Conditional Cash Transfer Program and Leprosy Incidence: Analysis of 12.9 Million Families From the 100 Million Brazilian Cohort

Abstract Leprosy is a neglected tropical disease predominately affecting poor and marginalized populations. To test the hypothesis that poverty-alleviating policies might be associated with reduced leprosy incidence, we evaluated the association between the Brazilian Bolsa Familia (BFP) conditional cash transfer program and new leprosy case detection using linked records from 12,949,730 families in the 100 Million Brazilian Cohort (2007–2014). After propensity score matching BFP beneficiary to nonbeneficiary families, we used Mantel-Haenszel tests and Poisson regressions to estimate incidence rate ratios for new leprosy case detection and secondary endpoints related to operational classification and leprosy-associated disabilities at diagnosis. Overall, cumulative leprosy incidence was 17.4/100,000 person-years at risk (95% CI: 17.1, 17.7) and markedly higher in “priority” (high-burden) versus “nonpriority” (low-burden) municipalities (22.8/100,000 person-years at risk, 95% confidence interval (CI): 22.2, 23.3, compared with 14.3/100,000 person-years at risk, 95% CI: 14.0, 14.7). After matching, BFP participation was not associated with leprosy incidence overall (incidence rate ratio (IRR)Poisson = 0.97, 95% CI: 0.90, 1.04) but was associated with lower leprosy incidence when restricted to families living in high-burden municipalities (IRRPoisson = 0.86, 95% CI: 0.77, 0.96). In high-burden municipalities, the association was particularly pronounced for paucibacillary cases (IRRPoisson = 0.82, 95% CI: 0.68, 0.98) and cases with leprosy-associated disabilities (IRRPoisson = 0.79, 95% CI: 0.65, 0.97). These findings provide policy-relevant evidence that social policies might contribute to ongoing leprosy control efforts in high-burden communities.


The 100 Million Brazilian Cohort
The 100 Million Brazilian Cohort baseline is an open cohort using data linkage built by the Centre of Data and Knowledge Integration for Health (CIDACS/FIOCRUZ) [1,2]. The cohort is based on the idea of a "cohort baseline" with information of over 114 million individuals that were registered during 2001 and 2015 in the Brazilian National Registry for Social Programs -Cadastro Unico (CadUnico), which can be exposed to a certain social interventions and can be assessed on different health outcomes.
For the purpose of this study, we considered the intervention as the Bolsa Familia Program (BFP) and our study outcome was new cases of leprosy registered in the Brazilian Notifiable Disease Registry (SINAN).
The baseline of the cohort that was produced by Cidacs/Fiocruz (Version 1) and linked with SINAN was available for the researches in January 2018 [3].

Brazilian National Registry for Social Programs -Cadastro Unico (CadUnico)
Description CadUnico is a national administrative system containing information on all individuals, and their families, applying for any social programs in Brazil [4].

Bolsa Familia program
Description BFP was created in 2003 to alleviate poverty and improve education and health [5]. Its implementation, in 2004, was followed by a formal analysis by the Brazilian government to define the BFP budget for each municipality in Brazil by estimating the number of individuals living in poverty and extreme poverty using the National Household Sample Survey (PNAD) [5]. Extremely poor families were considered to be those living with under 60 BRL per capita a month in 2007 and poor families were those living with under 120 BRL per capita a month in the same year, with frequent adjustments over the following years (Table S1).

Eligibility and cash payments
Enrollment to BFP is conditional on the family being registered in CadUnico and being eligible for the program (i.e., families living in extreme poverty or poverty). Each municipality was responsible for implementing the program and, due to administrative delay, eligible individuals could start receiving benefits at any time after application in CadUnico and consequently, in the 100 Million Brazilian Cohort baseline [2]. BFP benefit is given to a primary recipient, who should preferentially be a woman. Families considered as extremely poor receive a fixed sum plus a supplementary benefit for each applicable family member (i.e., children or adolescents up to 18 years old and women who are pregnant or breastfeeding) with a maximum of five supplementary benefits allocated per family. Families classified as poor receive only the supplementary benefits for applicable family members (Table S1). In 2011, the benefit system was extended to a maximum of seven supplementary benefits per family, and an extra supplement was provided to families who after allocation of BFP benefits remained classified as extremely poor, in order to allow these families to overcome the extreme poverty threshold [5]. Receipt of BFP benefits is subject to compliance with certain conditions: children or adolescents must attend school for at least 80% of school days, both children aged 0-6 and breastfeeding women must be monitored by health professionals, and pregnant women must receive prenatal care. Families receiving BFP benefits are monitored by social assistants to ensure that they maintain compliance with these conditions. In addition, families whose financial situation improves sufficiently, hence no longer meeting BFP poverty eligibility criteria, will continue receiving benefits for up to 2 years after they cease being eligible.

Brazilian Notifiable Disease Registry (SINAN)
Description SINAN was first created in 1993 but only regulated in 1998, becoming the official Brazilian information system for over 40 notifiable health conditions, including leprosy [6]. Following detection of a leprosy case, health professionals, in both private and public services, are required to notify the electronic system, reporting date of detection and clinical and sociodemographic characteristics of the patient and update information regarding the treatment [6].

Methods for and bias assessment
Succeeding linkage between the baseline of the 100 Million Brazilian Cohort and SINAN-leprosy using the CIDACS-RL linkage tool [7], all registries from the baseline cohort received a similarity score (ranging from 0-1) attributed to a possible link with SINAN-leprosy. After that, we randomly select 10,000 linked pairs, and two independent researchers classified the pairs as true or false matches. Linkage accuracy was evaluated by calculating the sensitivity and specificity for various cut-offs of similarity score [7]. For each strata of similarity score, we were able to estimate the true positives, the false positives, the number of missed links (false negatives) and classified the remaining registries as non-links.
In addition, to evaluating potential bias due to the linkage process, we assessed whether the incidence of leprosy cases in the baseline of the 100 Million Brazilian Cohort was representative of the expected leprosy incidence in the total study population and by sociodemographic categories. We used sociodemographic characteristics available from the 2010 Brazilian population census which were also available in the cohort baseline and SINAN-leprosy: sex, age, ethnicity, region of residence and living in rural versus urban areas.
We estimated the mean new case detection rate (NCDR) in the Brazilian population during the study period, as the ratio of yearly mean new leprosy cases from 2007-2014 obtained from SINAN-leprosy, over the total Brazilian population in 2010 [8]. The NCDR estimated from the total population was used to calculate the  (Table   S2). There were no major differences in proportion of expected and observed linked cases overall, by sex or area of residence. Nevertheless, we observed that 15% fewer cases were linked among individuals with brown/mixed ethnicity, 7% fewer cases were linked in the Northeast, and 6% fewer cases were linked in the North region of Brazil. We also observed more cases linked among individuals over 25.

Definition of exposure groups
If eligible, families can start receiving BFP in any point in time after enrolment in cohort, but the majority of BFP beneficiaries start receiving BFP benefit within 6 months. Therefore, we defined exposure to BFP depending on whether households received benefits within 6 months after registration to the cohort. Therefore, families contributed to the analyses only as beneficiary or as non-beneficiary families.
We defined three datasets comprising: i) the overall sample, ii) households living in high burden municipalities for leprosy, and iii) households not living in priority municipalities for leprosy. In each of the three datasets, we excluded households that applied after July 2014 and stayed in our cohort for less than 6 months, and households where the first new leprosy case occurred before application to the cohort or during the first 6 months (see Figure S2).

Estimating logistic regression for each groups of exposure
For each dataset, we estimated the probability of receiving BFP benefit given the baseline covariates using multiple logistic regression (Table S3). Covariates included sex, age, ethnicity, education level and work of the head of the household (oldest family member), region and area of residence, house ownership, housing material, indicators of basic sanitation services (i.e., water supply, electricity, sewage and waste collection), per capita income in quintiles and year of application in the cohort baseline. Propensity score distribution for each dataset is shown in Figure S3.

Matching and post-estimation
Using 1:1 nearest-neighboring matching with a caliper of 0.05, we were able to match: all but 102 beneficiary families within 6 months in the overall sample, all but 19 beneficiary families living in highburden municipalities for leprosy, and all but 153 families living in non-high-burden municipalities for leprosy ( Figure S2). In the overall Brazilian sample, 0.6% (24,437/4,272,847) of the non-BFP families were matched 10 or more times with BFP families; 1% (18,744/1,837,065) of the non-BFP families living in high-burden leprosy municipalities; and 0.3% (7,090/2,435,712) of the non-BFP families living in nonpriority municipalities for leprosy control were matched 10 or more times with BFP families living in the same type of municipalities. Table S4 shows the distribution and the estimated standardized mean difference (SMD) across matching covariates between families exposed and non-exposed to BFP benefit in each of the three datasets after matching (Table S4). We also show, for the overall Brazilian matched cohort, the SMD before and after matching ( Figure S4).

Definition of exposure groups
To account for the fact that some families had a delay to over 3 years between application to the cohort and first receipt of BFP benefits, we defined exposure to vary over time. To account for that in the matching procedures, in addition to defining BFP families as those that start or did not start receiving BFP within 6 months, we also performed 3 additional analysis in order to, in each of them, include families exposed and families not-exposed to BFP in each given amount of time after applying to the cohort. In the first analysis, we defined as BFP beneficiary families those that start receiving BFP between the sixth month and 1 year of registration. In the second analysis, we defined as BFP beneficiary families those that start receiving BFP between 1 and 2 years of registration. In the third analysis, we defined as BFP beneficiary families those that start receiving BFP between 2 and 3 years of registration. Families that started receiving BFP after 3 years of registration were considered to be non-beneficiary families.
In each of the three additional analysis, we excluded (i) BFP beneficiary families already included in the previous analysis (e.g., we excluded families that were already defined as exposed to BFP within 6 months in the analysis considering BFP between 0.5 and 1 year after registration). Also, as in the primary analysis, we excluded; (ii) families where the first new leprosy case within the family unit occurred during the period of BFP definitions (e.g., for those families defined as beneficiaries or non-beneficiaries within six months and 1 year after applying to the cohort, we excluded all leprosy cases that occurred within the same period).

Estimating logistic regression for each groups of exposure, matching and postestimation
For each of the exposed groups (i.e., three additional datasets), we estimated the probability of receiving BFP benefit given the same baseline covariates of the primary analysis using multiple logistic regression.
To perform the matching, we used 1:1 nearest-neighboring matching with a caliper of 0·05. The datasets were matched separately and combined to the primary analysis producing a single/full matched cohort.
Using this approach, families could be matched first as unexposed families and later as exposed, but never twice in the exposed groups. As in the primary analysis, after matching, we estimated leprosy incidence rate ratio between BFP and non-BFP families both using Mantel-Haenszel method and by using Poisson regression further adjusting for income and accounting for cluster robust standard errors by family. We performed the analysis for all families and separately for those living in high and low-burden leprosy municipalities.

Follow-up time calculation
As previously described for the primary analysis, for each matched cohort, the contribution of person-years at risk for each family began at the point where BFP exposure was defined (i.e., at 6 months, 1 year, 2 years and 3 years after registration in the cohort) and ended on December 31st 2014 or at diagnosis of the first new leprosy case in the family. Additionally, for unexposed families who later became exposed, the contribution of person-years for that family ended at the time of first receiving BFP benefits.

Results
The majority of families (4,328,630 families) received BFP within 6 months of applying to our cohort (primary analysis), 1,030,528 families received BFP within 6 months to 1 year (Group 2), 1,090,132 families received BFP between 1 to 2 years after applying (Group 3), and 453,930 received BFP between 2 to 3 years of applying to the cohort (Group 4). After combining the three groups with the primary analysis, our analysis included 13,506,522 matched families and yielded similar results to the primary analysis (Table   S6).

INVERSE PROPABILITY OF THE TREATMENT WEIGHTING (IPTW)
We used the same framework of analysis of the propensity score matching to estimate the effect of the treatment on the treated (ATT) using weights. First, we estimated the propensity score (ps) of receiving the BFP given the sociodemographic covariates of the cohort baseline for the overall sample and separately for each group of families living in high-burden or non-high burden municipalities for leprosy. Second, we estimated the weights for BFP beneficiary families (weight=1) and for non-BFP beneficiary families (weight= E(ps)/(1-E(ps)).
We estimated the incidence rate ratios (IRRs) of new case detection of leprosy for BFP and non-BFP beneficiary families for the overall sample (Table S6). The IRR was estimated using Poisson regression using inverse probability of treatment weighting (IPTW) with further adjustment for income. We investigated the dose-response effect of BFP participation on leprosy using analyses stratified by duration of exposure to BFP benefits (i.e., 0-6 months of fyr, 6-12 months, 1-2 years, 2-3 years, and 3+ years).
Similarly to the main analyses using propensity score matching, we also investigated the association of BFP participation with the secondary outcomes of operational classification (i.e., PB versus MB) and presence of disabilities at diagnosis (i.e., G0D versus G1D/G2D).

LEPROSY TRENDS AMONG BENEFICIARIES AND NON-BENEFICIARIES
We estimated leprosy trends in our dynamic cohort during 1st January 2007 to 31st December 2014 per semester of application into the study cohort. To do that, for each semester, we calculated the number of leprosy cases detected in the specific semester divided by the cumulative number of individuals among applicant families. The NCDR per 100,000 individuals per semester was calculated for BFP beneficiary and non-beneficiary families within 6 months after application to the cohort (Web Figure S2).

Eligibility for BFP in Brazilian Reais Benefit in Brazilian Reais
Year

Web Table 4 Standardized mean differences (SMD) between BFP beneficiaries (BFP) and non-beneficiaries (non-BFP) after matching for the overall sample of
Brazil and in subsamples of high and non-high-burden municipalities for leprosy.