Early effects of antiretroviral therapy on work performance: preliminary results from a cohort study of Kenyan agricultural workers

Objective: This paper estimates the impact of antiretroviral therapy (ART) on days harvesting tea per month for tea-estate workers in Kenya. Such information is needed to assess the potential economic benefits of providing treatment to working adults.

Methods: Data for this analysis come from company payroll records for 59 HIV-infected workers and a comparison group of all workers assigned to the same work teams (reference group, n = 1992) for a period covering 2 years before and 1 year after initiating ART. Mean difference tests were used to obtain overall trends in days harvesting tea by month. A difference in difference approach was used to estimate the impact of HIV/AIDS on days working in the pre-ART period. Information on likely trends in the absence of the therapy was used to estimate the positive impacts on days harvesting tea over the initial 12 months on ART.

Results: No significant difference existed in days plucking tea each month until the ninth month before initiating ART, when workers worked -2.79 fewer days than references (15% less). This difference grew to 5.09 fewer days (27% less) in the final month before initiating ART. After 12 months on ART, we conservatively estimate that workers worked at least twice as many days in the month than they would have in the absence of ART.

Conclusions: Treatment had a large, positive impact on the ability of workers to undertake their primary work activity, harvesting tea, in the first year on ART.

Introduction

In 2004, Fox et al. documented the decline in work performance associated with untreated HIV/AIDS in a treatment-naive population of tea pluckers in western Kenya ^[1]. Since then an HIV/AIDS treatment program has been introduced in the region. We report here an analysis of the impact of antiretroviral therapy (ART) on one indicator of work productivity for tea pluckers, days plucking tea each month, over the first 12 months of ART.

Methods

The study site is a large tea plantation in Kericho District of the southern Rift Valley Province of Kenya. The plantation's workforce includes more than 10 000 employees on permanent contract. Tea is harvested in every month of the year, with some seasonal and annual variation.

In 2003, the population HIV prevalence in the study area was estimated at 14.3% (11.3% among men, 19.1% among women) ^[2]. The tea company maintains a central hospital which started to provide free ART to eligible employees and dependents in April 2004 with support from a donor-funded program. Eligibility was based on standard criteria for initiating ART in resource-limited settings. The company has a non-discrimination policy towards persons with HIV/AIDS and provides paid leave for medical appointments.

The present study was restricted to permanent employees working primarily as tea pluckers. Tea pluckers work in 'gangs' of pluckers who work in the same fields at the same time.

The index workers (the treatment group) were 59 Kenyan adults who were permanently employed as tea pluckers, provided consent, were hired prior to 2002 (i.e. had a complete work history), and started ART between April 2004 and December 2005.

Study enrollment began in March 2006 and occurred as workers visited the ART clinic for regular appointments. Potentially eligible pluckers were referred for enrollment by clinic staff. Based on the number of workers who attended the clinic and met our inclusion criteria, we estimate 150 workers were eligible for recruitment. Of these, 64 discussed participation with the study nurse and 59 enrolled in the study. After consenting, no workers died during the time period included in this analysis.

The reference individuals (the comparison group) were all 1992 permanently employed workers in the same gangs as the index workers who were hired prior to January 2002. The HIV status of the reference participants was unknown, but was estimated to be 14.3% ^[2]. The presence of HIV-infected individuals within the reference population may bias test results towards the null and underestimate differences between index the treatment group and HIV-negative workers.

Data was extracted on the number of days spent plucking tea per month by index and reference individuals from company payroll databases. The ART start dates for index workers was extracted from company hospital records.

Based on the date when each index worker began ART, we defined the 'duration on ART' for each gang (the index worker and all reference individuals in the index patient's gang) as the difference in months between the calendar date month and the month when the index worker in that gang initiated ART. For example, if the index worker in gang 10 began ART in September of 2004, the duration on ART variable for the index and all reference participants in gang 10 equals 0 for September 2004, -8 for January of 2004, and 3 for December 2004. For gangs including multiple index patients (11 gangs), we randomly split gang members into sub-gangs and randomly assigned a sub-gang to each index worker.

As the index workers initiated ART on various days during the month, month 0 was interpreted as being on ART for less than 1 month. For each index worker, this analysis includes 24 full months prior to the initiation of therapy (months -24 to -1), the month therapy begins (month 0), and 12 full months after therapy initiation (month 1 to 12).

To evaluate the impacts of HIV/AIDS and ART on work performance, three types of comparisons were made. First, we estimated the mean number of days that index and reference participants spent plucking tea by month over the study period. Differences in means by month were tested using a t-test with unequal variances. Second, we calculated the difference in days plucking per month from between month -12 and -1 (i.e. the mean change from month -12 to -1) for index and reference individuals separately. We then estimated the impact of HIV/AIDS over the final year before initiating therapy by calculating the difference between index and reference participants in terms of their change in mean days plucking from month -12 to -1 (a difference in difference estimate). This was done using linear regression with robust standard errors adjusted for within-gang correlation. Third, a similar difference in differences between month 1 and 12 was estimated using assumptions about likely trends for days plucking had index workers not initiated ART.

Ethical review was provided by the Boston University Medical Campus, the Kenya Medical Research Institute, and the Walter Reed Army Institute for Research.

Results

Index workers were slightly older (41.5 versus 38.6 years) and somewhat more experienced (10.8 versus 9.9 years) than reference patients. Index workers were more likely than reference participants to be female (55.9 versus 34.4%, respectively), reflecting the higher HIV prevalence among women in the region. The median CD4 cell count for index workers at ART initiation was 145 [interquartile range (IQR) 86-231].

For the 37 months of this study, Table 1 and Fig. 1 show the mean number of days plucking tea per month at quarterly intervals for index and reference individuals. In each month between -24 and -12, there was no significant difference between index and reference participants. Index workers then worked significantly fewer days (P < 0.05) plucking tea than reference participants for the remainder of the study period (months -9 to 12).

Table 1

Fig. 1

The large decline in days plucking for index workers from the month before beginning ART (month -1) to month 0 reflects two phenomena: a decline in days worked due to illness and/or treatment-related side effects, and a company policy to assign extra sick leave to patients beginning ART.

Index workers worked 3.22 fewer days in month -1 in comparison with month -12, whereas reference workers worked 0.79 more days in month -1 in comparison with month -12. The difference between index and reference participants in changes in days plucking from months -12 to -1 is the impact of HIV/AIDS over this time period. This difference was 4.01 fewer days for index workers (P = 0.004), a 22% decline in days plucking per month in the final year before beginning ART.

In month 12, index workers worked 7.51 more days than they did in month 1, while reference workers worked 1.03 fewer days. The difference in difference estimate is 8.55 more days for index patients than reference participants over the period (P < 0.001).

Figure 1 and Table 1 show that index workers plucked tea an average of 14.38 days per month by the twelfth month on ART. To identify the impact of ART after 1 year on therapy, however, we need an estimate of the number of days that index workers would have worked had they never initiated therapy. Data about this counterfactual outcome are not available, but we consider three plausible hypothetical possibilities. First, index workers worked 13.68 days in month -1, and experienced a mean decline of 3.22 days between months -12 and -1. If this downward trend in days plucking had persisted over the following 12 months, index workers would have worked 10.46 days in their twelfth month on ART, when their actual total was 14.38 days. ART thus allowed them to work 3.92 more days in month 12 than would otherwise have been possible.

We know, however, that this simple linear projection is too optimistic. Studies from resource-limited settings report median survival times of 6-19 months following an AIDS diagnosis ^[3], with a median of 11 months. Thus, rather than a linear decline over time, it is more likely that a negative concave trend would have been observed because of deaths among index workers. If, in the absence of treatment, 50% of index workers had died by 12 months, with the rest experiencing the slow decline discussed in the previous paragraph, an average of approximately 5 days plucking tea in month 12 is perhaps a more reasonable estimate of what would have happened in the absence of ART. In this case, 9.48 additional days plucking in month 12 is the estimated impact of ART in month 12.

Finally, index workers worked 7.51 more days in month 12 than in month 1 and reference workers worked 1.03 fewer days in month 12 than in month 1. The difference in difference estimate of 8.55 additional days plucking tea between months 1 and 12 for index workers provides a conservative alternative estimate of the impact of ART after 12 months on therapy.

Discussion

The results presented here show a clear, positive trend in days spent plucking tea over the initial year on ART for index workers. Based on the discussion above, we conclude that index workers worked an additional 7.5 to 9.5 days (109 to 207% more) in month 12 beyond what they would have worked in the absence of ART.

As with any analysis of intervention impacts in a nonrandomized control setting, selection biases in either study group may be important limitations to this analysis. In this study index workers could have differed from the reference participants even prior to HIV infection, in which case differences between index and reference workers could merely reflect this pre-existing difference. Table 1 and Fig. 1 indicate, however, that index workers plucked tea for the same number of days per month as the general population during their second year before initiating ART. Our sample of 59 index workers could also be systematically different from the eligible population in ways that could affect their work performance after initiating ART. Although we have no data supporting this possibility, we cannot rule it out. A final limitation of this analysis, which would tend to bias the results towards the null, is that the HIV status of the reference participants is unknown. Although some reference workers are HIV infected, they were alive and working at the end of this study period, and only some percent would have progressed to the point of experiencing significant impairments. Nonetheless, the significant impacts identified above (7.5-9.5 additional days plucking in month 12) can be considered conservative estimates.

Due to these limitations, the results reported here are best interpreted as preliminary evidence on the ability of tea pluckers to return to work harvesting tea after initiating therapy. More definite conclusions about the longer run impacts of ART on work performance will require both a larger sample of index workers and a longer period of follow up. The cohort we are following will ultimately include more index workers and a follow-up period of two full years on ART.

Acknowledgements

We are deeply grateful to the workers who consented to be part of this study. We thank the administration, estate managers, and medical staff of the participating tea company for access to data and their generous assistance in accessing and interpreting records. We also thank, without implicating, the journal reviewers and editors for their useful advice on this manuscript. This paper is published with permission from the Director of the Kenya Medical Research Institute.

Sponsorship: Funding for the research presented in this paper was provided by the Fogarty International Center of the National Institutes of Health through the International Studies in Health and Economic Development (ISHED) Program (Grant Number 5R01TW7181-3).

Disclaimer

The views expressed here are the opinions of the authors and are not to be considered as official or reflecting the views of the Walter Reed Army Institute of Research, the US Army, the US Department of Defense, and the Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc.

Authors' contributions

Bruce A. Larson contributed to the conception and design of this study, data management and analysis, interpretation of results, and drafting/completion of the paper. Matthew P. Fox contributed to the conception and design of this study, data management and analysis, interpretation of results, and drafting/completion of the paper. Sydney Rosen contributed to the conception and design of this study, interpretation of results, and drafting/completion of the paper. Margaret Bii contributed to the design of this study, data development and management, interpretation of results, and review of the paper. Carolyne Sigei contributed to the design of this study, data development and management, and review of the paper. Douglas Shaffer contributed to the design of this study, interpretation of results, and drafting/completion of the paper. Fredrick Sawe contributed to the design of this study, interpretation of results, and review of the paper. Monique Wasunna contributed to the conception and design of this study, review of results, and review of the paper. Jonathon L. Simon provided overall guidance into the conception and design of this study and contributed to the analysis, interpretation, and writing of the paper.

There are no conflicts of interest.