The first reported HIV case in India was documented in the city of Chennai in 1986. A little more than two decades later, approximately 2.4 million individuals were living with HIV/AIDS in the country . Recent years have seen rapid progress in the provision of antiretroviral therapy (ART) for HIV-infected individuals with AIDS. In the 6-year period beginning 2004, India's National AIDS Control Organization scaled up the number of ART Centres providing free treatment from eight to 239, and a total of 300 743 people were receiving ART by January 2010 [2–4]. Despite these efforts, ART coverage stood at 23–28% based on the 2009 WHO ART eligibility guidelines and 36–55% based on the 2006 WHO guidelines .
As the Indian government and donors consider how to further respond to the epidemic, evidence on the economic impacts of ART can be useful for modeling purposes and policymaking. As functional capacity deteriorates during advanced stages of HIV infection, there is a high likelihood that labor productivity will also decline substantially . Households in which the HIV-infected individual is a working-age adult therefore risk losing both income and livelihoods. Given its effectiveness in reducing morbidity and mortality among adults with AIDS [7–10], ART has the potential to reverse many of the negative economic consequences of HIV infection. However, evidence from India on the employment and income impact of ART has been limited to date . Recent studies in Africa have documented large increases in labor force participation and labor productivity among adults who receive ART [12,13]. Here we examine the impact of ART on the employment and income of adults receiving care from a government program in the Indian state of Tamil Nadu.
The Tamil Nadu Family Care Continuum Program (TNFCC) is a family-centered program that was initiated in September 2005 by the Tamil Nadu State AIDS Control Society. Implemented at three large medical college hospitals located in three cities (Chennai, Salem, and Tirunelveli), TNFCC provided a comprehensive range of free care to HIV-infected adults and children. This consisted of clinical care (including ART), nutritional supplements and home-based care. The clinical care included routine medical care, diagnosis and treatment of opportunistic infections, and ART for individuals with CD4 cell counts below 200 cells/μl. The nutrition care included nutrition assessments, counseling, and macronutrient and micronutrient supplements. Macronutrient supplements were provided to meet the additional needs of HIV patients as per WHO guidelines (10% more calories for asymptomatic patients and 20% more calories for patients with AIDS). The ingredients of the supplements, which were in powder form and came in 500-g packets, included wheat (225 g), corn flour (100 g), lentils (50 g), finger millet (25 g), and vegetable oil (25 g). Each packet provided 2000 kcal and carbohydrates, protein, fat, and crude fiber. Adult patients on ART received about 8 packets/month and pre-ART adults received about 4 packets/month, but the prescribed dose varied based on the patient's age and lifestyle. The micronutrient supplements came in multivitamin tablets containing essential vitamins and minerals; adult patients were instructed to take two tablets daily.
Outreach workers made home visits and focused on providing ART adherence support, motivating and encouraging all patients to make their monthly hospital visits, and connecting families to social services such as income generating activities, legal services, and housing.
TNFCC received financial support from the Government of Tamil Nadu and Children's Investment Fund Foundation, and technical assistance from the nongovernmental organization Solidarity and Action Against the HIV Infection in India.
Enrollment of patients into the study began in October 2005 and continued through May 2006. In order to measure the family-level impacts of TNFCC, the original study design called for interviews with all adult patients in the program who had children and all pediatric patients and their caretakers. When they enrolled in TNFCC, adult patients and caretakers of pediatric patients were referred to the study's research assistants for a baseline interview. These face-to-face interviews took, on average, 45 min to complete. Due to time constraints, of the 3752 eligible patients enrolled in TNFCC during the enrollment period, a convenience sample of 1664 patients was chosen for the cohort. At each clinic, the study team interviewed as many consecutive patients as they were able to each day, starting with the first patient who arrived at the clinic and ending with the patient whose interview finished at the clinic's closing hour. Of the 1664 patients, 1597 were adults (1543 HIV-positive parents and 54 caretakers) and 67 were children.
The analysis for this study excluded pediatric and adult patients who were not naive to ART at the time they enrolled in TNFCC. Patients missing CD4 cell count information, ART status, and information about key economic characteristics were also excluded from the analysis. As a result, a total of 1238 adult patients were included in the analysis.
Using a structured survey, the cohort was interviewed in the clinic at the time of enrollment in TNFCC (baseline) and every 6 months thereafter for a period of 2 years. In addition to demographic information, data were collected on four economic outcomes: whether patients participated in economic activities during the week prior to interview, the number of hours they worked during the week prior to interview, individual income earned in the past 30 days, and individual income earned in the past 6 months. Information on the BMI, CD4 cell count, and ART initiation date was obtained from the patients' clinical records.
At each of the four 6-month follow-up observations after baseline, our analysis classified patients as belonging to the comparison group if they were not receiving ART or to the treatment group if they were receiving ART. At each observation, we defined four binary variables that indicated whether the patient had been receiving ART for 6, 12, 18, or 24 months. For example, if a patient had not been receiving ART for at least 6 months at the time of an observation, the four binary variables equaled 0 and the patient was part of the comparison group. If a patient initiated ART at baseline, during the 6-month observation the binary variables had the following values: the variable ‘6 months of ART’ equaled 1 and the other three binary variables equaled 0. During the 12-month observation, ‘6 months of ART’ equaled 0, ‘12 months of ART’ equaled 1, and the other two binary variables equaled 0. If a patient initiated ART at 6 months instead, the variable ‘6 months of ART’ equaled 1 during the 12-month observation and equaled 0 during all other observations. The duration of time on ART at each 6-month observation was thus defined by the four binary variables. During periods when a patient was not receiving ART, the patient was part of the comparison group.
Impacts on the outcome variables at 6, 12, 18, and 24 months following ART initiation were estimated by a linear regression model with individual (i.e. patient) fixed effects. The fixed effects adjust for unobserved patient characteristics that were constant through the study period but might have influenced baseline levels. The model also included standard errors that were clustered at the patient level. The way in which the four binary ART indicator variables were defined allowed for measurement of impacts at each 6-month interval relative to baseline.
To identify the impact of ART, the model examined the trends in outcomes of patients from the time they initiated ART and used the comparison group to adjust those trends for factors that were independent of ART provision, such as secular trends in employment and income in the study areas. This adjustment, which was done by including a binary indicator for each follow-up period in the regression model, was essential for attributing outcome changes in treated patients to ART. Statistical analyses were performed using STATA version 10.0 (StataCorp, College Station, Texas, USA).
The study received human subjects approval from Duke University and by local ethical review boards.
Table 1 describes the baseline characteristics of the 1238 patients included in the analysis. Three hundred and sixty patients were initiated on ART at baseline, followed by 44, 52, and 59 patients (155 in total) who initiated ART at 6, 12, and 18 months after baseline, respectively. The remaining 723 patients remained asymptomatic throughout the study period and did not receive ART. Thus, for the purpose of our statistical analysis, all patients contributed person-time to the study, with 360 contributing person-time only as ‘treated’ patients, 155 contributing person-time both as ‘treated’ and ‘comparison’ patients, and the remaining 723 contributing person-time only as ‘comparison’ patients.
We compared characteristics of the 360 patients who initiated ART at baseline (the baseline ART cohort) and the 155 patients who initiated ART during the study (the postbaseline ART cohort) with the 723 patients who never initiated ART during the study (no-ART cohort). Patients initiating ART at baseline were significantly older, less likely to be illiterate, and less likely to be widowed than patients in the two other cohorts. A significantly higher proportion of the baseline ART cohort was male. Marital status and secondary school completion rates were similar across cohorts. The postbaseline ART cohort was more similar to the no-ART cohort than the baseline ART cohort, as age was the only demographic characteristic that differed significantly.
Turning to economic outcomes, 28% of the baseline ART cohort reported having worked in the 7 days prior to the baseline interview, compared with 44% of the no-ART cohort (P < 0.01). The baseline ART cohort also worked significantly fewer hours in the 7 days prior to the baseline interview (11.4 h compared with 17.7 h, P < 0.01). Despite the baseline ART cohort working less, there were no significant differences in income earned in the past 30 days and 6 months (US $1 equaled roughly 45 rupees at the time of the study). On the basis of means reported in Table 1, average hourly wages ranged from US $0.20–$0.30. Table 1 also shows that the postbaseline ART cohort did not have significantly different employment and income than the no-ART cohort.
The difference in baseline health status among the three cohorts of patients is also evident in Table 1. The baseline ART cohort had the lowest CD4 cell count and BMI, followed by the postbaseline ART cohort. The no-ART cohort was significantly healthier than both of the other cohorts at baseline.
Table 2 provides an indication of the health improvement experienced by ART patients over time. The baseline ART cohort was the one that was followed on treatment for the longest duration. At each 6-month observation, the table reports the mean CD4 cell count for this cohort and the no-ART cohort. Over time, there was a large increase in the mean CD4 cell count of the ART cohort, from 128 cells/μl at baseline to 465 cells/μl at 24 months of ART. The no-ART cohort's mean CD4 cell count remained stable between 465 cells/μl at baseline and 494 cells/μl at 24 months.
Of the 1238 patients interviewed at baseline, 850 were interviewed at 24 months. Attrition of patients from the interview cohort occurred for a number of reasons including mortality, loss to follow-up, and patients not having been successfully referred to the study team.
Table 3 reports the results from estimating the regression model with patient fixed effects. In the model, employment trends for patients who initiated ART after baseline contributed to the estimated 6, 12, and 18-month impacts of ART. The 24-month impact of ART, on the other hand, was identified only by the experience of the baseline ART cohort. The results in columns 1–4 indicate that patients receiving ART experienced a statistically significant increase in all the employment and income outcomes. Even after adjusting for secular time trends with data from the comparison group, employment increase due to ART occurred rapidly: within 6 months after ART initiation, there was an increase of 10% in the probability of being economically active (P < 0.01). This increase was over and above employment changes for the entire patient population. In addition, during the first 6 months, reported weekly hours worked by ART patients rose by 5.55 (P < 0.01). Employment and income levels remained significantly higher than baseline as ART duration increased. Compared with baseline, at 24 months after ART initiation there was an increase of 25.1% in the probability of being economically active (P < 0.01) and of 10.89 in weekly hours worked (P < 0.01). Similarly, income earned by patients increased over time, although these increases took longer to become statistically significant. At 18 and 24 months after initiation of ART, monthly income was rupees 170 (P < 0.01) and 331 higher (P < 0.01) than baseline, respectively.
As shown by the coefficients of the binary follow-up indicator variables in columns 1–4 of Table 3, employment outcomes for the comparison group also improved over time. Relative to baseline employment levels, at 6 months, these patients had an increase of 10% in their probability of being economically active (P < 0.01). At 24 months, the probability of being economically active was 12% higher than at baseline (P < 0.01). Significant increases were also found for hours worked and income earned. Figure 1 shows the changes in number of hours worked in the past week (relative to baseline) for patients who initiated ART at baseline and comparison patients who did not receive ART throughout the study period. The number of hours worked increased for both groups, but the increase for the baseline ART cohorts was over and above that for the no-ART cohort.
Because the sex composition of the baseline ART group is different from that of the comparison group, columns 5 and 6 of Table 3 report the employment impacts of ART for men and women separately. Although the probability of being employed increased significantly for both men and women, the impact of ART was almost twice as large for men compared with women. At 24 months of ART, the increase in the probability of being employed was 32.1% for men (P < 0.01), and 14.5% for women (P < 0.01).
Finally, when we exclude the postbaseline ART cohort and limit the analysis to the baseline ART and the no-ART cohort, we found that the employment and income impacts of ART remained large and statistically significant. As shown by columns 7–10 of Table 3, the impacts of ART were larger for the baseline ART cohort, particularly at 6 and 12 months. The income impacts were positive but were not statistically significant until 24 months. Separate analysis of the employment and income impacts for the postbaseline ART cohort was limited by the small cohort size.
Although previous studies have documented decreased HIV-related work absenteeism and increased labor supply and income for adults receiving ART, this study is among the first to implement an evaluation that includes a comparison group of pre-ART HIV-positive patients who receive the same program services as patients who receive ART [12,13]. In the absence of experimental data, using data from pre-ART patients to adjust for confounding factors can result in more accurate estimates of the economic impact of ART. During the 24 months following ART initiation, the results show that even after adjusting for trends in employment and income that are experienced by pre-ART patients, ART patients' labor market participation, number of hours worked, and income earned were all significantly higher than at baseline.
The results are consistent with findings of several studies of ART in Africa. For example, a study in Kenya reported a 6-month increase in labor force participation that was one-third higher than baseline levels . This study showed a similar 6-month increase and an even larger increase after 2 years of ART. The persistence of these employment impacts is striking, as it suggests that ART provision can result in longer-term economic gains.
Another important result is that the employment impact of ART was smaller for women compared with men. Similar findings have been reported in Kenya , and one explanation for this is that the employment outcomes examined in this study do not include domestic work (such as cooking and housekeeping) that tends to be performed disproportionately by women. This may explain why market employment does not increase as much for women due to ART. Factors such as stigma and discrimination may also explain this result and further examination of this issue is essential.
ART resulted in an increase in income earned by adults, especially after 18 and 24 months of treatment. Given the informal labor market in which many patients worked, income varied considerably across patients and across time. This is a common problem associated with analyzing income data in low-income settings; hence, the estimated income impacts of ART were less robust than the employment impacts.
The results show that the comparison group of pre-ART patients also experienced a significant improvement in economic outcomes. These increases could have been due to the nutritional support and home-based care they received. However, they could also have been driven by broader macroeconomic factors in the study areas, and it is not possible to distinguish between the effects related and unrelated to TNFCC services. This analysis did allow for the attribution of a significant part of the increase in economic outcomes of ART patients to ART provision. The possibility that additional macronutrient supplementation provided to ART patients may be driving some of the employment and income increases should also be noted. The combination of nutrition supplements with ART is becoming increasingly common, and although the additional supplements to ART patients were not so large as to be the primary reason for the economic impacts reported here, it could be that nutrition complementarities were essential for achieving those impacts.
Several limitations of this study should be acknowledged. A first set of limitations stem from the study not being based on random assignment of ART among equally eligible adults. Relative to the HIV-infected but asymptomatic adults who are in the comparison group, untreated adults with CD4 cell counts below 200 cells/μl would undoubtedly experience a more rapid decline in health and presumably economic outcomes over the 2-year study period. This implies that our results could be underestimates of the true economic impact of ART. When compared with the counterfactual of not providing ART, the patients receiving ART would experience an even larger economic benefit. A second related bias is that, like the treatment group, the comparison group received nutritional support and home-based care. This again would mean that the comparison group's health and economic outcomes would not deteriorate as much as a true comparison group's, thereby resulting in an underestimate of the impact of ART.
Related to the nonexperimental nature of the study, selection of patients into TNFCC and into comparison or treatment groups is another important issue. These patients might be different from other HIV-infected adults in the study area. Moreover, patients who enroll in TNFCC with advanced HIV disease might also be different from those who enroll when they are asymptomatic. As an example, the proportion of men was significantly higher in the baseline ART cohort compared to the postbaseline ART cohort and the no-ART cohort. This could be due to the higher HIV prevalence in India among men and because men tended to come to the clinic at a later stage of the disease than women when the TNFCC program began. To allay some concerns about the sex differences in the treatment and comparison groups, it is reassuring that the employment impacts of ART remained positive and statistically significant when we conducted the analysis separately for men and women. The effects of other differences between the treatment and comparison groups, such as differences in education and marital status, are harder to predict. Further analysis of larger cohorts from various ART programs will be necessary to ensure that the results in this study are generalizable.
Finally, we acknowledge that because not all patients remained alive for the 24-month study period, it is possible that the longer-term impacts overestimate the average level of improvement in economic activity that can be expected as a result of ART. This study did not collect information on mortality; limited resources as well as the challenges of working in an urban area made it difficult to conduct active surveillance and follow-up with patients who did not return to the clinic. Due to the passive recording of mortality, we are therefore unable to correct for it.
Despite these limitations, the fact that the comparison group in this study comprised HIV-infected adults who self-selected into TNFCC makes the results presented here more convincing than other results comparing ART patients to uninfected adults. A larger sample of patients and more varied economic settings would be necessary before drawing stronger conclusions. However, as early evidence on the long-term economic impacts of ART in India, this study suggests that the negative economic impact of HIV/AIDS may be mitigated through ART provision. Although policy makers have recognized the importance of scaling up access to ART, there remains substantial unmet need in many countries. These results suggest that closing this coverage gap can yield sizable economic returns.
We thank Tamil Nadu State AIDS Control Society for its leadership and for initiating TNFCC; the physicians, nurses, case managers, nutritionists, counselors, lab technicians, and other support staff, who provided clinical care and nutritional services to patients in the Kilpauk Medical College Hospital in Chennai, the Government Mohan Kumaramangalam Medical College Hospital in Salem, and the Government Medical College Hospital in Tirunelveli; Solidarity and Action Against the HIV Infection in India for coordinating with NGOs in the community; hospital-based and community-based NGOs who provided home-based care to patients and their families; the DGHI M&E data managers and assistants; and Nalini Tarakeshwar for feedback.
Authors' contributions: M.M. and H.T. designed the study. Data collection was performed by G.S., V.A., and R.M.S. Data were analyzed by H.T. and interpreted by H.T., A.J., and M.M. The manuscript was written by H.T., A.J., G.S., and M.M. and was edited by S.K.A., M.P., S.R., and S.K. All authors were involved in the decision to submit the manuscript for publication.
The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of The World Bank, its Executive Directors, or the governments they represent.
Funding source: Children's Investment Fund Foundation.