact on the functioning and quality of life. While life expectancy continues to increase in all European Countries, studies in disability trends have not shown such a clear improvement, resulting in increases, decreases or stagnation in disability prevalence

• ver time, and the debate on health expectancies of European populations is still open (Olshansky et al.

1991; Mathers 1997; Robine et al. 1998; Waidmann&Manton 2000; Robine& Michel 2004; Egidi 2013; Jagger 2016). The actual economic crisis has further complicated this debate, and most studies focusing on this issue show how the effects of the crisis may affect the health behaviours of populations differently. The recession could directly and indirectly affect health in the medium period, because it could have an impact

• n reducing consumption related to food and health (Walker, 2010). This effect may be complex; in

focusing on developed countries, some researchers claim that in the short run the recession might actually lead to improvements in health, because individuals could possibly modify their unhealthy habits (Gerdtham, 2006; Rhum, 2008; Lonides, 2008, Suhrcke et al., 2012). However, several studies conducted on a wide range of high income European countries have found a noteworthy increase in suicide and the evidence generally suggests that unemployment and financial insecurity is linked to increased risks of mental health problems (WHO, 2014). Even if it is difficult to highlight the short term effects of the recession on health (Karanikolos et al., 2013; Escolar-Pujolar et al., 2014; Stuckler et al., 2011), some national and international studies in Europe have shown recent changes in individual behaviours, in terms of prevention and use of healthcare services (Zavras, 2013; Gili et al., 2013; De Vogli et al., 2013) that could affect the future health status of the

• population. The aim of this study is to analyse the Italian pattern of disability, employing data from the

2010-2013 longitudinal release of the Italian “Statistics on Income and Living Conditions” survey (EU-SILC), carried out by the National Institute of Statistics. Data and methods EU-SILC sample design is based on a four year rotational panel: every year, longitudinal information are provided by three independent panels (named P1, P2 and P3 in Figure 1), followed during 2, 3 and 4 years

• respectively. Each panel is representative of the national population at the first wave and its changes during

the period of observation. This sample design allows for a flexible use of the panels depending on the estimates of interest. On the

• ne hand, a single panel may be used to achieve estimates on four years transitions, relying on a relatively

small sample size (e.g. panel 1 is based on a sample of about 7400 individuals aged 40 or above in 2010, and 4600 individuals in 2013: the reduction in the sample size is due to natural events as death or migration, and mostly to attrition). On the other hand, different panels may be pooled together to rely on a bigger sample size, but a shorter period of observation has to be selected. We prefer to focus on two years transitions (2012-2013) and three panels (P1, P2 and P3) are pooled together, providing a sample size of

16590 individuals (i.e. those interviewed in 2012 and who in 2013 are either interviewed or died - see table 1). Figure 1 – EU-SILC Sample design W1 P4 W1 W2 P3 W1 W2 W3 P2 W1 W2 W3 W4 P1 2010 2011 2012 2013 Table 1 shows that 72.68% of the individuals interviewed in 2012 provide information on limitations or died in 2013 (respectively 71.6% and 1.08%). About 4% of individuals did not provide any information about limitations in 2012: being a sensitive information, respondents are allowed to refuse to answer and missing values are not imputed. Moreover, no information is available in 2013 for the remaining 23.65% of the sample (i.e. 14.7% and 8.95% of non-limited and limited individuals respectively). Individuals with missing values, named “attritors” from now onwards, cannot be included in the analysis. If attritors differ from non- attritors, bias estimates may be achieved. We observe much higher levels of strongly limited individuals among attritors, for both men and women, also when using longitudinal weights to control for self-

• selection. In principle, the EU-SILC Used Data Base provides other longitudinal weights to be used for

estimating transitions between the last two years of the longitudinal component, but they cannot be used in our analysis because we need to assign a weight also to individuals who die between 2012 and 2013. As a consequence we estimated ad-hoc weights for non-attritors, correcting for the selection due to attritors. Then we estimate the parameterized transition probabilities between 2012 and 2013 following the Interpolation of Markov Chain approach - IMaCh , proposed by Laditka and Wolf (1998) and developed by Lièvre and Brouard (2003), which has been applied in several analyses dealing with health. We calculate the age-specific forward and (recently developed in version 0.99 of IMaCh) backward period prevalence; the first corresponds to the prevalence which will be observed in a younger cohort if the transitions rates between the various disability states as well as mortality, remained constant in the future. For example, a lower period prevalence of disability at old ages compared to the corresponding cross- sectional prevalence indicates that disability is currently declining. On the other hand, the backward prevalence is the prevalence which would have been observed in the past at younger ages, in an cohort reaching an old age today. In comparison to the forward prevalence which depends only of the transition rates, the backward prevalence depends also of the cross-sectional prevalence. With the ageing process and changing conditions, the cross-sectional prevalence will usually rely between the backward and forward prevalences, showing what was the past pattern and what will be the future pattern.

Table 1 - Distribution of individuals (age 40+) interviewed at the baseline, by state at the beginning and end

• f the period under examination

2012-2013 Total % Total with ad hoc weights % Disability-free at both dates (1-1) 8243 36.11 10997 48.2 Disability-free to disability (1-2) 2010 8.81 2665 11.7 Recovered from disability (2-1) 1783 7.81 2521 11.0 Remained disabled (2-2) 4307 18.87 6271 27.5 Died from disability-free (1-3) 64 0.28 74 0.3 Died from disability (2-3) 183 0.80 299 1.3 Total in analysis 16590 72.68 22826 100.0 Missing from disability free 3356 14.70

• Missing from disability

2042 8.95

• Information on health missing at the

base line 838 3.67

• Total missing

6236 27.32

• Total

22826 100,0 22826 100.0

Results For each age we estimated the probability of transition between health states for different covariates and for the total sample. As expected, the probability of entry into disability increases with age, and it is higher among women than for men. On the other hand, the probability of recovery decreases with age, and it is higher among men than for women. On the basis of transition probabilities, we estimate life expectancy at all ages. Figure 2 shows the comparison between our estimates and those based on national statistics. People aged 70 can expect to live 6.74 years in disability-free state, given that they were in that state initially, but the expectation is reduced to 5.6 years if they were in the disabled state at age 70. The corresponding health expectancies for the disabled state are 10.9 and 11.7 years respectively. Figure 3 and 4 show the estimation of the age-specific cross-sectional prevalence of disability (from the first wave of the survey) compared to the "period" prevalence, or forward period prevalence (computed from the age-specific flows observed between two waves), for men and women. The forward prevalence corresponds to the prevalence which will be observed in a cohort if the transitions rates between the various disability states as well as mortality, remained constant in the future. The two curves clearly

• verlap, indicating that disability isn’t currently increasing or declining for both sexes.

Figure 4 compare cross-sectional and forward with backward prevalence of disability for Southern regions: backward prevalence is the prevalence which would have been observed in the past at younger ages, in a cohort reaching an old age today; it describes the former situation of current population. The proximity of the backward prevalence with the cross-sectional prevalence let us verify that the pattern of a reduced prevalence of disability for younger adults prevailed in Italian Southern regions since many years, while the distance of both curves from the forward prevalence tends to prove that these exceptions will disappear in a very near future.

Figure 2 – Life expectancy at different ages with 95% confidence interval and comparison with life expectancy obtained from vital statistics* Figure 3 - Stable and observed prevalence with 95% confidence interval for men (left) and women (right) Figure 4 - Stable and observed prevalence with 95% confidence interval for Southern Italian regions

5 10 15 20 25 30 35 40 45 50 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 ISTAT 2012 Silc 2012-13 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 Observed prevalence Stable prevalence 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 Observed prevalence Stable prevalence

References

Brouard N (1980) Espérance de vie active, reprises d’activité féminine: un modèle. Revue économique 31:1260–1287 Brouard N and Shih YC. Forward and backward prevalences: changes in living arrangements observed with the US Health and retirement surveys. Communication at the REVES Conference 8 June 2016 Burricand C. et al. (2014). EU-Silc / SRCV, ambition et apport d’une approche longitudinale et transversale des revenus et des conditions de vie, en France et en Europe, Economie et statistique, 5 (18), pp. 469-470. Cambois E, Robine JM, Brouard N (1999) Life expectancies applied to specific statuses. A history of the indicators and methods of calculation. Popul Engl Sel 11:7–34 De Vogli R. (2013). Financial crisis, austerity, and health in Europe, Lancet, pp. 382-391. Gili M. et al. (2013). The mental health risks of economic crisis in Spain: evidence from primary care centres, 2006 and 2010, European Journal of Public Health, 23(1), pp. 103–108. Giudici C., Arezzo MF., Brouard N. (2013), Estimating health expectancy in presence of missing data: an application using HID survey, Springer, p. 518. Hoem J, Fong M (1976) A Markov chain model of working life tables. Working paper 2 Laboratory of Actuarial Mathematics, University of Copenhagen Lièvre A, Brouard N, Heathcote CR (2003) The estimation of health expectancies from cross-longitudinal

• surveys. Math Popul Stud 10:211–248

Rogers A (1975) Introduction to multi regional mathematical demography. Wiley, England Willekens F (1979) Computer program for increment-decrement (multistate) life table analysis: a user’s manual to lifeindec. Working papers of the international institute for applied systems analysis Zavras D. et al. (2013), Impact of economic crisis and other demographic and socio-economic factors on self-rated health in Greece, European Journal of Public Health.