A cohort study was conducted using households living in two favelas—Heliópolis and Vila São José—in São Paulo, Brazil. The inclusion criterion was people living in the aforementioned favelas. Heliópolis favela is the largest and most densely populated shantytown located close to the downtown area of São Paulo city. Vila São José is a favela located south of São Paulo city. Further characterization of the study site can be found in Manfrinato et al. .
The study began in the early period of the pandemic, in April 2020, and the same householders were interviewed six months apart: October 2020, and April 2021.
A standardized online questionnaire was employed to derive data on socioeconomic and demographic characteristics, finances, and FI during three periods of the study. Interviewers were previously trained.
In wave 1, an online questionnaire was conducted by the observatory ‘De Olho na Quebrada’ (Eyes in the Quebrada, slang for “slum”) in Heliópolis for one week. In Vila São José, the community leader sent a questionnaire to households that he had contact with. In both communities, questionnaires were sent to some households through an instant messaging application, WhatsApp, and were then shared with other residents in these communities. Participants reported their skin color as defined by the Brazilian Institute of Geography and Statistics as White, Mulatto, Black, Yellow, or Indigenous. Data about participants’ education level, occupation, number of children and elderly people in the household, existing morbidity among residents in the household, number of residents in the household, family income, and social assistance coverage from (i) the Bolsa Família Program (BFP) (federal conditional cash transfer program); (ii) the Auxílio emergencial (federal cash transfer offered during the COVID-19 pandemic); (iii) food kits called cesta básica; and (iv) tickets for students (a city program offered during school closure). The Auxílio emergencial fund was offered to address political, sanitary, and economic crises. This fund was available from April to December 2020, Int$254 (R$600), and ended at half the value. From April to October 2021, it was available at a quarter of the value (Int$64).
During waves 2 and 3, trained interviewers collected information via phone. Socioeconomic, demographic, social assistance program coverage, and additional information about difficulties related to buying gas canisters for cooking (electric stoves are uncommon in Brazil), water supply for cooking, handwashing, and showers, hand sanitizers, and personal hygiene products (toothpaste, shampoo, and hand soap). In the third wave, data on rental expenses were collected. If participants did not respond after three contact attempts on different days and times, this was considered follow-up loss.
Family income was assessed, including monetary value received from the social assistance program. The monetary values in Reais (R$) to international dollars (Int$) were calculated while considering the purchasing power parity for 2020 (conversion factor 2·362) . The Int$ currency is used to compare the purchasing power of basket goods and services prices in different currencies from the Organisation for Economic Co-operation and Development (OECD).
Food insecurity experience
Data on FI were collected according to the Brazilian Food Insecurity Scale—Escala Brasileira de Insegurança Alimentar (EBIA) in Portuguese . The short version of the scale comprises five questions and presents high sensitivity and specificity, compared to the original EBIA scale. The short version was not classified based on FI severity . However, it is useful for the fast diagnosis of families in FI and as a dichotomized variable (presence or absence of FI). The householder responsible for food acquisition or preparation was required to respond to the scale. Households that reported FI by answering “yes,” were scored with 1 point, whereas those who responded with “no” were scored with 0 points, with a maximum of five points in the questionnaire. The five questions were based on assessing: (1) anxiety and concerns about the ability to obtain food; (2) whether households were too poor to buy more food; (3) whether the quality and variety of food have been compromised, particularly for nutritious food; (4) quantity reduction; and (5) skipping meals. In the present study, we adapted a short version of the FI over the past month. We created a dichotomized household food security variable: households with food security and quite FI (EBIA total score≤2) and FI when (EBIA total score ≥3). EBIA=1 was considered safe, based on model-based cutoffs of the EBIA in the original scale . Concerns about food acquisition, monotony of food, or reduction of food amount brought about by money shortage are characteristics of households experiencing FI. However, due to the small sample size of households with food security, EBIA=2 was also considered in the same group as food security.
This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving research study participants were approved by the Universidade Federal de São Paulo (CAAE 30805520.7.0000.5505). Online consent was obtained from all participants. The questionnaire only opens after online consent. The present study complies with the guidelines for Strengthening the Reporting of Observational Studies in Epidemiology.
The distribution of sociodemographic and economic characteristics, EBIA questions, and the prevalence of FI severity are presented according to the study period. To test the differences according to each wave, we used the Pearson x2 or Kruskal–Wallis tests.
Persistence of FI during the follow-up
The outcomes of interest were (i) FI persistence from wave 1 to wave 2 (“no,” “yes”) and (ii) FI persistence during the three waves: “no/never,” “sometimes in FI” (FI in one or two waves during the study), or “always in FI”. The independent variables were sociodemographic characteristics in wave 1, per capita income and social assistance coverage in waves 1 and 2, and income changes in different periods (waves 1 and 2; waves 2 and 3; and waves 1 and 3). Sociodemographic variables were gender (male, female), age (continuous), education level (<9, ≥9 years), race/ethnicity (White, non-White), number of dwellers (<4 or ≥4), presence of children aged below 10 years, and presence of elderly people (>60 years old). Economic variables were per capita income (cut-off value was the median value among families in food security for each wave) and social assistance program coverage (whether or not they received assistance) of scholar tickets, BFP, auxílio emergencial, food basket, and number of social assistance programs received. The categories “no/never” or “sometimes” in this variable were combined because of the reduced sample size for “never.”
To identify the factors associated with (i) FI persistence from waves 1 to 2 and (ii) FI persistence during the three waves, the prevalence ratio (PR) was calculated using the Poisson regression model. PR adjustment was conducted while considering a conceptual model . The variables were maintained if a significance level of p≤0·20 was noted. Final model was adjusted for education
A significance level of 5% was adopted. Statistical analyses were performed using STATA 14.0 (Texas, USA).