Study design and setting

This is a retrospective cohort study with data from Brazil.

Regarding the study setting, Brazil is the largest country in South America by land area; it has 26 states and a Federal District (Brasília, the capital). The country is divided into five major regions: North, Northeast, South, Southeast, and Central-West, with a total of 5,570 municipalities (Fig. 1). In 2018, 64% of the population was covered by the Family Health Strategy, and 49% of Basic Health Units used electronic medical records, though this rate varied across regions, from 29% in the Northeast to 87% in the South¹⁷.

Created using ArcGIS 10.5 (https://www.esri.com/).

The country is located in South America. With an area of 8,510,417.771 km², it is considered the fifth largest country in the world by territorial extension. In South America, it occupies nearly 50% of the total area, being the largest country in the region, as well as the third largest country in the American continent¹⁸. According to the Brazilian Institute of Geography and Statistics (IBGE), the Brazilian population reached 203,080,756 inhabitants on August 1, 2022¹⁹. Brazil has a Human Development Index (HDI) of 0.760, considered high for the country, and a Gini Coefficient of 0.59, indicating significant social inequality, with a poverty rate of 31.6%²⁰.

Study population and information sources

The study population included cases of DM diagnosed with COVID-19, recorded in the Influenza Epidemiological Surveillance Information System (SIVEP-GRIPE), and cases of DM diagnosed with TB, recorded in the Notification Diseases Information System (SINAN) of the Department of Informatics of the Unified Health System (DATASUS), in Brazil from 2020 to 2022. Both systems aim for epidemiological surveillance and monitoring of disease cases. These systems have retroactive data, allowing for the registration and consultation of information online throughout the patient’s treatment journey. The systems aim to maintain a unique record for each patient, where the entire treatment history is entered. For each patient, one or more treatments are opened, and follow-ups are recorded for each treatment²¹.

Statistical analysis

Spatial analysis

To detect spatial clusters of the comorbidities DM-TB and DM-COVID-19, the technique known as Scan Statistics or Scan Statistic was used. Developed by Kulldorff and Nagarwalla (1995)²², this technique creates circles that cover the entire study area centered on the centroids of each analyzed territorial unit. The radius of these circles can range from zero up to a limit determined by the researcher²³.

The identification of spatial clusters occurs by calculating the number of events within each circle. If the observed number is significantly higher than expected in the region z defined by the circle, a cluster is identified. Otherwise, the radius of the circle is increased to test a new centroid, and this process is repeated until all centroids are evaluated. The hypotheses tested are as follows: H0 indicates the absence of clusters in the study region, while H1 suggests that region z is a cluster²⁴.

The numerous generated circles are tested using Monte Carlo simulations based on the null hypothesis. During the centralization process, the Log Likelihood Ratio (LLR) of each possible cluster is calculated from the difference between the observed and expected incidence within and outside the circular window, where a p-value is assigned, according to the following formula²⁵:

where “O” represents the observed cases, “E” the expected cases, “O_in” and “E_in” denote the observed and expected numbers within the circular window, respectively, and “E_in” is calculated by multiplying TB deaths by the population of the census sectors. The larger the LLR value, the lower the probability that the detected cluster occurred by chance. Additionally, the Spatial Relative Risk (SRR)^26,27 is calculated for each statistically significant cluster, comparing the risk within the cluster to the risk outside of it²⁸. The spatial scanning analysis was conducted using SaTScan 10.0.2 software (https://www.satscan.org/).

After the cluster is identified, the software calculates the SRR, obtained using the following formula²⁹:

where “N” is the total number of cases, “N_Z” is the number of cases in cluster Z, “E_A” is the expected number of cases in the region under the null hypothesis, and “E_Z” is the expected number of cases in area Z under the null hypothesis. The SRR value determines whether the cluster is of risk (SRR > 1) or protection (SRR < 1)^30,31.

Additionally, thematic maps were created from the scanning analyses, displaying the risk clusters identified using ArcGIS 10.5 software (https://www.esri.com/).

Binary logistic regression

To investigate the main factors associated with the outcomes of the comorbidities DM-TB and DM-COVID-19, binary logistic regression analysis was conducted based on variables from the SIVEP-GRIPE and DATASUS databases. Two separate analyses were performed, one for TB-DM and another for DM-COVID-19, with the objective of identifying the factors associated with unfavorable outcomes for both comorbidities. The dependent variable was dichotomized as unfavorable outcome (1) and favorable outcome (0), as illustrated in Fig. 2.

For DM-TB, the following variables were considered: age group, gender, race/ethnicity, pregnancy status (yes/no), education level, type of entry (new case, TB discovered post-mortem, TB relapse, TB relapse after abandonment), histopathology (positive, suggestive of TB, not suggestive of TB), Directly Observed Treatment (DOT) (yes/no), life history (alcoholism, mental illness, incarceration, homelessness, immigrant status, drug use, smoking), diagnosis by rapid molecular test for TB, drug resistance (isoniazid resistance, isoniazid and rifampin resistance, resistance to other drugs), and HIV testing.

For DM-COVID-19, the following variables were considered: age group, gender, race/ethnicity, education level, diagnostic criterion (COVID-19), pregnancy (1st trimester, 2nd trimester, 3rd trimester), comorbidities (cardiopathy, asthma, Down syndrome, neurological disease, kidney disease, obesity), and COVID-19 vaccination status.

Categorization

The dichotomized dependent variable indicated whether there was a favorable or unfavorable outcome for TB (0 and 1, respectively). This variable was analyzed in conjunction with all independent variables, which were also dichotomized (0 and 1).

The independent variables considered for the DM-TB regression, along with the dichotomization process into 0 and 1 values, were as follows:

• Age: 1 to 5 years (1 = Yes; 0 = No); 6 to 19 years (1 = Yes; 0 = No); 40 to 49 years (1 = Yes; 0 = No); 50 to 59 years (1 = Yes; 0 = No); 60 to 69 years (1 = Yes; 0 = No); 70 to 79 years (1 = Yes; 0 = No); 80 to 89 years (1 = Yes; 0 = No); above 90 years (1 = Yes; 0 = No).

• Sex: (1 = Male; 0 = Female).

• Pregnancy: (1 = Yes; 0 = No).

• Race/Color: White (1 = Yes; 0 = No); Black (1 = Yes; 0 = No); Yellow (1 = Yes; 0 = No); Brown (1 = Yes; 0 = No); Indigenous (1 = Yes; 0 = No).

• Educational level: No schooling (0 = Yes; 1 = No); Complete high school (1 = Yes; 0 = No); Complete higher education (1 = Yes; 0 = No); Incomplete higher education (1 = Yes; 0 = No).

• Case classification: TB discovered post-mortem (1 = Yes; 0 = No); TB relapse (1 = Yes; 0 = No); TB relapse post-abandonment (1 = Yes; 0 = No).

• Directly Observed Treatment (DOT): (1 = Received; 0 = Do not received).

• Histopathology: (1 = Positive; 0 = Negative).

• Comorbidities and drug use: Alcoholism (1 = Yes; 0 = No); Mental illness (1 = Yes; 0 = No); HIV status (1 = Positive; 0 = Negative); Illicit drug use (1 = Yes; 0 = No);

• Smoking (0 = Yes; 1 = No).

• Vulnerable populations: Incarceration (1 = Yes; 0 = No); Homelessness (1 = Yes; 0 = No); Immigrant (1 = Yes; 0 = No).

• Government beneficiary: (1 = Yes; 0 = No).

• Diagnosis by Tuberculosis Rapid Molecular Test—TRM-TB: (1 = Yes; 0 = No).

• Drug resistance: Isoniazid resistance (1 = No; 0 = Yes); Isoniazid and rifampicin resistance (1 = No; 0 = Yes); Resistance to other drugs (1 = No; 0 = Yes).

The independent variables considered for DM-COVID-19 regression:

• ICU Admission: (1 = Yes; 0 = No).

• Sex: (1 = Male; 0 = Female).

• Diagnostic criterion: Laboratory (1 = Yes; 0 = No); Epidemiological (1 = Yes; 0 = No); Clinical (1 = Yes; 0 = No); Imaging (1 = Yes; 0 = No).

• Age: 1 to 5 years (1 = Yes; 0 = No); Age 6 to 19 years (1 = Yes; 0 = No); Age 30 to 49 years (1 = Yes; 0 = No); Age 50 to 59 years (1 = Yes; 0 = No); Age 60 to 69 years (1 = Yes; 0 = No); Age 70 to 79 years (1 = Yes; 0 = No); Age 80 to 89 years (1 = Yes; 0 = No); Age 90 years or older (1 = Yes; 0 = No).

• Gestational age: Pregnant—1st Trimester (0 = Yes; 1 = No); Pregnant—2nd Trimester (0 = Yes; 1 = No); Pregnant—3rd Trimester (0 = Yes; 1 = No).

• Race/Color: White (1 = Yes; 0 = No); Black (1 = Yes; 0 = No); Yellow (1 = Yes; 0 = No); Brown (1 = Yes; 0 = No); Indigenous (1 = Yes; 0 = No).

• Educational level: No schooling (1 = Yes; 0 = No); Incomplete Elementary School (1 = Yes; 0 = No); Complete Elementary School (1 = Yes; 0 = No); Complete High School (1 = Yes; 0 = No); Complete Higher Education (1 = Yes; 0 = No).

• Comorbidities: Cardiopathy (1 = Yes; 0 = No); Down Syndrome (1 = Yes; 0 = No); Asthma (1 = Yes; 0 = No); Neurological Disease (1 = Yes; 0 = No); renal disease (1 = Yes; 0 = No); Obesity (1 = Yes; 0 = No).

• COVID-19 Vaccination Scheme: (1 = Complete; 0 = Not completed).

Statistical analysis

Initially, exploratory analyses were conducted to check for collinearity among independent variables using the Variance Inflation Factor (VIF). Variables with a VIF greater than 10 were removed from the statistical modeling ^32,33.

Modeling was performed using the Backward stepwise selection method, starting with a full model and removing one variable at a time based on the minimization of the Akaike Information Criterion (AIC) ^34,35. The final model was chosen based on the lowest AIC value, and the Odds Ratio (OR) with 95% Confidence Intervals (CI) was calculated.

To validate the final model, Hosmer–Lemeshow, likelihood ratio, Cox-Snell, Nagelkerke, and McFadden tests were performed. The predictive ability and accuracy of the models were assessed by the area under the ROC (Receiver Operating Characteristic) curve and its corresponding CIs ³⁶. All analyses and validation tests were conducted using RStudio software.

Ethical aspects

This study was approved by the Research Ethics Committee of the Ribeirão Preto School of Nursing, University of São Paulo, in accordance with the Guidelines and Regulatory Standards for Research with Human Subjects, Resolution No. 466/2012 of the National Health Council, under Certificate of Presentation for Ethical Appreciation No. 57933622.4.0000.5393, issued on June, 06th 2022. There was no consent required to participate in the study, as the study was conducted using secondary data from registered in the SINAN and SIVEP-GRIPE databases. The databases used for the study are anonymous, so it is not possible to identify the subjects included in the study. All methods in the study were performed in accordance with relevant guidelines and regulations.