Leprosy physical disabilities in the 100 Million Brazilian Cohort

Leprosy continues to be an important cause of physical disability in endemic countries such as Brazil. Knowledge of determinants of these events may lead to better control measures, as targeted interventions may mitigate their impact on affected individuals. This study investigated such determinants among the most vulnerable portion of the Brazilian population. A large cohort was built from secondary data originated from a national registry of applicants to social benet programs, spanning the period 2001 to 2015 and including over 114 million individuals. Data were linked to the Leprosy disease notication system utilizing data from 2007 until 2014. Descriptive and bivariate analyses lead to the multivariate analysis using a multinomial logistic regression model with cluster-robust standard errors. Associations were reported as Odds Ratios with their respective 95% condence intervals


Background
Chronic infections with Mycobacterium leprae have the potential to cause lasting nerve damage and physical disabilities [1,2]. Among patients with leprosy, physical disabilities arise as a result of late diagnoses and/or insu cient treatments. The incidence of leprosy-associated disabilities among newly detected cases is, therefore, an important indicator of gaps in population-level leprosy control strategies.
Leprosy cases are classi ed as: Grade 0 disability (G0D) when the muscle strength and the sensitivity of these segments are preserved; Grade 1 (G1D) when there is decreased muscle strength and/or decreased sensitivity; and Grade 2 (G2D) when there are visible deformities in the hands, feet, and/or eyes [3,4].
As part of the 2016-2020 Global Leprosy Strategy, the WHO has set a target of reducing the rate of newly diagnosed leprosy patients with G2D to be less than 1 per million population [4]. Within Brazil, a country with a high leprosy new case detection rate of 13.7/100,000 population as of 2018, the National Leprosy Disease Program has similarly prioritized reducing the rate of diagnoses with G2D as a primary goal. From 2012 to 2016, the mean rate of leprosy new case detections with G2D in Brazil was 10.5 per 1 million inhabitants, with an average of 2,042 people diagnosed annually with leprosy-related G2D in this period [5]. In the last decades, Brazil has adopted extensive public health measures to improve the assessment and prevention of leprosy-related physical disabilities [6]. Nevertheless, a systematic review conducted by Vieira et al. (2018) [7], indicates that the proportion of leprosy cases involving disability among children < 15 years remains high in Brazil, re ecting active transmission and challenges for case detection.
Although there have been large-scale studies in Brazil studying the social determinants of leprosy incidence and treatment default [8,9], the risk factors for leprosy-associated disability at the time of diagnosis remain inadequately investigated. Using nationwide linked data from the 100 Million Brazilian Cohort, this study seeks to use large-scale data to identify risk factors for having leprosy-related physical disabilities at the time of diagnosis.

Study Design and Population
The 100 Million Brazilian Cohort [10,11] was built by linking the health and administrative records of individuals registered in the Cadastro Único para Programas Sociais (CadÚnico), a national registry for social assistance programs in the country. This database was created at the Centro de Integração de Dados e Conhecimentos para Saúde at Oswaldo Cruz Foundation (Cidacs, Salvador, Bahia, Brazil) and is part of the Center's mission to evaluate the impact of social determinants and policies on health. The cohort includes administrative records from over 114 million individuals who applied for social assistance between 2001 and 2015.
For this study, we extracted a de-identi ed dataset including exposure data related to geographic factors (i.e. region, urbanicity, and residence in a 'high-burden cluster'), household living conditions (i.e. household density, housing materials, water source, electricity source, sewage disposal, and waste disposal), family per capita income (i.e. relative to the Brazilian minimum wage), and individual socioeconomic indicators (i.e. sex, age, self-identi ed race/ethnicity, literacy, educational attainment, and employment status). For children younger than 18 years, education and employment were reported as the education level and employment status of the head of the family (here de ned as the oldest member of the family). Among the geographic factors examined, we investigated risks within clusters of higher incidence, as de ned by Penna et al. (2009) [12] based on epidemiological data from 1980 until 2007.
These clusters were described as 29 spatial clusters comprising 789 municipalities and were devised to facilitate decision-making for leprosy control in Brazil. Although these were de ned more than ten years ago, a recent study [13] analyzed the spatial distribution of leprosy in selected endemic regions of the country comparing the periods 2001-2003 versus 2010-2012 and concluded that there is signi cant overlap of clusters comparing both time periods.
As previously described [8,9,14,15] the data from the 100 Million Brazilian Cohort was then linked to leprosy noti cation records from 2007 to 2014 in the national noti able disease system, Sistema de Informação de Agravos de Noti cação, SINAN-leprosy. Additional variables from the SINAN data included the operational classi cation of leprosy (i.e. paucibacillary or multibacillary [PB or MB]), the number of skin lesions, and the grade of disability at diagnosis.
The study population for this investigation included members of the 100 Million Brazilian Cohort followed from 1 January 2007 and 31 December 2014. Cohort members were excluded if they: (i) were diagnosed with leprosy prior to registration in CadÚnico, (ii) belonged to family units without one member aged over 15 years (i.e., children registered separately from their families), (iii) had less than one day of follow-up, (iv) were relapsed leprosy cases or (v) did not have information on grade of disability at diagnosis.

Statistical Analysis
Descriptive analysis was performed to assess the distribution of the independent variables, followed by bivariate analysis with the outcome (presence of any degree of disability) to assess the strength of association between independent variables and grade of disability at diagnostic. Those with a p-value less than 0.1 were considered eligible for the multivariate model.
For the multivariate analysis, a multinomial logistic regression model with cluster-robust standard errors (i.e., accounting for familial clustering of covariates) to estimate the adjusted odds ratios (OR) was used, with grade zero disability cases used as the reference category.

Results
The study included 21,565 new leprosy cases, from the original cohort of 33,905,423 individuals ( Fig. 1). At the time of diagnosis, 15,095 (63.1%) cases had G0D, while grades 1 and 2 represented 21% (5,026) and 6% (1,444), respectively. In the multivariate model, 16,376 cases were included, as missing values for some variables prevented a number of cases from being included.
In the study population, the mean age of newly detected leprosy cases was 37.6 years old, varying by the grade of disability (G0D 34.9; G1D 43.4; G2D 45.7) ( Table 1). Overall, cases were equally distributed by sex, and 49.6% of the cases were female. The majority of the cases identi ed as mixed race ("pardo") (72.1%), had up to 4 years of schooling (61%) and were literate (79.3%). Almost half (49.1%) were employed. Although 81.8% earned up to 0.5 minimum wage, 11.3% reported no source of income. Most lived in urban areas (79.5%), and the regions with the highest percentage of cases were the Northeast (40.4%) and the North (23.6%). The majority (69.3%) lived in brick or cement-made dwellings, with publicly provided water, garbage collection and electricity. However, 67% of reported using a homemade tank as a sewage disposal system. The majority of the cases lived in municipalities that belonged to the epidemiologically-de ned high incidence clusters (63.8%) There were more MB cases than PB (59.2 vs 40.8%) at time of diagnosis (Table 1).
As far as G2D is concerned, those older than 15 years old were more likely to present it (OR 1.95; 95%CI 1.51-2.50) and having less education also showed a similar association, with a clear dose response gradient (ORs for no schooling, ≤ 10 years of education, > 10 and ≤ 14 years were respectively 1.91, 1.64, 1.31, all p-values < 0.05). Being a multibacillary patient was also a risk factor (OR 8.22; 95%CI 6.51-10.38). On the other hand, being female (OR 0.61; 95%CI 0.53-0.70), and living in a high incidence municipality (OR 0.67; 95%CI 0.58-0.78) decreased the odds of presenting G2D at diagnosis (Table 2). Other protective factors included living in the north, Northeast and centre-west region.

Discussion
This study investigated factors associated with leprosy-related disability in a large Brazilian set of cases. Among 16,376 new cases of leprosy analysed, were less likely to present grade 1 or grade 2 physical disabilities women, those living in the North, Northeast and Center-West regions or in high incidence clusters, in urban areas and with greater household crowding. In contrast, being over 15 years of age, with a lower level of schooling, not working and being multibacillary increased the chances of presenting grade 1 or grade 2 physical disabilities.
The higher likelihood of leprosy-related disabilities found among those older than 15 years is similar to previous studies. In a hyperendemic area of the Center-West region of Brazil, the estimated risk ratio of grade 2 disability was 5.3 times higher among patients aged ≥ 45 years [16]. In the state of Minas Gerais, a retrospective study showed that age above 15 years was an important risk factor for the development of physical disability in leprosy patients [17]. A study of patients residents in the state of Maranhão showed a progressive increase in the chances of developing physical disability among those older than 15 years, ranging from 3 to 10.4 times more [18]. Considering the duration of the disease is directly related to age and, given the chronic pro le of leprosy effects, increasing age may result in more advanced disabilities [17,19].
Regarding gender, some studies did not identify an association between gender and the level of disability [20][21][22]. However, as in this study that used data from all over Brazil, other studies report higher grades of physical disability among male individuals with leprosy [17,23].
Men are generally more exposed to M. leprae and have reduced contact with health care, which may delay diagnosis and increases the risk of developing physical disabilities [24]. Data from the Ministry of Health for the general population show that between 2012 and 2016 the detection rate of new cases with physical disability grade 2 was much higher in males with 15.2 cases per 1 million men, while the rate in women was 6.1 cases per 1 million women [5]. Cultural factors may explain the difference between the studies because women may be more likely to seek health assistance than men [18], while men go when the disease is in a more advanced stage.
Our study also suggests that higher levels of education were negatively associated with the presence of physical disabilities at diagnosis, which is consistent with the literature [16,17]. Higher education may be associated with a better understanding of the disease and, consequently, better access and utilization of health services. Regular treatment and evaluation, as well as self-care, are aspects that prevent the worsening of leprosy cases [17,25].
The fact that cases from the Northeast and the North regions were less likely to present G1D and G2D contrasts the ndings from Freitas and colleagues (2016) [13], which showed greater proportions of G2D in municipalities with higher incidence rates of leprosy. They hypothesize that "in these municipalities, at least in the short term, a consequence of increased surveillance actions may be the initial increase in the 'detected' cases of the disease. In turn, this increase may lead to increased tracing of people who have had contact with it and greater detection of cases with grade 2 disability, which previously were not identi ed. This hypothesis may explain the nding that municipalities with a greater proportion of cases presenting with grade 2 disability also had higher average leprosy incidence rates". However, the areas with higher endemicity do not have a better structure surveillance and care system, as they are systematically poorer. The clusters are located in areas that are more vulnerable.
Therefore, we hypothesize that this fact is likely due to a more sensitive health staff and surveillance system to case detection, therefore more capable of detecting leprosy cases earlier. Assuming that disability is a marker for late diagnosis, it is expected that regions of high endemicity will show a lower chance of patients presenting with grades 1 and 2 disability. G2D, as already mentioned may indicate a late diagnosis and a suboptimal surveillance system. According to Penna et al. (2009) [12], access to primary health care units has improved mainly in rural areas and small towns, improving the diagnosis of leprosy in the rst decade of this century. However, as her work emphasizes, "the diagnosis of skin diseases depends on the cultural importance given to skin lesions, as well as health-seeking habits among the population." The study by Freitas et al. (2016) [13] looked at risk factors, estimated rate ratios (RR), and identi ed a high NCDR in the Midwest and North regions compared to the South, large cities and greater urbanization, median and high illiteracy rate, income inequality (Gini index), domiciles' agglomeration, worse sanitation condition, and percentage of cases with grade 2-disability.
Although we found similar evidence that individuals living in urban areas were at a greater risk of leprosy detection than individuals living in rural areas, we did not nd evidence of an association of household density with leprosy risk in the full cohort. It is, however, noteworthy that in subgroup analyses increased household density (more than one resident per two rooms) was associated with an increased leprosy risk in children, a group indicative of active transmission [8].
The association between the proportion of multibacillary leprosy and presentation of G2D has been shown in the past [16,26,27]. Studies conducted in some Brazilian cities indicate that at the time of diagnosis, the educational level variables and operational classi cation are statistically associated with the development of physical disabilities (p < 0.05). It is emphasized that multibacillary individuals are twice as likely to develop sequelae as paucibacillary individuals [28].
Our study has several strengths: the large sample size and extensive follow up period allowed us to evaluate determinants of disability to the extent that is rarely possible. This study linked data from over 100 million individuals and was able to assess factors associated with physical disability in an unprecedented way, also as we were able to evaluate a wider range of variables present in CadÚnico. Unlike other studies, we analyzed the most vulnerable fraction of the Brazilian population, as this is the pro le of individuals enrolled in CadUnico.
Nevertheless, our study has some limitations. The use of secondary data originated from routine surveillance activities always brings the issue of completeness of information. We did not have complete information on disability evaluation at diagnosis (n = 1,557) and at discharge. The latter was poorly collected to an extent that does not permit to analyze. Efforts should be undertaken to stress the importance of performing this evaluation at discharge and record it in the information systems. Other factors associated with disability were not available in our database and therefore, could not be assessed, such as health services characteristics and patients' perception and knowledge about leprosy.

Conclusion
Our study indicated that worse socioeconomic conditions might act as barriers to early diagnosis, which increases the risk of developing physical disabilities. Our ndings suggest the need for early and quali ed diagnosis of leprosy in endemic regions and especially in regions considered non-endemic that may present high rates of hidden prevalence. Focus could be given to younger patients, which re ect recent and active transmission. It is imperative to train staff in less endemic areas to become sensitive to leprosy, aiming at reducing under detection. In addition, data linkage proved a powerful tool to shed more light on identifying potential causal factors of disabilities among the poorest Brazilian population.
There is a need for further studies on GIF-related socioeconomic and clinical factors at the end of treatment, issues we could not address in our study. Future research should also explore if the ndings found in this work will replicate among relapses or reinfections; considering we focused exclusively on new cases. Besides early diagnosis and timely treatment, social protection policies and initiatives are key to lead us to effective leprosy control -evidence that has been put forth a century ago[29] and yet remains valid.

Consent for publication
Not applicable

Availability of data and materials
The data that support the ndings of this study are available from Center of Data and Knowledge Integration for Health (Cidacs -https://cidacs.bahia. ocruz.br/) but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of Cidacs.

Competing interests
The authors declare that they have no competing interests.  1 Univariate logistic regression model accounting for household cluster. 2 Final model of multinomial logistic regression accounting for household cluster with exclusion of the missing data. 3 For all the tests and for permanence of the variables in the final model was used the significance level of 5%. Figure 1 Study population selection owchart from the 100 Million Brazilian Cohort.