We used a mixed ecologic study design (temporal and multiple groups). Analysis units were the neighborhoods (geographic) and quarters (temporal) during the study years. The study population consisted of CS cases reported to the Brazil Information System of Notifiable Diseases (SINAN) during 2013–2017 for those living in the city of Rio de Janeiro, which was geocoded by neighborhood. Only municipal health units (HU) patients were included in the study (i.e., notifications from federal/state units and private/psychiatric hospitals were excluded).
The geographical units of study were 160 geocoded neighborhoods in SINAN (Fig. 1). Four data sources were used: (i) SINAN's Gestational Syphilis and Congenital Syphilis registry databases, (ii) reports and spreadsheets on the dispensation of BP provided by the Pharmaceutical Assistance Center of the Municipal Health Secretariat (SMS-Rio), (iii) population data of the census tracts and neighborhoods of the 2010 Census of the Brazilian Institute of Geography and Statistics, and (iv) live birth registry data from the Live Birth Information System of the municipality of Rio de Janeiro.
The disease incidence rate was estimated by dividing the number of CS cases reported in children < 1 year of age by 1,000 live births per trimester in the city of Rio de Janeiro. The study outcome was the significantly increasing incidence rate of CS in Rio de Janeiro, considering the relative difference in the incidence rate of the previous quarter compared with that in the index quarter under analysis.
The significant increase in CS incidence was estimated, as the quarterly rate variation was greater than 200% (estimated percentage considering 2 times the value of the CS incidence compared with the previous quarter). The outcome was classified as dichotomous (yes/no).
The BP supply measure for use in Gestational Syphilis considered the ratio between two Benzatin Pencil Dispenser bottle quantities: (a) the number of bottles dispensed to health facilities in each neighborhood and (b) the number of bottles necessary to treat pregnant women diagnosed with syphilis and their respective partners residing in each neighborhood.
The BP Dispensation data for the HU in each neighborhood of the city of Rio de Janeiro were extracted from electronic spreadsheets provided by SMS-Rio, where there were three ways the number of BP vials were dispensed: (1) for the HU of each neighborhood, (2) to the General Coordinators of Primary Care of the Program Area (CAP) with their respective neighborhoods, and (3) to both the HU in specific neighborhoods and to coordinate the program area of the same coverage area.
To obtain the number of vials dispensed to the HU of each neighborhood, the following estimates were made according to the type of data: (1) for type 1 data, the sum of the number of vials dispensed to the HU of each neighborhood; (2) for type 2 data, the number of vials distributed to each HU in the neighborhoods was first estimated based on the number of notifications of Gestational Syphilis in these units, and then the number of vials in the HU of each neighborhood was summed; and (3) for type 3 data, a combination of the two previous methods was used.
Since not all of the BP that arrived at the HU was used to treat Gestational Syphilis, we assumed that 80% was used to treat pregnant women with syphilis and their partners. This assumption is justified by two factors: (i) BP is used to treat other infections also prevalent in the HU, such as rheumatic fever, upper respiratory tract infections, soft tissue infections (erysipelas, impetigo), pneumonia, and penicillin-sensitive Streptococcus pneumoniae; (ii) in order to stop the growth of CS cases, SMS-Rio prioritized using BP to treat pregnant women with syphilis.
To determine the amount of BP vials needed to treat pregnant women and their partners, the dosage of each individual treatment was considered. In the absence of available data on the clinical classification (staging) of the disease, the therapeutic regimen with three doses of BP was considered, with each dose using two vials of BP (1,200,000 IU), which is six vials used (7,200,000 IU) for the treatment of pregnant women.
The calculation for the number of BP vials needed to treat the pregnant women with reported CS was: the total number of reported cases of gestational syphilis in the HU of each neighborhood times 6, plus the 100% scenario of treated partners, where the scheme was used with three doses. Once obtained from the estimates of the number of dispensed and needed bottles, the supply measure used in this study was calculated by the following formula:
BPS i – benzathine penicillin supply indicator of the neighborhood i;
X i – Number of benzathine penicillin vials dispensed (and used) for treatment of Gestational Syphilis and partners in the neighborhood i;
Y i – Number of vials of benzathine penicillin needed to treat pregnant women and partners in neighborhood i.
The benzathine penicillin supply indicator (BPS) was calculated by quarter, as it was observed that the dispensation did not occur regularly throughout the year. Thus, the quarter was considered as a time unit because it represents the homogeneous dispensation of BP.
As the indicator of supply is a ratio whose values range from 0 to ∞, when BPS had a value between 0 and 0.99, it was considered a shortage (yes), and there was no shortage (no) when BPS was greater or equal 1.
In this paper, we use a theoretical model design based on the studies by Kilmarx et al. (1997)9 and Thomas et al. (1999)10, which aimed to evaluate community factors that influence the incidence of syphilis. Thomas et al. (1999)10 found that factors affecting syphilis transmission can be categorized into those acting at the individual level (e.g., number of sexual partners) or at the socio-physical level, such as availability of treatment services for curable infections. In the model created by Thomas et al. (1999)10, three categories of the socio-physical environment were considered: geographical, architectural, and technological and sociocultural.
Based on this information, the theoretical model used was proposed in four dimensions: sociodemographic, economic, geographical, and prenatal care. Sociodemographic information on the neighborhood populations used included (i) the proportion of women who had children under the age of 20, (ii) the rate of live births, and (iii) the illiteracy rate in the population over 18. In the economic dimension, information covered (iv) the proportion of people living in extreme poverty, (v) the Gini coefficient, and (vi) the unemployment rate. The (vii) homicide rate in the neighborhoods and (viii) proportion of the population living in subnormal clusters was also studied in the geographical dimension. Regarding the prenatal care characteristics of the population, we used (ix) the proportion of pregnant women with a non-treponemal test performed, (x) the proportion of women who had more than six prenatal consultations during their pregnancy, and the (xi) proportion of treated partners.
To evaluate the association between the BP shortage and a significant increase in CS incidence, the negative-inflated zero-binomial regression model (longitudinal model) was used. Crude and adjusted odds ratios were calculated to express the magnitude of the studied associations. The selection of adjustment variables was considered using Directed Acyclic Graphs (DAG). Figure 2 shows the relationships between covariates, exposure, and outcome. This approach also involved the strategy of identifying the minimum set of variables necessary to obtain the adjusted measurements.