An integrated dataset of malaria notifications in the Legal Amazon

Prevalence and correlates of childhood anemia in the MINA-Brazil birth cohort study

OBJECTIVE

This study aimed to describe the prevalence and predictors of childhood anemia in an Amazonian population-based birth cohort study.

METHODS

Prevalence of maternal anemia was estimated at delivery (hemoglobin [Hb] concentration < 110 g/L) in women participating in the MINA-Brazil birth cohort study and in their children, examined at ages one, two (Hb < 110 g/L), and five (Hb < 115 g/L). Moreover, ferritin, soluble transferrin receptor, and C-reactive protein concentrations were measured in mothers at delivery and in their 1- and 2-year-old children to estimate the prevalence of iron deficiency and its contribution to anemia, while adjusting for potential confounders by multiple Poisson regression analysis (adjusted relative risk [RRa]).

RESULTS

The prevalence 95% confidence interval (CI) of maternal anemia, iron deficiency, and iron-deficiency anemia at delivery were 17.3% (14.0-21.0%), 42.6% (38.0-47.2%), and 8.7% (6.3-11.6)%, respectively (n = 462). At one year of age (n = 646), 42.2% (38.7-45.8%) of the study children were anemic, 38.4% (34.6-42.3%) were iron-deficient, and 26.3 (23.0-29.9) had iron-deficiency anemia. At two years of age (n = 761), these values decreased to 12.8% (10.6-15.2%), 18.1% (15.5-21.1%), and 4.1% (2.8-5.7%), respectively; at five years of age (n = 655), 5.2% (3.6-7.2%) were anemic. Iron deficiency (RRa = 2.19; 95%CI: 1.84-2.60) and consumption of ultra-processed foods (UPF) (RRa = 1.56; 95%CI: 1.14-2.13) were significant contributors to anemia at 1 year, after adjusting for maternal schooling. At 2 years, anemia was significantly associated with maternal anemia at delivery (RRa: 1.67; 95%CI: 1.17-2.39), malaria since birth (2.25; 1.30-3.87), and iron deficiency (2.15; 1.47-3.15), after adjusting for children's age and household wealth index.

CONCLUSIONS

Anemia continues to be highly prevalent during pregnancy and early childhood in the Amazon. Public health policies should address iron deficiency, UPF intake, maternal anemia, and malaria to prevent and treat anemia in Amazonian children.

Anomaly Detection in Endemic Disease Surveillance Data Using Machine Learning Techniques

Disease surveillance is used to monitor ongoing control activities, detect early outbreaks, and inform intervention priorities and policies. However, data from disease surveillance that could be used to support real-time decisionmaking remain largely underutilised. Using the Brazilian Amazon malaria surveillance dataset as a case study, in this paper we explore the potential for unsupervised anomaly detection machine learning techniques to discover signals of epidemiological interest. We found that our models were able to provide an early indication of outbreak onset, outbreak peaks, and change points in the proportion of positive malaria cases. Specifically, the sustained rise in malaria in the Brazilian Amazon in 2016 was flagged by several models. We found that no single model detected all anomalies across all health regions. Because of this, we provide the minimum number of machine learning models top-k models) to maximise the number of anomalies detected across different health regions. We discovered that the top three models that maximise the coverage of the number and types of anomalies detected across the thirteen health regions are principal component analysis, stochastic outlier selection, and the minimum covariance determinant. Anomaly detection is a potentially valuable approach to discovering patterns of epidemiological importance when confronted with a large volume of data across space and time. Our exploratory approach can be replicated for other diseases and locations to inform monitoring, timely interventions, and actions towards the goal of controlling endemic disease.

Notification of malaria cases in the Brazilian Amazon Basin from 2010 to 2020: an analysis of the reporting times

BACKGROUND

As controlling malaria transmission remains a public-health challenge in the Brazilian Amazon basin, the National Surveillance System for Malaria (SIVEP-MALARIA) has registered malaria notifications for over fifteen years helping in the decision-making on control and elimination. As a surveillance database, the system is prone to reporting delays, and knowledge about reporting patterns is essential in decisions.

METHODS

This study contains an analysis of temporal and state trends of reporting times in a total of 1,580,617 individual malaria reports from January 2010 to December 2020, applying procedures for statistical distribution fitting. A nowcasting technique was applied to show an estimation of number of cases using a statistical model of reporting delays.

RESULTS

Reporting delays increased over time for the states of Amazonas, Rondônia, Roraima, and Pará. Amapá has maintained a similar reporting delay pattern, while Acre decreased reporting delay between 2010 and 2020. Predictions were more accurate in states with lower reporting delays. The temporal evolution of reporting delays only showed a decrease in malaria reports in Acre from 2010 to 2020.

CONCLUSION

Malaria notifications may take days or weeks to enter the national surveillance database. The reporting times are likely to impact incidence estimation over periods when data is incomplete, whilst the impact of delays becomes smaller for retrospective analysis. Short-term assessments for the estimation of malaria incidence from the malaria control programme must deal with reporting delays.

Prolonged Breastfeeding and the Risk of Plasmodium vivax Infection and Clinical Malaria in Early Childhood: A Birth Cohort Study

BACKGROUND

Relatively few Amazonian infants have clinical malaria diagnosed, treated and notified before their first birthday, either because they are little exposed to an infection or remain asymptomatic once infected. Here we measure the proportion of children who have experienced Plasmodium vivax infection and malaria by 2 years of age in the main transmission hotspot of Amazonian Brazil.

METHODS

We measured IgG antibodies to 3 blood-stage P. vivax antigens at the 1- and 2-year follow-up assessment of 435 participants in a population-based birth cohort. Children's malaria case notifications were retrieved from the electronic database of the Ministry of Health. We used multiple Poisson regression models to identify predictors of serologically proven P. vivax infection and clinical vivax malaria during the first 2 years of life.

RESULTS

Overall, 23 [5.3%; 95% confidence interval (CI): 3.5-7.8%) children had antibodies to ≥2 antigens detected during at least one follow-up assessment, consistent with past P. vivax infection(s). Fifteen (3.4%; 95% CI: 2.1-5.6%) children had clinical vivax episodes notified during the first 2 years of life; 7 of them were seronegative. We estimate that half of the infections remained unnotified. Children born to women who experienced P. vivax infection during pregnancy were more likely to be infected and develop clinical vivax malaria, while those breast-fed for ≥12 months had their risk of being P. vivax -seropositive (which we take as evidence of blood-stage P. vivax infection during the first 2 years of life) decreased by 79.8% (95% CI: 69.3-86.7%).

CONCLUSION

P. vivax infections in early childhood are underreported in the Amazon, are associated with anemia at 2 years of age, and appear to be partially prevented by prolonged breastfeeding.

On multifactorial drivers for malaria rebound in Brazil: a spatio-temporal analysis

Background

Malaria incidence in Brazil reversed its decreasing trend when cases from recent years, as recent as 2015, exhibited an increase in the Brazilian Amazon basin, the area with the highest transmission of Plasmodium vivax and Plasmodium falciparum. In fact, an increase of more than 20% in the years 2016 and 2017 revealed possible vulnerabilities in the national malaria-control programme.

Methods

Factors potentially associated with this reversal, including migration, economic activities, and deforestation, were studied. Past incidences of malaria cases due to P. vivax and P. falciparum were analysed with a spatio-temporal Bayesian model using more than 5 million individual records of malaria cases from January of 2003 to December of 2018 in the Brazilian Amazon to establish the municipalities with unexpected increases in cases.

Results

Plasmodium vivax incidence surpassed the past trends in Amazonas (AM), Amapá (AP), Acre (AC), Pará (PA), Roraima (RR), and Rondônia (RO), implying a rebound of these states between 2015 and 2018. On the other hand, P. falciparum also surpassed the past trends in AM, AC, AP, and RR with less severity than P. vivax incidence. Outdoor activities, agricultural activities, accumulated deforestation, and travelling might explain the rebound in malaria cases in RR, AM, PA, and RO, mainly in P. vivax cases. These variables, however, did not explain the rebound of either P. vivax and P. falciparum cases in AC and AP states or P. falciparum cases in RR and RO states.

Conclusion

The Amazon basin has experienced an unexpected increase in malaria cases, mainly in P. vivax cases, in some regions of the states of Amazonas, Acre, Pará, Amapá, Roraima, and Rondônia from 2015 to 2018 and agricultural activities, outdoor activities, travelling activities, and accumulated deforestation appear linked to this rebound of cases in particular regions with different impact. This shows the multifactorial effects and the heterogeneity of the Amazon basin, boosting the necessity of focusing the malaria control programme on particular social, economic, and environmental conditions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-021-04037-x.

An analysis of malaria in the Brazilian Legal Amazon using divergent association rules

In data analysis, the mining of frequent patterns plays an important role in the discovery of associations and correlations between data. During this process, it is common to produce thousands of association rules (ARs), making the study of each one arduous. This problem weakens the process of finding useful information. There is a scientific effort to develop approaches capable of filtering interesting patterns, balancing the number of ARs produced with the goal of not being trivial and known by specialists. However, even when such approaches are adopted, the number of produced ARs can still be high. This work contributes by presenting Divergent Association Rules Approach (DARA), a novel approach for obtaining ARs that presents themselves in divergence with the data distribution. DARA is applied right after traditional approaches to filtering interesting patterns. To validate our approach, we studied the dataset related to the occurrence of malaria in the Brazilian Legal Amazon. The discovered patterns highlight that ARs brought relevant insights from the data. This article contributes both in the medical and computer science fields since this novel computational approach enabled new findings regarding malaria in Brazil.