This qualitative process evaluation provides contextual and process information which goes beyond the trial design and methodology to explain the trial findings . Importantly, it sheds light on some of the reasons for the poor fidelity to the intervention protocol of the trial, and in so doing, explains to some extent why the effects of the intervention were small. It also provides insights into the perceived value of economic support in improving TB treatment outcomes, as well as into participants’ broader views on the concept of economic assistance for the improvement of health outcomes. These findings may help to inform further research as well as the large scale implementation of similar interventions, in other contexts . There were two important factors that detracted from fidelity to the intervention protocol: nurses’ perceptions of the inequity of the criteria for receiving vouchers; and logistical issues involved in voucher administration. Nurses modified the delivery of these vouchers in several ways but, perhaps most importantly for this trial, they rejected the eligibility criteria as being unjust. Working in a context of widespread and deep poverty, it seemed unfair to them to give vouchers only to those who met the eligibility criteria when some of those who met the criteria needed the voucher less and some who didn’t meet the criteria needed it more. Although the eligibility criteria for receipt of a voucher (which did not include socioeconomic status) were clearly outlined to nurses at the start of the trial, the imperative to ration the vouchers was stronger. Nurses frequently ration the food parcels that may be provided to needy patients as part of their routine TB treatment because there is usually insufficient food to give to every patient, and they seemed to treat the vouchers in the same way. This rationing impacted on the delivery of the vouchers and reaffirms the importance of the ‘street level bureaucrat’ as an implementer . Lipsky  suggests that ‘street level bureaucrats’ (public service workers who have important roles in delivering government services, have constant interaction with members of the public, and are able to use their own discretion in carrying out their activities) are not passive media through which the policy or intervention passes from the designers to the recipients. On the contrary, implementers can be seen as active in interpreting and, if considered necessary, modifying policies or interventions, to the extent that they can may constitute a level of policy-making themselves . Logistical problems also impacted on fidelity to the delivery of the intervention. In some clinics it was administratively easier for nurses to give out all the vouchers at the end of the month, but this required patients to come back to the clinic at this time to receive them. Because the study team was small, there were also logistical difficulties with delivering vouchers to all clinics on time. Voucher books were personally delivered to all clinics and collected from all shops by the principal investigator and one assistant, with staff at clinics and shops required to sign proof of delivery or collection. This meant that vouchers were sometimes not available for patients on their appointment dates, which necessitated another visit to the clinic to collect them. Although the frequency of these occurrences were not quantified and seemed a less important barrier to the implementation of the vouchers than the nurses’ rationing, they were raised by some interviewees in this process evaluation. Such logistical issues should be considered in the replication of similar interventions in other settings. The coordination of delivery of vouchers to clinics and collection of vouchers from shops requires considerable organization and a dedicated staff complement. Setting up the infrastructure to manage the voucher system may be difficult where health systems are weak and resources very limited.
Related to issues of logistics are mechanisms for control of the distribution and use of the voucher. There were very few cases of leakage of the vouchers to those for whom they were not intended. In only one case was the theft of a voucher reported and even in this case the theft was reported before the voucher could be redeemed. More common was the fear that lay Directly Observed Treatment Supporters (DOT supporters)  who might collect vouchers on behalf of patients who were too ill to collect them themselves, would either not give the vouchers to the patients or would buy food for themselves. Even this, however, was expressed by only a few participants. The level of leakage of vouchers in this study was thus very small compared to the levels experienced at provincial and national levels with other social grants. For example, in a different province of South Africa, 3000 cases of social grant fraud were recently handed over to the Special Investigating Unit for prosecution. Although it is possible that cases of leakage would increase if vouchers were delivered on a larger scale, the system used in this study is promising in terms of minimizing that potential.
Like patients in Mexico’s conditional cash transfer program , we found that many patients shared their voucher purchases with their families, and that this was consistent with their social values. Although this may have diluted the effect of the vouchers for the index patients, for many it was inconceivable that they should keep the voucher to themselves. Like other social grants in South Africa, the material benefits of the grant are generally distributed throughout the household, so that the household is the unit that benefits rather than the individual recipient . Therefore, for this voucher (as for other social grants) it is important to note that the impact of policies targeted at individuals will be mediated by household dynamics . If the value of the voucher had been larger the dispersion of this benefit may have improved the nutritional status of other household members and so reduced their risk of contracting or developing TB [27–29]. However, because it was relatively small these sharing practices may have meant that neither the index patients nor their households could benefit maximally from the vouchers.
Nurses were of the view that patients who received the voucher came back to the clinics to collect their tablets regularly, and that even known defaulters returned to the clinic to resume treatment. One of the reasons for the perceived improvement in adherence is that most patients who were interviewed valued the food purchases that they could make with their vouchers, and that this food helped to avoid having to take tablets on an empty stomach. This is consistent with the ‘enabling’ effect of economic support to poor patients who are ill; that is, that such support enables adherence by minimizing some of the barriers to adherence imposed by poverty.
This qualitative process evaluation showed that patients generally felt that the value of the voucher (R120.00) was small, and indeed it was, compared to the Child Support Grant (valued at R250.00 per month at the time of the trial), the Old Age Pension, and the Disability Grant (both valued at R1080.00 per month at the time of the trial). However, it was about a fifth of the value of the median per capita income in KwaZulu-Natal around the time of the trial , and most participants said it was helpful, particularly in enabling them to buy the food to take their tablets with. This was a powerful theme in the patients’ interviews, with most saying that it was impossible to take tablets on an empty stomach. In this sense, the voucher enabled patients to take tablets where this may have been difficult without it.
Acknowledging the link between poverty and TB, the TB managers and nurses interviewed in this study agreed in theory with the principle of social assistance for people who are poor and ill. However, managers raised concerns about the impact of a financial transfer to patients with TB in terms of the development of dependency on the grant which is a widespread concern in the country . However, this fear was not realized in our study, which suggested that the more often patients received vouchers the more likely they were to achieve treatment success . It is interesting to note, however, that concerns around dependency on grants and their perverse incentive effects go back a long way in South Africa’s history, and were important debates at the time of the Lund Commission in 1995  and the Carnegie Commission in 1932 . These concerns persist today and relate largely to the concept of the ‘deserving poor’, a concept first articulated in the Poor Laws of Elizabethan England and they remain an area of debate for welfare states today . In essence, the poor who deserve assistance from the state are felt to be those who are unable to work, such as the very young, the elderly, and the disabled. However, in current day South Africa the opportunities for formal employment are diminishing and those which are available are increasingly for people with a completed secondary or tertiary education . Social grants are an important, perhaps even a crucial, means of survival for the poor in this country . Further research is therefore needed to explore whether these perverse effects are indeed found in practice in order to inform ongoing debates.