Designing and implementing a socioeconomic intervention to enhance TB control: operational evidence from the CRESIPT project in Peru

To provide evidence to assist national TB programs considering implementing TB-related socioeconomic interventions, our intervention exclusively targeted TB-affected households (i.e. was “TB-specific”) rather than targeting all households living below the poverty line. The reasons for this decision were: encouraging results from the TB-specific ISIAT project; [39] the urgent need for impact assessment and operational evidence for TB-specific socioeconomic interventions; the lack of existing TB-specific or TB-sensitive socioeconomic initiatives with which to feasibly collaborate in Peru; and the achievability of focusing on relatively small numbers of TB patients versus much larger, operationally unmanageable numbers of people at risk of TB in the wider community [52]. In addition, it was expected that by working exclusively with TB-affected families we would generate evidence concerning those sections of the community most at risk of TB.

Cash transfers directly into bank accounts were selected as the most locally-appropriate way to deliver economic support because in the impoverished shantytown communities of the study site there were many: local bank agencies; food or material vouchers had poor accessibility and acceptability; direct cash transfers posed a security risk; and transfers using mobile-phone technology potentially overlooked the most vulnerable patients [53] and were prone to handset loss/theft, damage, or faults.

Deciding on the size and duration of cash transfers was difficult because this has varied considerably in past projects [47, 49]. Learning from similar regional projects [45, 49], our local catastrophic costs findings [7], and ongoing liaison and site visits from key policy-makers from WHO, Pan-American Health Organisation and the World Bank, we aimed to completely mitigate TB-related direct out-of-pocket expenses, which was expected to be equivalent to 10 % of median TB-affected household annual income in the study site [7]. This amount was: empirically believed to be too small to act as a perverse incentive; [34, 35] affordable for a TB program in a low income country (expert opinion suggests that a socioeconomic intervention that adds less than 50 % to the cost of biomedical treatment but reduces TB risk by 30–40 % would be likely to justify policy change and widespread implementation); [54, 55] large enough so that poverty-related TB risk factors in TB-affected households may be reduced; and that incentivized and enabled TB-affected households to achieve the shared goals of the Peruvian National TB Program and CRESIPT project.

We designed the intervention so that cash transfers would be provided throughout treatment but weighted towards the first 2 months, when TB-affected households incur the majority of hidden costs (Additional file 1: Figure S1a and b) [7, 56, 57].

We stratified cash transfer incentives into “double” and “simple” incentives. Double incentives were made for meeting the condition “optimally” (i.e. monthly adherence missing less than two daily tablets). Simple incentives were made for meeting a condition “acceptably” (i.e. monthly adherence in which two or more tablets had been missed but the patient had not abandoned treatment). Figure 3 summarizes seven different potential scenarios of TB patients and the total amount of cash transfer incentives they would receive. Were a participant with non-MDR TB to receive all the double incentives available throughout treatment, they would receive a total of 230 US Dollars; for all simple incentives, they would receive a total of 115 US dollars (Additional file 1: Figure S1a and b). In situations in which TB treatment routinely extended beyond 6 months, such as HIV-TB co-infection (9 months) or multi-drug resistant (MDR) TB (18 to 24 months), cash transfers continued throughout the duration of treatment. The decision to stratify simple and double incentives was taken in order to encourage a positive feedback loop of behaviour change through graded incentives whilst increasing the opportunity for vulnerable patient groups to receive cash transfers even when they could not achieve conditions optimally.

Of 10 banks visited, formal meetings were organised with four that aimed to: create a relationship with the bank to achieve sustainable cash transfers throughout the study; identify charge-free appropriate accounts; create a “virtual” way of opening accounts to minimize paperwork, time spent “in branch” and travel-related patient costs; establish a mutually suitable day on which to accompany patients to open accounts; and to clarify the bank’s accessibility in our study sites (i.e. branches and agencies).

The banks we consulted raised similar concerns about the proposed intervention, including: infection risk; cash transfer flow; and difficulties opening accounts with patients who have no national identification, fixed abode, or are illiterate. We initially chose one bank that appeared to be more likely to overcome these issues because it had a social inclusion department with previous involvement in successful microfinance initiatives.

Recruited patients with a negative sputum smear microscopy test (indicating low infectiousness) were accompanied by our project staff to open a bank account. The account holder’s details were then relayed to our project office with a copy of the bank’s original documents. In the case that the patient was a minor, did not have legal capacity, wished for another household member to be the named bank account holder, or had prolonged sputum smear positivity, then a household member was selected by the patient or household to be the named bank account holder. Patient transport and time costs were reimbursed by our project.

The patient’s incentive card (Additional file 1: Figure S1a and b) was updated by the field nurses when each condition was achieved. Confirmation of completion was made through liaison with the patient, review of CRESIPT project records (e.g. participatory community meeting attendance) and Peruvian National TB Program records and treatment cards (e.g. medication adherence verification). Signed incentives cards were returned to a project administrator who double digitized the data. Thus, a weekly list of patients, their bank account details and required transfers was generated. The same day, this list was submitted electronically to a member of the social inclusion department of the bank, and the virtual transfers made. The transaction codes and receipts generated were double digitized in the CRESIPT project database and delivered to the patients in the health post by the CRESIPT field nurses.

Of 1299 potential cash transfers, 964 (74 %) were achieved (of these, 885/964 [92 %] were achieved optimally and 79/964 [8 %] sub-optimally), 259 (19 %) were not achieved, and 76 (7 %) were yet to be achieved. Thus, 964 conditional cash transfers totalling 61,120 Peruvian Soles (20,373 US dollars) were made to TB-affected households for meeting the conditional goals of the Peruvian National TB Program and CRESIPT project (Fig. 4a and b). The average cash transfer amount received by each TB-affected household over the course of the intervention was $173 USD.

The significantly higher number of patients declining to participate seen in the urban rather than peri-urban communities may have been due to the fact that CRESIPT project activities were new in this area or reflect distinctions between these communities:

“We don’t fully understand the demographic differences and challenges between the urban and peri-urban communities” [CRESIPT Project Investigator]

The field staff reported that some patients were not willing to participate because: i) they thought that CRESIPT project staff were part of a governmental body; ii) they did not want to register their true address with either the TB program, a bank, or the CRESIPT project in order to keep “under the radar” [CRESIPT Project Nurse] especially those formerly-incarcerated or involved with “pandillas” (drugs gangs); iii) they did not wish our team to visit their home because their household was unaware of their diagnosis or they frequently moved location; or iv) the incentives were insufficient to match the money they lost for participating in project activities and, more importantly, continuing on their treatment. In addition, patients with recognised “high-risk” factors for treatment default such as prolonged treatment courses (e.g. MDR TB and/or HIV), mental illness, illegal drug use, homelessness, or being formerly-incarcerated were difficult to recruit and retain. Because these patients did not consent to participate we could not formally characterise their reasons for declining. This lack of engagement and formal feedback is concerning given that such groups may have benefited most from the intervention.

“Conditions are appropriate but you need to provide additional support to those people who find it harder to meet those conditions” [Ex-TB Patient Civil Society Representative]

“Some patients would never open a bank account because they don’t want to register their name” [CRESIPT Project Nurse]

“[A negative aspect of the CRESIPT project intervention is] giving an economic incentive to a patient with drug-dependency and for that matter any other benefit/incentive such as food baskets (which some of these patients sell to buy drugs)” [Peruvian National TB Program Nurse]

To combat some of these issues, extra care was taken during the informed consent process to reassure potential participants that the CRESIPT project team is a non-governmental research organisation with no connections to the justice system and that no project activity, especially household visits, was mandatory. In an attempt to address the needs of participants with HIV and/or MDR TB, we explicitly specified that cash transfers for adherence were provided throughout treatment, regardless of treatment duration. This longevity meant that TB-affected household support, staff-household relationship and financial benefits of the cash transfers were refined to be more equitable and responsive to the ongoing needs of patients with HIV and MDR TB.

Hard-to-reach populations and/or difficult urban settings such as those in which our intervention was implemented, may be characterised by violence, illegal drug use and severe marginalisation that are also associated with TB. These populations and those with comorbidities (such as diabetes, HIV, or mental illness) and/or MDR TB may require differential levels of intervention including prolonged or enhanced conditional cash transfers and social support. Future studies may investigate the barriers, feasibility and impact of delivering TB-specific socioeconomic interventions to challenging, vulnerable groups in such settings.

We changed bank-provider because the initial bank-provider: had limited geographical accessibility; provided inconsistent information “in branch” (resulting in some patients opening accounts with maintenance fees); was reported during feedback sessions to have been stigmatizing towards patients, not due to TB (the branch staff were unaware of patients’ diagnoses) but possibly due to other sociocultural factors such as poverty or appearance; and introduced account maintenance charges to previously charge-free accounts.

“Some patients lost faith in the project due to the hidden bank charges” [CRESIPT Project Nurse].

The new bank-provider, while not having a specific social inclusion department, provided improved coverage and accessibility because of a greater density of agency micro-branches in local shops that facilitated participant transactions. While the new bank-provider overcame the challenges described above, these experiences have led us to question whether banks are the most appropriate delivery strategy. Indeed, conditional cash transfer programs in Sub-Saharan Africa have predominantly used specified pay points to pay participants in cash rather than banks which may be less accessible to the poor and may have user fees [60]. However, banks have been the favoured partner agent in existing conditional cash transfer programs in Peru (JUNTOS), Brasil (Bolsa Familia) and Mexico (Progresa) with the co-ordination of national cash transfer programs being centralised through national banks in these countries [60]. This level of coordination may only be suitable in countries with comparatively developed financial infrastructure, information and communications technologies, and accessibility to bank branches or micro-agencies. We have reviewed other modalities of conditional cash transfers in greater detail elsewhere [20]. Future research into implementation of socioeconomic interventions may compare the effectiveness and cost-effectiveness of cash delivery mechanisms including mobile phone vouchers, mobile banking, automated or other pay points, or innovative strategies, for which rigorous evidence is currently lacking. To achieve optimal impact, implementers of conditional cash transfer programs may work more closely with the local communities and civil societies to establish how a program can be adapted to be appropriate and acceptable in that specific setting.

During focus group discussions, the internal research committee debated what the most important objective of cash transfers is: mitigating TB-related costs; avoiding catastrophic costs; or reducing poverty-related TB risk factors. This confounded deciding the cash transfer amounts. To address this challenge, we analysed TB patients’ costs, which demonstrated that direct out-of-pocket expenditure was 10 % of that household’s annual income [7]. These results, together with additional data characterising annual household income for TB-affected households in the study site, informed the cash transfer amount necessary to match direct out-of-pocket expenditure and subsequently avoid catastrophic costs. The optimal cash transfer size is likely to depend on the intervention setting and proposed outcomes of the intervention. This will require baseline evaluation of local TB-related costs prior to planning for and implementing cash transfers of suitable amounts. Further research is required to evaluate how cash transfer size affects intervention impact and cost-effectiveness [61]. It is noteworthy that the strategy we adopted in this research was a “costs-mitigation” rather than a “poverty-reduction” strategy and involved a cash transfer amount that were appropriate and feasible for the local setting. While long-term poverty reduction would be an appealing additional goal, this would likely require greater socioeconomic support than national TB programs could afford.

During focus group discussions, the delay between incentivized behaviour and cash transfers was noted as a barrier to achieving project conditions due to the lack of a tangible “reward” and accompanying positive reinforcement loop. For example, when a household attended a participatory community meeting, it could be 1 to 2 weeks before they received the corresponding cash transfer.

“Patients want immediate and tangible gratification on the same day as they complete their condition” [Ex-TB Patient Civil Society Representative].

Cash transfers were initially delayed due to the flow of information from the field, to the research office, to the bank. While cash transfer flow improved during the project, patients and households stated that such delays made household budgeting difficult:

“Due to the initial cash transfer delays, some patients didn’t get the money when they most needed it” [CRESIPT Project Nurse].

Participants reported that they would prefer to receive cash transfers at the beginning of the month for their adherence in the subsequent month rather than wait until the end of the month. As has been reported in other settings [45], the waiting was perceived as financially stressful and, on occasion, demoralising. Learning from this setting-specific qualitative feedback, in the planned CRESIPT study in these same communities, we will combine unconditional monthly cash transfer provided to all TB patients taking treatment with supplementary conditional cash transfers for meeting CRESIPT project and National TB Program conditions. Furthermore, we have self-imposed penalties on our project if incentives do not reach the patient’s bank account within 1 week of confirmation that the condition has been met. Specifically, if a delay occurs, their cash transfer is doubled.

Project conditions requiring involvement of “100 %” of the TB-affected household in order to receive the cash transfer (e.g. attendance at participatory community meetings) were hard to achieve. In addition, the amount of these cash transfers was fixed and independent of household size (see Fig. 3) and thus felt to be inequitable because larger households received a lower cash transfer amount per household member. There was, therefore, a perceived challenge in balancing operational simplicity (e.g. fixed-amount incentives) while responding to patient household needs (e.g. variable incentive depending on household size). Consequently, on the basis of this qualitative evaluation of the implementation process, we suggest designing relevant incentives to be more equitable and responsive to household size: a fixed amount added to the patient’s cash transfer for each member of their household on completion of the condition. In this way, larger households will receive the same amount per household member as smaller households.