Estimating probabilities of malaria importation in southern Mozambique through P. falciparum genomics and mobility patterns

By integrating human mobility data with *P. falciparum* genomics, this study developed a novel Bayesian approach to quantify malaria importation in southern Mozambique, revealing that nearly half of infections in targeted elimination districts are imported—primarily from Inhambane province—and highlighting the need for fine-scale, source-targeted strategies to achieve malaria elimination.

The Gist

Imagine malaria as a stubborn weed in a garden. For years, health workers in southern Mozambique have been pulling these weeds out, trying to clear the garden completely. But there's a problem: even when they pull all the weeds in their own patch, new seeds keep blowing in from neighboring fields. These "imported" cases are the biggest obstacle to getting rid of malaria entirely.

This paper is like a high-tech detective story that combines travel diaries and parasite DNA to figure out exactly where these new malaria seeds are coming from.

Here is the story of how they solved the mystery, explained simply:

1. The Mystery: Two Neighbors, Two Different Problems

The researchers focused on two neighboring districts in southern Mozambique: Magude and Matutuine.

• Magude is like a quiet, rural village tucked away in the hills, bordered by a massive national park. It's hard to get in and out of.
• Matutuine is like a busy coastal town near the capital city, with good roads and lots of tourists.

Even though they are neighbors, the malaria situation was very different. The researchers wanted to know: Why is malaria still showing up in Matutuine but barely in Magude? Are the cases happening locally, or are people bringing them in from elsewhere?

2. The Detective Tools: The "Genetic Fingerprint" and the "Travel Log"

To solve this, the team used two main clues:

• The Travel Log: They asked patients, "Where did you go in the last month?" This is like checking a passenger's boarding pass.
• The Genetic Fingerprint: They took a tiny sample of the malaria parasite from the patient's blood and sequenced its DNA. Think of this as taking a photo of the parasite's face to see who its "family" is.

They built a special computer model (a "Bayesian approach") that acted like a super-smart judge. It weighed the travel log against the genetic fingerprint to decide: Is this parasite a local resident, or did it just arrive on a bus from another province?

3. The Big Discovery: The "Inhalambane" Connection

The results were fascinating:

• The "Local" vs. "Imported" Split: In the quiet district of Magude, almost all cases were local (grown right there). But in the busy district of Matutuine, nearly half of all malaria cases were imported.
• The Source: Where were these imported cases coming from? The genetic fingerprints pointed to one main culprit: Inhambane province. It's like finding out that 60% of the weeds in Matutuine's garden were actually seeds blown in from Inhambane.
• The Travel Link: People in Matutuine were traveling to Inhambane much more often than people in Magude. The genetic data confirmed that the parasites in Matutuine were "cousins" to the parasites in Inhambane.

4. Why This Matters: The "Seed" Analogy

Imagine you are trying to keep a house free of mice.

• Old Strategy: You set traps inside the house. If you catch a mouse, you remove it. But if a mouse keeps sneaking in through the front door, you'll never win.
• New Strategy (This Paper): This study tells us exactly which door the mice are using. It says, "Hey, 60% of the mice in this house are coming from the neighbor's garden (Inhambane)."

This changes the game. Instead of just setting more traps inside the house (treating local patients), health workers can now:

1. Treat the Source: Go to Inhambane and clear the weeds there so fewer seeds blow over.
2. Check the Travelers: Screen people coming back from Inhambane before they enter Matutuine.
3. Save Money: In Magude, where people don't travel much, they can focus on local traps. In Matutuine, they need to focus on the border and the travelers.

5. The Takeaway

This paper is a game-changer because it proves that genetics + travel data = a superpower for malaria elimination.

It shows that in the "last mile" of getting rid of malaria, you can't just look at one village. You have to look at the whole neighborhood and the roads connecting them. By understanding that Matutuine's malaria problem is actually a "travel problem" linked to Inhambane, health officials can now design a smarter, cheaper, and more effective plan to finally clear the garden of malaria weeds for good.

Technical Summary

1. Problem Statement

Malaria elimination in low-transmission settings is critically hindered by parasite importation, where infected individuals travel from high-transmission areas to low-transmission zones, sustaining local transmission chains. Current strategies often rely on:

• Epidemiological data: Travel reports (often incomplete or biased).
• Genomic data: Assessing genetic relatedness to infer transmission flow.
• Mobility data: Mobile phone records or geospatial modeling.

The Gap: No existing method simultaneously integrates epidemiological, human mobility, and parasite genomic data to provide individual-level probabilities of whether a specific malaria case is imported or locally acquired. Previous studies using these data sources separately often yielded conflicting conclusions due to intrinsic biases in each data type.

2. Methodology

The authors developed a novel Bayesian framework to classify clinical cases as "imported" or "local" by integrating three distinct data streams:

A. Data Collection

• Study Area: Southern Mozambique (Maputo Province), specifically the low-transmission districts of Magude (rural, bordering Kruger National Park) and Matutuine (urban-adjacent, high connectivity).
• Samples:
  • Targeted: 200 P. falciparum positive clinical cases from Magude and Matutuine (2022) with complete travel history and sequencing coverage.
  • Surveillance: 1,467 samples total from 9 provinces across Mozambique to establish national genetic baselines.
• Genomics: Used the MAD4HatTeR targeted amplicon sequencing panel (165 microhaplotype loci) to determine genetic diversity and Identity-by-Descent (IBD).
• Mobility: Collected detailed travel reports (destination, duration, dates) for the previous 28 days.

B. The Bayesian Model

The model calculates the probability that an infection occurred in area $A$ (travel destination) versus area $B$ (local residence) given the parasite genome $G$:

$$P(I_A | G) = \frac{P(I_A) \cdot P(G | I_A)}{P(G)}$$

Where:

• $P(I_A)$ (Epidemiological/Mobility Prior): Proportional to the time spent in area $A$ ($T_A$) and the transmission intensity proxy ($PR_A$, derived from mRDT positivity rates in children).
• $P(G | I_A)$ (Genetic Likelihood): Estimated as $R'_A$, the fraction of samples from area $A$ that are IBD-related (IBD > 0.1, $p < 0.05$) to the study sample.
• Classification: Cases with an importation probability $> 50\%$ were classified as imported.

C. Statistical Analysis

• Genetic Structure: Assessed spatial genetic relatedness ($R$) across provinces and regions (South, Centre, North).
• Risk Factors: Used Firth's logistic regression (to handle small sample sizes and separation issues) to identify factors associated with importation (e.g., district, season, occupation, travel destination).
• Genetic Metrics: Calculated Complexity of Infection (COI) and effective COI (eCOI) using the MOIRE software.

3. Key Results

A. Genetic Population Structure

• Regional Differentiation: A strong genetic divide exists between Southern Mozambique and the Central/North regions.
  • Relatedness ($R$) within Centre/North: 0.028.
  • Relatedness between South and Centre/North: 0.017 (significantly lower, $p < 0.001$).
• Spatial Scale:
  • Large Scale (>100 km): Strong isolation-by-distance; genetic relatedness decreases as geographical distance increases.
  • Medium Scale (10–100 km): No significant correlation, suggesting high mobility or heterogeneous transmission within provinces.
  • Micro Scale (<10 km): Significant correlation re-emerges, with the highest relatedness found within the same household, indicating fine-scale transmission networks.

B. Importation Rates

• Overall Importation: 43.5% (87/200) of infections in the elimination-targeted districts were classified as imported.
• District Disparity:
  • Matutuine: High importation rate (48.6%, 84/173).
  • Magude: Low importation rate (11.1%, 3/27).
  • Statistical Significance: Matutuine had 6.6 times higher odds of a case being imported compared to Magude ($OR=6.6$, $p < 0.001$).
• Source of Importation: The primary source for Matutuine was Inhambane Province (62% of imported cases), followed by Gaza and Zambézia.

C. Risk Factors and Correlates

• Travel Destination: Traveling to provinces outside Maputo (specifically Inhambane, Gaza, Zambézia, Nampula) significantly increased the odds of importation ($>80\%$ probability).
• Seasonality: Importation rates were higher during the dry season (54.7%) compared to the rainy season (38.2%), likely because local transmission drops while travel continues.
• Genetic Complexity: Imported cases had significantly higher eCOI (1.80 vs 1.53) and COI (2.80 vs 2.21) than local cases, confirming they originated from areas with higher transmission intensity.
• Demographics: No significant association was found with age, sex, or pregnancy.

4. Key Contributions

1. Novel Methodology: Introduced a Bayesian approach that fuses travel history, epidemiological intensity proxies, and parasite genomics to generate individual importation probabilities. This overcomes the limitations of using any single data source.
2. Fine-Scale Resolution: Demonstrated that while genomics can distinguish inter-provincial transmission, it also reveals household-level transmission clusters, validating its utility for both macro and micro surveillance.
3. Quantification of Heterogeneity: Revealed that even within the same province (Maputo), neighboring districts (Magude vs. Matutuine) have vastly different importation dynamics, necessitating district-specific elimination strategies.
4. Source Identification: Pinpointed Inhambane Province as the critical reservoir sustaining transmission in Matutuine, a finding that would be difficult to ascertain with travel data alone due to reporting biases.

5. Significance and Implications

• Tailored Elimination Strategies: The study argues against "one-size-fits-all" interventions.
  • Matutuine: Requires importation-focused interventions, such as testing/treating travelers from Inhambane and reducing transmission in Inhambane itself.
  • Magude: With low importation, strategies should focus on reactive case detection and local vector control rather than travel screening.
• Cost-Effectiveness: Interventions targeting travelers and source provinces are most critical during the dry season when importation drives the majority of cases.
• Scalability: The framework is adaptable to other P. falciparum endemic regions and potentially P. vivax (with adjusted travel windows), offering a robust tool for the "last mile" of malaria elimination.
• Policy Impact: Provides evidence for the Mozambique National Malaria Control Program to shift resources toward cross-provincial collaboration and targeted traveler screening.

Limitations Noted

• Sampling Bias: Samples from non-study districts were collected via convenience sampling at health facilities, potentially underestimating importation from those areas.
• Travel Reporting: Reliance on self-reported travel (28-day window) may miss unreported trips or longer incubation periods.
• Transmission Proxy: Used mRDT positivity in children as a proxy for transmission intensity; monthly incidence data would be more precise.

In conclusion, this study demonstrates that combining genomic surveillance with mobility data provides a powerful, high-resolution lens for understanding malaria dynamics, enabling precise, data-driven interventions to achieve elimination in southern Mozambique.