A thirteen-year spatiotemporal analysis of malaria surveillance across Assam has revealed that while overall incidence substantially declined between 2011 and 2023, transmission transitioned into a highly focal pattern characterized by persistent geographic and demographic concentrations. Researchers analyzed district-level annual case records compiled from the National Center for Vector Borne Diseases Control alongside demographic projections from the International Institute for Population Sciences. The investigation tracked shifting disease dynamics across three distinct developmental groups: children aged zero to four years, school-aged children aged five to fourteen years, and adults older than fifteen years. The findings established that the statewide decline was accompanied by persistent transmission foci in forested border districts and central hill tracts, with the highest relative incidence rates observed among school-aged individuals and the highest total case volume concentrated among adult males.

Geographically, Assam covers 78,438 square kilometers in northeastern India, representing approximately 2.39 percent of the country’s total land area, with a projected population of roughly 36.16 million distributed across thirty-five administrative districts. Most residents, approximately 84.2 percent, live in rural areas characterized by an average population density of 398 people per square kilometer. The physical terrain is defined by the fertile and flood-prone valleys of the Brahmaputra and Barak rivers, coupled with a warm, humid climate experiencing more than 2,000 millimeters of annual monsoon rainfall. These recurring floods and heavy rains routinely create extensive, standing water bodies that serve as favorable vector breeding grounds. Compounding these environmental factors are historical burdens of Plasmodium falciparum malaria, complex land-use patterns, insecticidal and drug resistance, and remote rural healthcare access barriers that complicate uniform disease suppression across the region.

To evaluate these disparate environmental and demographic risks across a uniform geographic framework, the investigators re-aggregated all case records from newly established administrative territories into a fixed baseline of twenty-seven parent districts aligned with the 2011 Indian Census structure. This alignment ensured spatial continuity across the entire 2011 to 2023 observational window without distortion from administrative boundary shifts. Expected case numbers for every district and calendar year were computed using internal standardization anchored to a pooled thirteen-year reference rate across the entire study period. By utilizing stratum-specific internal denominators, the researchers estimated standardized incidence ratios that permitted precise tracking of district-level fluctuations over time, while avoiding invalid direct numerical cross-comparisons between divergent demographic groups that possess distinct underlying baseline distributions.

Methodologically, the team employed a complementary two-stage statistical strategy incorporating both frequentist spatial cluster detection and Bayesian hierarchical modeling. In the initial phase, space-time scan statistics implemented through SaTScan version 10.1 used cylindrical scanning windows governed by discrete Poisson models to screen for localized clusters exhibiting higher-than-expected case counts. Significant transmission clusters were identified using 999 Monte Carlo replications evaluated at a significance threshold of p less than 0.05. In the subsequent phase, Bayesian hierarchical spatio-temporal models were implemented via the Integrated Nested Laplace Approximation approach within R-INLA. The spatial neighborhood dependency was structured using a queen contiguity matrix and a Besag-York-Mollié prior that divided spatial variation into a spatially structured conditional autoregressive component and an unstructured independent heterogeneity term.

Temporal trends were examined by comparing first-order and second-order Bayesian random walk priors across two distinct observational likelihood formulations: Negative Binomial and Zero-Inflated Poisson distributions. Model selection criteria utilizing the Deviance Information Criterion and the Watanabe-Akaike Information Criterion demonstrated that Zero-Inflated Poisson models decisively outperformed Negative Binomial specifications by hundreds of points across all demographic strata. For example, among males aged five to fourteen years, the Zero-Inflated Poisson model yielded a Watanabe-Akaike score of 1547.40 compared to 1915.04 for the standard Negative Binomial model. This model performance showed that once excess zero counts arising from localized elimination were captured through the zero-inflation parameter, active case counts followed standard Poisson variance. Furthermore, second-order random walks offered superior fit for younger cohorts, reflecting smooth programmatic declines, whereas first-order walks better suited adult cohorts whose occupational exposures in agriculture, tea plantations, and forestry prompted independent annual fluctuations.

Over the full study duration, a total of 147,007 malaria cases were documented across the state, exhibiting substantial variations across age cohorts. The adult demographic older than fifteen years recorded the largest absolute burden, contributing 90,801 cases, or 61.8 percent of the statewide total, with males accounting for 54,516 cases compared to 36,285 cases in females. School-aged children aged five to fourteen contributed 41,546 cases, representing 28.7 percent of the state total, yet exhibited the highest overall incidence rate at 0.45 per 1,000 person-years, compared to 0.34 per 1,000 person-years in children aged zero to four who accumulated 14,660 cases, and 0.30 per 1,000 person-years in adults. In both the zero to four and five to fourteen age cohorts, recorded cases were higher among females than males, contrasting sharply with the male-dominated case volumes observed in the older adult cohort.

Geographic disparities were pronounced throughout the thirteen-year span, with the central hill district of Karbi Anglong consistently displaying the highest caseload across all demographic categories. Karbi Anglong accounted for 5,152 cases or 35 percent of all infections in children aged zero to four, 11,760 cases in school-aged children, and 21,884 cases or 24 percent of the total adult burden. In sharp contrast, several districts showed minimal or absent transmission; for example, Darrang reported zero cases among children aged zero to four, Dibrugarh recorded only 291 adult cases over the entire thirteen years, and districts such as Sivasagar and Nalbari maintained similarly negligible levels. Preliminary space-time scan analysis revealed intense early clustering between 2011 and 2015, notably in Karbi Anglong where observed-to-expected ratios reached 21.10 for school-aged males and 20.24 for school-aged females, alongside an observed-to-expected ratio of 47.79 among young males in Chirang between 2011 and 2012.

The Bayesian posterior relative risk mappings demonstrated that transmission foci shifted dynamically from central highland tracts toward the northwestern districts bordering Bhutan. During the early study years from 2011 to 2015, elevated relative risks were heavily concentrated in Chirang, Udalguri, and Karbi Anglong. In 2018, Udalguri experienced acute epidemic surges, with posterior relative risk climbing to 32.64 in young females and 33.99 in young males, alongside values of 24.02 in school-aged females and 22.67 in school-aged males. By 2023, standardized relative risk values had fallen below 1.0 across the vast majority of districts following widespread vector-control scale-ups, but Kokrajhar emerged as a dominant remnant hotspot, recording relative risks of 20.41 in young females, 18.59 in young males, 20.14 in school-aged females, 20.76 in school-aged males, and 21.25 in adult females.

Temporal random walk evaluations identified a rapid programmatic reduction in relative risk between 2011 and 2018 across all six demographic cohorts, coinciding with the expanded distribution of rapid diagnostic tests, long-lasting insecticidal nets, and artemisinin-based combination therapies. Following 2018, this downward trajectory stabilized into a plateau from 2020 through 2023. Although modest upward inflections of approximately 0.2 relative risk units appeared in adult cohorts and young males during the final three years, the authors noted these changes were statistically non-significant because the corresponding 95 percent Bayesian credible intervals widened substantially and spanned the study-wide baseline of 1.0. This widening reflected boundary behavior inherent to random walk priors at the terminus of time series coupled with sparse counts in near-elimination districts, rather than demonstrable epidemic resurgences, though disruptions from the COVID-19 pandemic may have influenced routine local surveillance operations.

The authors outlined several methodological limitations inherent to the surveillance framework. Case data originated from passive surveillance through the national control program, making them subject to potential underreporting and detection bias in inaccessible rural and forest communities. In addition, diagnostic practices evolved over the study duration as rapid diagnostic tests were deployed alongside conventional microscopy, potentially influencing detection sensitivity over time. Because the investigation relied on aggregated district-level counts rather than individual health records, the associations remain subject to the ecological fallacy and could not capture personal behavioral risk factors such as bed net usage, housing structure, local migration, or precise occupational exposures. Nevertheless, the convergence of Bayesian smoothing and scan statistics demonstrates that malaria in Assam has contracted into persistent, focal reservoirs that require targeted surveillance rather than uniform statewide interventions.

Subject of Research: Spatiotemporal analysis of age- and sex-specific malaria dynamics in Assam from 2011 to 2023

Article Title: Age and sex specific spatiotemporal dynamics of malaria in Assam from 2011 to 2023

Article References: Vavilala, H., Mopuri, R., Kadiri, M. R., Panda, L. P., & Mutheneni, S. R. (2026). Age and sex specific spatiotemporal dynamics of malaria in Assam from 2011 to 2023. Discover Public Health, 23(1), Article 1763. https://doi.org/10.1186/s12982-026-02838-6

Image Credits: AI Generated

DOI: 10.1186/s12982-026-02838-6

Keywords: Malaria, Assam, Spatial Epidemiology, Public Health, Infectious Diseases, Disease Surveillance, Bayesian Modeling, Vector-Borne Diseases, specific, spatiotemporal, dynamics, scientific research