Differences in labour participation between people living with HIV and the general population: Results from Spain along the business cycle.

BACKGROUND: HIV/AIDS (Human immunodeficiency virus/Acquired immune deficiency syndrome) not only has a strong impact on the health of the worldwide population but also affects the labour status of HIV-positive people. The primary aim of this paper is to compare the labour participation of people living with HIV (PlwHIV) with the labour participation of the general population along the last business cycle in Spain.
METHOD: The data used are from the Hospital Survey on HIV-AIDS, with a total sample size of 4,651 PlwHIV and the Labour Force Survey from 2001 to 2010, with a total sample size of 660,674 individuals as general population. Propensity Score Matching method was used to analyse the differences between the labour participation of PlwHIV and the general population. Additionally, several specific models categorised into different subgroups (gender, education, source of infection and level of defences) were also performed.
RESULTS: We identified a convergence in labour participation across the period in the two populations considered: PlwHIV was 23% less likely to have a job than the general population during 2001-2002 and 14% less likely during 2009-2010. This convergence is mainly explained by two facts: first, the positive evolution of people infected by sex; second, the change in the PlwHIV population composition with a decreasing weight of people infected by drug use throughout the decade. Thereby, at the end of period, there was no statistical difference in the employment rate between PlwHIV infected through sex and the general population but there was strongly difference in PlwHIV infected through drugs.
CONCLUSION: Inmunological status, source of infection and level of education play a relevant role among the PlwHIV population when comparing their labour participation with the general population. In spite of this positive result, the likelihood of being employed in HIV-positive people continues to be different from that of non-carriers. Our study shows that institutional features of labour markets are relevant and should be considered in comparison between countries.

Introduction

Labour participation has been identified as a significant social issue facing people living with HIV (PlwHIV). A notable characteristic of this population group is that the vast majority are of working age. More specifically, it was estimated that almost 90% of PlwHIV worldwide are between 25 and 49 years old [1].

Until the introduction of highly active antiretroviral treatments, disease progression would have driven out a vast majority of PlwHIV from the labour force [2-5]. Therapeutic advances since the 1990s have substantially increased life expectancy and improved the living standards of this population. Consequently, the labour participation of PlwHIV has clearly improved, increasing the likelihood of keeping a job and reducing the difference in the employment rate among PlwHIV and the general population [6-10].

Nevertheless, although therapeutic advances have gradually improved this population group’s labour integration, this does not mean that being diagnosed with HIV has no effect on labour participation. Several studies have suggested that the decline in the percentage of PlwHIV employed may be a rational consequence of adapting their expectations to a negative health shock that implies a shorter life expectancy [11]. Other papers have highlighted the association between health status and likelihood of employment. The immunological status but also the psychological aspects [12-18], as well as the importance of other diseases, physical problems or the presence of fragility associated with premature ageing are relevant aspects to explain the probability of employment [17, 19–21]. In addition to these findings, the literature agrees that in addition to health status, other socioeconomic characteristics, such as age, sex or level of education, are as important as the previous ones to explain the labour participation of PlwHIV [9, 10, 12, 14].

The positive progression of PlwHIV labour participation over the last decade could be affected by changes in business cycle. As literature has showed, the recent economic crisis had played a relevant role in health and labour markets for non-HIV people, especially for those with poor health status or with chronic diseases [22-27]. Moreover, a recent study has concluded that improvements in HIV care are not necessarily conducive to significant improvements related to the employment status of PlwHIV [28]. This finding could be due to the recent economic crisis. Despite advances in medical treatment, at a time of economic recession, the unemployment rate of PlwHIV in France has been affected by a greater extent than that of the general population [29]. The authors conclude that in time of economic crisis, a part of PlwHIV could have been excluded from the labour market.

It is important to stress that economic crisis in Europe has been asymmetric. The duration and depth of the crisis has been different across countries and their labour markets have behaved very differently. For example, in Spain, the economic crisis began in 2008 and, according to figures from the National Statistical Institute, concluded in 2014. By the end of 2016, the Spanish economy had not recovered GDP values prior to the crisis [30]. The Spanish labour market has been particularly affected by the crisis, as the unemployment rate has sharply increased (from 8% in 2007 to 25% in 2014), while employment has decreased from 2007 to 2014 [31].

The objective of our study is to compare the labour participation of PlwHIV residing in Spain with the labour participation of the general population along the business cycle. In this way, we want to test whether the result found by Annequin et al in France [29] is comparable with the Spanish crisis or if the asymmetry in the economic crisis and labour markets across countries leads to differential results. Thus, we will investigate whether the differences in employment rate between PlwHIV vs the general population have been increased or decreased over the last decade, comparing three different points of time: before the economic crisis (2001–2002 and 2006–2007) and during the first years of the economic crisis (2009–2010) in Spain. Conversely, another objective of our work is to identify profiles for which the differences in the employment rate between PlwHIV and the general population have decreased over the period studied and, on the contrary, profiles where the differences in labour have become more acute.

Methods

Data

Two different datasets were used to carry out this analysis, one to obtain information related to PlwHIV and another to acquire information regarding the general population. For the first group, the Hospital Survey on HIV/AIDS patients by the General Secretariat of the AIDS Plan [32] was used. The aim of this survey was to analyse the socio-demographic characteristics of PlwHIV. The survey was a one-day cross-sectional study (repeated on an annual basis) where information was collected using a questionnaire completed by the physician during or immediately after the patient’s consultation. Between 69 and 86 centres participated in this survey during the period under analysis with a total sample size of 4,651 PlwHIV. The target population was people with HIV/AIDS, diagnosed and treated in hospital outpatient departments, day hospital centres or hospital admissions. PlwHIV receiving hospital treatment for reasons unrelated to HIV were not included. The information collected by the Hospital Survey on HIV/AIDS contains clinical information on immunological status, such as the level of defences measured by CD4 cell count (CD4 count/μl≥500 indicates high level of defences, 200≤CD4/μl <500 indicates medium level of defences and 200≤CD4/μl indicates low level of defences) and most likely cause of transmission. There is also information related to demographic variables such as gender, age, educational level and employment status.

To acquire information about the employment status of the general population, the Labour Force Survey from 2001 to 2010, as performed by the Spanish Statistics Institute [33], was used. This survey was designed to provide detailed results on a national level. The total sample for the period under analysis was 660,674 individuals for the years selected. Performed every three months, the study considered the population living in family dwellings on Spanish national territory. The study’s main aim is to reveal the characteristics of that population group with regard to the labour market, collecting information, such as gender, age, marital status, education and other individual characteristics. Although the Labour Force Survey does not provide information on whether or not respondents are HIV-positive, since the prevalence rate of HIV among the general Spanish population was 0.3% for this period, we can assume a margin of error that the respondents are not HIV-positive [34].

Because the analysis is focused on labour participation, we only included individuals in their working age (16–64 years old) and PlwHIV who do not live in prison (that population reached just 3% of the total PlwHIV).

Variables

The main variables used in the analysis are; a dichotomous variable indicates whether individual works or not, gender, age (from 16 to 64 years old), level of education (no education, primary, secondary and tertiary education), and whether an individual is in treatment group or control group (1 if the individual suffers from HIV, 0 otherwise).

Statistical analysis

Propensity Score Matching method was used to analyse the differences in the labour participation of PlwHIV residing in Spain compared with the labour participation of the general population during three different periods of time: years 2001–2002, 2006–2007 and 2009–2010, using logistic regressions to estimate the propensity scores. Due to the treatment assignment (HIV-positive) being not a random process and the differences between individuals in the treatment and control groups were recorded with observable variables (such as age, gender or education), a matching method was therefore the most suitable method [35-37]. The idea was that individuals from the control group (general population) were matched to individuals in the treatment group (those with HIV); thus, the key assumption was that for each treated individual, an untreated individual could be found.

Since we observed either treated (HIV-positive) or untreated (general population) individuals, we were unable to note the causal effect of interest, Yi1-Yi0. We may thus consider the Average Treatment Effect on the Treatment Group (ATE): E(Y1 − Y0|X,D = 1) = E(Δ|X,D = 1), where D is a dummy variable indicating status as HIV-positive (1) or part of the general population (0). This magnitude measures the difference in the average outcome (being employed) of those HIV-positive individuals and those individuals from the general population.

Propensity Score Matching method may have been difficult to accomplish if determined by multiple variables (meaning matching all participants in the treatment group with all participants in the control group with the same characteristics). To avoid this problem caused by dimensionality, it was necessary to use the propensity score for determination purposes, demonstrating that if (Y0, Y1) ⊥ D|X and 0