Cash interventions to improve clinical outcomes for pulmonary tuberculosis: systematic review and meta-analysis.

Abstract Objective To assess cash transfer interventions for improving treatment outcomes of active pulmonary tuberculosis in low- and middle-income countries. Methods We searched PubMed®, Embase®, Cochrane Library and ClinicalTrials.gov for studies published until 4 August 2017 that reported on cash transfer interventions during the treatment of active pulmonary tuberculosis in low- and middle-income countries. Our primary outcome was a positive clinical outcome, defined as treatment success, treatment completion or microbiologic cure. Using the purchasing power parity conversion factor, we converted the amount of cash received per patient within each study into international dollars (Int$). We calculated odds ratio (OR) for the primary outcome using a random effects meta-analysis. Findings Eight studies met eligibility criteria for review inclusion. Seven studies assessed a tuberculosis-specific intervention, with average amount of cash ranging from Int$ 193–858. One study assessed a tuberculosis-sensitive intervention, with average amount of Int$ 101. Four studies included non-cash co-interventions. All studies showed better primary outcome for the intervention group than the control group. After excluding three studies with high risk of bias, patients receiving tuberculosis-specific cash transfer were more likely to have a positive clinical outcome than patients in the control groups (OR: 1.77; 95% confidence interval: 1.57–2.01). Conclusion The evidence available suggests that patients in low- and middle-income countries receiving cash during treatment for active pulmonary tuberculosis are more likely to have a positive clinical outcome. These findings support the incorporation of cash transfer interventions into social protection schemes within tuberculosis treatment programmes.


Introduction
Tuberculosis remains one of the top 10 causes of death worldwide, with the highest burden of disease in low-and middle-income countries. 1 In these countries, the disease disproportionately affects the most vulnerable populations. 1,2 In 2015, the World Health Organization's (WHO's) End TB Strategy set the goal of a 90% reduction in tuberculosis deaths, an 80% reduction in tuberculosis incidence rate and zero catastrophic costs for tuberculosis-affected families by 2030. 3 These goals explicitly acknowledge the need to both directly treat people infected with the disease and address social determinants of health to improve tuberculosis outcomes.
Social protection policies protect individuals or households during periods when they are unable to financially support themselves because of a range of conditions, such as illness or disability. 4 Cash transfer interventions, defined as cash payments provided to selected beneficiaries by formal institutions, are one form of social protection that has been proposed in the setting of tuberculosis. 5,6 Such interventions can either be tuberculosis-specific or tuberculosis-sensitive. 6 Tuberculosis-specific interventions target directly tuberculosis patients and their households, and are typically incorporated into existing tuberculosis treatment programmes. 6 A tuberculosis-sensitive intervention is part of a broader social protection scheme, potentially affecting tuberculosis outcomes by targeting communities and groups that are at high risk for tuberculosis. The effect on health outcomes, cost-effectiveness and feasibility of these two strategies are not well established and likely to vary based on the local social protection and health-care infrastructure.
Since a review in 2011 on the effects of cash transfer interventions on tuberculosis outcomes in low-and middleincome countries was inconclusive, 7 we assessed the current state of the evidence for such interventions. We were especially interested if cash transfer to people receiving treatment for active pulmonary tuberculosis affects their clinical outcomes.

Methods
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 8 The review protocol is available from the corresponding author.
To identify studies on the use of cash transfer interventions during the treatment of active pulmonary tuberculosis in low-and middle-income countries, we searched the online databases PubMed®, Embase®, Cochrane Library and ClinicalTrials.gov. We used the search string "Tuberculosis" AND ("financial support" OR "token economy" OR "reimbursement" OR "economic burden" OR "incentives" OR "cash transfer" OR "enablers") to identify studies published Objective To assess cash transfer interventions for improving treatment outcomes of active pulmonary tuberculosis in low-and middleincome countries. Methods We searched PubMed®, Embase®, Cochrane Library and ClinicalTrials.gov for studies published until 4 August 2017 that reported on cash transfer interventions during the treatment of active pulmonary tuberculosis in low-and middle-income countries. Our primary outcome was a positive clinical outcome, defined as treatment success, treatment completion or microbiologic cure. Using the purchasing power parity conversion factor, we converted the amount of cash received per patient within each study into international dollars (Int$). We calculated odds ratio (OR) for the primary outcome using a random effects meta-analysis. Findings Eight studies met eligibility criteria for review inclusion. Seven studies assessed a tuberculosis-specific intervention, with average amount of cash ranging from Int$ 193-858. One study assessed a tuberculosis-sensitive intervention, with average amount of Int$ 101. Four studies included non-cash co-interventions. All studies showed better primary outcome for the intervention group than the control group. After excluding three studies with high risk of bias, patients receiving tuberculosis-specific cash transfer were more likely to have a positive clinical outcome than patients in the control groups (OR: 1.77; 95% confidence interval: 1.57-2.01). Conclusion The evidence available suggests that patients in low-and middle-income countries receiving cash during treatment for active pulmonary tuberculosis are more likely to have a positive clinical outcome. These findings support the incorporation of cash transfer interventions into social protection schemes within tuberculosis treatment programmes.
between the databases' inceptions and 4 August 2017. We also manually reviewed reference lists of identified systematic reviews, relevant articles and abstracts from the Union World Conference on Lung Health 2011-2016.

Eligibility criteria
We considered clinical trials and observational studies published in English, Spanish or French that assessed cash transfer interventions directed at people initiating treatment for microbiologically confirmed or clinically suspected active pulmonary tuberculosis. We used the WHO definition for tuberculosis and the 2017 World Bank's classification of low-and middleincome countries. 9, 10 We included studies that reported standard outcomes of treatment completion, microbiologic cure or treatment success, which includes both treatment completion and cure. 10

Study selection and data collection
After removing duplicate records, two reviewers independently screened titles and abstracts of all records for inclusion in full-text review. After screening, two different reviewers independently applied eligibility criteria to each full-text article. Two reviewers then proceeded to data extraction using a standardized form created for the study (Box 1). Disagreements were settled by consensus among all authors.
To better understand the relative amount of cash distributed in the included studies, we converted the average and maximum possible amount of cash received per patient within each study into international dollars (Int$) using the purchasing power parity conversion factor, and then adjusted for inflation into 2016 Int$ with the local inflation conversion factor. 11 If the average amount of cash received by patients in the intervention group was not reported in the article, we contacted the authors to provide the figures.
Because tuberculosis disproportionately affects the poorest households within a given context, 12 we estimated the average amount of cash received per patient as a proportion of annual individual income by dividing the aver-age amount of cash received per patient by the median income per capita of the lowest quintile of that country from the time period of the study. 11 Householdlevel income data were not available to estimate the interventions as proportion of annual household income.

Assessment of bias
For the randomized study, we assessed risk of bias using the Cochrane Collaboration Risk of Bias Tool, and defined a randomized study as overall high risk of bias if the trial met criteria for high risk of bias in more than one assessed domain. 13 We assessed risk of bias within non-randomized studies using the Newcastle-Ottawa Scale, defining a non-randomized study as overall high risk of bias if it had zero stars in any of the three assessed categories. 14 We generated a funnel plot to evaluate Box 1. Type of data extracted from identified studies on cash interventions to improve tuberculosis outcome We extracted data on location; urban and rural setting; time frame; study design; number of subjects; age and gender of participants; HIV prevalence; number with microbiologically confirmed tuberculosis; number with confirmed or suspected MDR and XDR tuberculosis; type of usual care for tuberculosis; annual individual or household income; whether the intervention was conditional; tuberculosis-specific or sensitive intervention; concurrently implemented cointerventions; primary and secondary outcomes.   publication bias for studies included in the meta-analysis.

Data analysis
All identified studies were included in a qualitative synthesis. After excluding studies at overall high risk of bias, we generated summary effect measures using a random effects model for our primary outcome of interest, the odds ratio (OR) of a positive clinical outcome, defined as either a treatment success; treatment completion, if a study did not report treatment success; or microbiologic cure, if a study did not report treatment success or treatment completion. If a study reported ORs adjusted for potential confounders we included these ratios in our analysis. We assessed heterogeneity by using the Cochran's Q test and the I 2 statistic. Among studies included in metaanalysis, we wanted to investigate sources of heterogeneity, including average amount of cash transfer, presence of non-cash co-intervention, treatment success rate in the control group, urban or rural setting, human immunodeficiency (HIV) prevalence, multidrug resistance (MDR) or extensive-drug resistance (XDR) tuberculosis prevalence and World Bank income classification. However, there was not enough information available to complete a random effects meta-regression model using any of these variables.
We used Comprehensive Meta-Analysis software version 3 (Biostat, Inc., Englewood, United States of America) and Review Manager Version 5.3 (The Cochrane Collaboration, London, United Kingdom of Great Britain and Northern Ireland) for data analysis.

Study selection
We identified 1537 publications and after removal of 639 duplicates, we screened 898 titles and abstracts yielding 100 full-text articles to be assessed for eligibility. Of these full-text articles, 92 were excluded (Fig. 1). We included eight eligible articles: one randomized control trial, 15 two non-randomized intervention studies, 16,17 and five observational studies, [18][19][20][21][22] comprising a total of 21 976 subjects. Table 1 summarizes the settings and populations of the included studies.

Study settings and populations
With the exception of one study that took place in 1989-1990, 17 the studies assessed cash transfer interventions between 2004 and 2015. The settings varied: one study took place in a rural clinic, 17 one in a large rural secondarycare facility, 16 four in urban centres, 15,[19][20][21] and two were nation-wide studies. 18,22 Three of the studies took place in countries currently on the WHO list of highburden countries for tuberculosis 16,19,22,23 and two other studies were in a country currently considered high burden for MDR tuberculosis. 15,21 One study focused on migrant workers, a high-risk group within an urban centre. 19 The remaining studies evaluated all tuberculosis patients identified within a given geographic or clinical service area. The control groups were either patients randomized to the non-intervention group, 15 living in a non-intervention area, 17,19 historical controls from the same population before the implementation of the intervention, 16,18,21 eligible for the intervention, but not yet receiving cash, because of administrative delay, 22 or not eligible for the intervention, because of insufficient financial need. 20 Prevalence of HIV seropositivity among the study populations was 0-15% in the six studies reporting the outcome. 15,16.14,17,18,22 Patients with MDR/ XDR tuberculosis were excluded from four studies, 16,18,19,22 two studies reported low prevalence (1-9%), 15,20 while two did not report on drug susceptibility. 17,21 Five studies reported free care for tuberculosis, 15,17,19,21,22 with the others not specifically commenting on the cost of care. 16,18,20 Participants in six studies received the WHO recommended directly observed therapy, short-course. 24

Tuberculosis-specific interventions
In total, seven studies evaluated tuberculosis-specific cash transfer interventions (Table 2). [15][16][17][18][19][20][21] Six of these studies were at least partially conditional on clinic attendance or treatment completion, 15-20 and one did not report whether the intervention was conditional. 21 Four studies described an additional transportation reimbursement. [17][18][19]21 Four studies did not report the average amount of cash received by patients in the intervention group. We contacted the authors of these studies and authors of two studies provided the amount, 16,18 while this information was not available for other studies. 19 15,16 including the CRESIPT project in Peru, the only identified randomized control trial. 15 The CRESIPT project distributed cash using bank deposit (hypothesizing that opening a bank account was empowering to the study subjects), 15 whereas other studies used actual cash 16,17,19,20 or did not report method of delivery. 18,21 Four studies included some additional non-cash co-intervention, including home visits, community meetings, food vouchers and psychological intervention. 15,17,18,21

Tuberculosis-sensitive interventions
Only one study described a tuberculosissensitive intervention, a nation-wide retrospective cohort study in Brazil of tuberculosis patients in the Bolsa Familia programme. The programme is a monthly cash transfer to poor people that is conditional on attending antenatal care, nutrition and vaccine monitoring for their children and that their young children attend school. 22 People with newly diagnosed non-MDR tuberculosis who received cash during treatment were compared to those who were eligible for cash at the time of treatment, but did not receive it, because of administrative delays. The average amount of total cash delivered to the intervention group was Int$ 101, representing an estimated 3.1% of annual individual income. Cash could be claimed by the patient monthly using a designated bank card.

Outcomes
Most studies (5) reported the primary outcome of treatment success, 15,16,[18][19][20] one reported treatment completion 21 and two reported microbiologic cure (Table 3). 17,22 Four studies controlled for potential confounders. 16,18,19,22 Two of the three studies that reported loss to follow-up found significantly less loss to follow-up in the intervention group. 15,16,18 Of the four studies which reported mortality, none found a difference between the intervention and control groups. [15][16][17][18] Bias   (Fig. 2). Fig. 3 shows the forest plot of the remaining tuberculosis-specific studies after excluding studies at high overall risk of bias. Patients receiving tuberculosisspecific cash transfer were more likely to have a clinical positive outcome than patients in the control groups (OR: 1.77; 95% confidence interval: 1.57-2.01), with I 2 = 0% (Q test P = 0.44).

Discussion
The findings of this systematic review and meta-analysis suggest that cash transfer interventions for patients in low-and middle-income countries initiating tuberculosis treatment may improve clinical outcomes. All studies reported improvement in treatment outcomes. However, the overall evidence is weak because we only identified one eligible randomized control trial. Additionally, half of the studies included some non-cash co-intervention and thus some of the positive effects seen may be related to the pooled effects of cash and non-cash interventions.
There are several possible mechanisms by which cash transfer interventions may improve clinical outcomes for tuberculosis patients during treatment. 6,25 Both tuberculosis-specific and tuberculosis-sensitive cash transfer interventions can act as direct povertyreduction measures by offsetting costs caused by the disease. These costs include both direct costs of treatment such as clinic fees, medication costs, travel and/or food, as well as indirect costs incurred through loss of wages. In particular, catastrophic costs, defined as tuberculosis-related costs which exceed 20% of the household's annual income, have been associated with adverse clinical outcomes. 1,2,26 Two studies have found that, on average, a person with tuberculosis in a low-and middle-income country will experience catastrophic costs as a result of the illness. 2 The definitions of the outcomes were: treatment success was positive clinical outcome; treatment completion was if a study did not report treatment success; and microbiologic cure was if a study did not report treatment success or treatment completion. b Derived from multivariable regression models.
(. . .continued) intervention provided cash equivalent to 173% of estimated annual individual income, 17 four provided cash equivalent to 10-20% of estimated annual individual income, 15,16,18,21 and two of these studies also reported the intervention as percentage of annual household income, between 3-6%. 15,21 The single tuberculosis-sensitive intervention we identified provided cash equivalent to 3.8% of estimated annual individual income. The difference between tuberculosisspecific and tuberculosis-sensitive interventions may reflect the findings that sensitive interventions are less likely to be effective and affordable by countries for offsetting tuberculosis-associated catastrophic costs than specific interventions. 27 However, tuberculosis-sensitive interventions also have the advantage of a broader poverty reduction impact, which might improve household economic resilience before a household member develops active tuberculosis infection.
Beyond simply offsetting costs, cash transfer interventions may also serve as an additional incentive for health-seeking behaviour, particularly when distribution is conditional on clinical followup or medication adherence. [28][29][30] Several systematic reviews have found a positive effect of conditional cash transfers in low-and middle-income countries on health behaviours and outcomes, including increased use of preventative services, improved childhood nutritional status, decreased self-reported episodes of illness and decreased HIV prevalence. 25,30,31 Another systematic review found that the impact of unconditional cash transfers on health services use and health outcomes was uncertain. 28 None of the interventions we identified had a completely unconditional cash transfer intervention. The incentive of a conditional intervention may be particularly important in tuberculosis care, where consistent adherence to a multiple-drug regimen for a prolonged treatment course is essential for optimal treatment outcomes. However, a metaanalysis of the effect of incentives and/ or enablers on medication adherence in tuberculosis was largely inconclusive, but primarily identified studies in high-income countries, where financial interventions may have less effect. 32 Tuberculosis-sensitive interventions are likely to lack a tuberculosis-specific incentive, although they may include other conditional elements unrelated to     22 The studies showed substantial heterogeneity in study design. However, there was no measured heterogeneity within the subset of studies with tuberculosis-specific interventions that were not at high overall risk of bias. Although factors related to the population, setting and intervention could cause heterogeneity in the effect size of the interventions, the available information from the limited number of studies did not allow us to determine the impact of these variables.
Whether cash transfers or goods and services, such as direct provision of food, vocational training, psychologic support and housing programmes, are preferable to improve health-related and other outcomes is currently under debate. [33][34][35] A recent meta-analysis found that noncash socioeconomic interventions, pre-dominantly food provision, may improve clinical outcomes in active tuberculosis. 36 To better understand which forms of social protection are most effective at improving clinical outcomes for tuberculosis, non-cash strategies should be studied comparatively and in combination with cash transfer interventions.
While beyond the scope of this review, the impact of cash transfer interventions on household and national or subnational outcomes, like contact screening and overall tuberculosis incidence, must also be considered. For example, a multivariable analysis found that municipalities in Brazil with higher coverage by the Bolsa Familia programme had a significant reduction in tuberculosis incidence compared to those with lower coverage. 37 In conclusion, we found some evidence that cash transfer interventions improve treatment outcomes in patients with active pulmonary tuberculosis in low-and middle-income countries, although the overall quality of this evidence is low. These findings support calls by WHO and others to incorporate cash transfer interventions into social protection schemes within tuberculosis treatment programmes. 1,6 In addition, high-quality research is needed to better understand the effectiveness of tuberculosis-specific and tuberculosissensitive cash transfer interventions, including understanding of the optimal amount, conditional feature, delivery method and implementation strategy. ■ Systematic reviews Cash interventions to improve tuberculosis outcomes Aaron Richterman et al.
Conclusión Las pruebas disponibles indican que los pacientes de países con ingresos entre bajos y medios que reciben dinero en efectivo durante el tratamiento de la tuberculosis pulmonar activa tienen más probabilidades de obtener un resultado clínico positivo. Estas conclusiones apoyan la incorporación de las intervenciones de transferencias de efectivo en los planes de protección social dentro de los programas para el tratamiento de la tuberculosis.