Early HIV treatment and labour outcomes: A case study of mining workers in South Africa

Abstract

This study examines whether labour outcomes of HIV‐infected workers treated with antiretrovirals are associated with the stage of the disease when commencing therapy. We use data on employment separation and absenteeism from the workplace health programme of South Africa's largest coal mining company over the period of January 2009 to March 2017 in a Cox proportional hazards model. When treatment was initiated at a CD4+ T cell count above 350 cells/μl, the risk of separating from the company was 37% lower and the risk of absence was 20%t lower than initiating at a CD4 count below 200 cells/μl, and these differences persist over time. Also, we find that workers initiating antiretroviral therapy at CD4 ≥ 350 have an 8% lower risk of absence prior to treatment. Although many companies and the South African government have adopted universal test‐and‐treat policies aiming to initiate all HIV‐infected people as early as possible, most HIV patients still start treatment late in the disease course when their CD4 counts have fallen to low levels. Our results indicate early HIV detection and treatment could have large productivity gains.

HIV treatment and worker absenteeism: Quasi-experimental evidence from a large-scale health program in South Africa

“…Given that its inclusion could result in a bad control or collider problem, we do not control for it in our baseline regression. Theoretically, the estimates of treatment would be biased downwards if those treated are sicker due to unobserved comorbidities resulting in higher health-related absenteeism (French et al, 2019). However, as can be seen in Appendix B, Table B2, Panel B, the effect of our treatment variable is not affected by the inclusion of tuberculosis as a control.…”

“…For instance, individuals with a higher unobserved motivation to work may have lower absenteeism but also are more likely to receive treatment earlier. Given that an earlier initiation of treatment is more effective (French et al, 2019), the effect of treatment would be biased upwards. This is captured by the individual-fixed effects 𝛿𝛿 𝑖𝑖 .…”

“…Non-randomized studies, however, are likely to suffer from endogenous treatment selection bias, because unobserved factors affecting the patient's and doctor's decision to initiate treatment can influence the outcome variable. A number of previous studies have estimated ART effects on labor outcomes (French et al, 2019, Habyarimana et al, 2010, Larson et al, 2013. Apart from the inclusion of fixed effects, which can only capture those unobserved confounding factors that are either constant over time or constant across all individuals, these studies have not fully addressed endogenous treatment selection.…”

“…Third, by only using individual-fixed effects, endogenous treatment selection is not fully controlled for in their analysis. French et al (2019) use the same firm dataset that we have used in this work. Their outcome variables are worker turnover and health-related absenteeism.…”

Economic consequences of better health: insights from clinical data

“…Quasi-experimental studies over the past 15 years, mostly in sub-Saharan Africa, have established that HIV treatment increases employment43444546 and worker productivity46 and decreases absenteeism 45. These effects are greater if treatment is started soon after HIV infection occurs rather than after the disease has progressed 4748…”