The temporal lagged association between meteorological factors and malaria in 30 counties in south-west China: a multilevel distributed lag non-linear analysis

Xing Zhao, Fei Chen, Zijian Feng, Xiaosong Li, Xiao‐Hua ZhouMalaria 2014 · DOI

TL;DR

This study investigates the complex, time-delayed relationship between weather patterns and malaria incidence in southwest China. By employing a novel statistical model, the researchers reveal that different meteorological factors (rainfall, humidity, temperature) have distinct lagged effects on malaria, and these effects are significantly influenced by local climate conditions (hot vs. cold weather). This nuanced understanding is crucial for improving malaria prediction and control strategies.

What Problem Does This Paper Address?

Malaria transmission is known to be influenced by meteorological factors, but the exact nature of this relationship is complex and often debated. Previous studies have yielded inconsistent findings, largely because they often fail to account for two critical aspects: the lagged effect (weather today doesn't cause malaria today, but days or weeks later) and the non-linear relationship (doubling rainfall doesn't necessarily double malaria risk). This lack of a clear, consistent understanding hinders the development of effective malaria early warning systems, especially in regions like southwest China, which has historically been a high-incidence area. The paper aims to untangle these complexities by examining the temporal, lagged, and non-linear associations between weekly weather data and malaria cases across diverse climatic conditions within southwest China.

Methods

The core of this research lies in the application of a Multilevel Distributed Lag Non-linear Model (MDLNM). This advanced statistical framework is designed to simultaneously capture both the non-linear relationships and the delayed effects of predictor variables (meteorological factors) on an outcome variable (malaria cases).

Here's a breakdown of the key methodological aspects:

  • Distributed Lag: Instead of assuming a single fixed lag period for a meteorological factor, the MDLNM allows for a flexible distribution of effects across a range of time lags. This acknowledges that the impact of weather can unfold over several weeks.
  • Non-linear: The model can capture relationships where the effect of a meteorological factor isn't simply proportional. For instance, the impact of a slight temperature increase might be different from a large one.
  • Multilevel: The model accounts for the hierarchical structure of the data, where observations are nested within counties. This allows for the estimation of both common effects across all counties and county-specific random intercepts, acknowledging that each county might have unique baseline malaria rates due to unmeasured factors.
  • Basis Functions: The MDLNM uses basis functions (specifically, 3rd-order polynomials in this study) to model the non-linear and lag effects flexibly.
  • Stratification: To account for regional climate differences, the 30 study counties were divided into two groups: "hot weather" and "cold weather" based on minimum temperature. This allows for a direct comparison of how weather impacts malaria under different climatic regimes.

This approach moves beyond simpler time-series analyses that might assume a single lag or a linear relationship, offering a more biologically plausible and statistically robust way to understand the complex interplay between weather and disease.

Datasets

The study utilized the following datasets:

  • Malaria Cases: Weekly malaria case data for 30 counties in southwest China, collected from 2004 to 2009. This data was obtained from the Chinese Centre for Disease Control and Prevention (CCDC).
  • Meteorological Data: Weekly data for four key meteorological variables:
  • Rainfall (mm)

    • Mean Relative Humidity (%)
    • Mean Minimum Temperature (°C)
    • Mean Maximum Temperature (°C) This data was collected from the Chinese Meteorological Data Sharing Service System, with monitoring stations located in or near the selected counties.
  • Population Data: Annual population data for each county from 2004 to 2009, obtained from the National Bureau of Statistics of China. This was used to adjust for population size in the models.

The study selected the top 30 counties based on average annual malaria incidence that also had available meteorological data, ensuring a focus on high-risk areas.

Key Insights and Ideas

  • Lagged Effects are Crucial: The study strongly reinforces that meteorological factors don't have an immediate impact on malaria. The "lagged" effect, meaning the delay between weather exposure and disease occurrence, is a critical determinant of the association.
  • Climate Matters: The distinction between "hot" and "cold" weather counties is not just an academic exercise; it fundamentally alters the relationship between weather and malaria. This highlights the need for localized models rather than one-size-fits-all approaches.
  • Rainfall's Extended Influence: Rainfall showed a relatively long lag range, suggesting it plays a significant role in creating breeding grounds for mosquitoes and facilitating the entire transmission cycle. This effect was more pronounced in hotter climates.
  • Humidity's Dual Role: Relative humidity showed a significant association with malaria in hot weather counties, particularly at earlier and later lags, suggesting its importance in parasite maturation and mosquito survival.
  • Temperature's Nuanced Impact: Minimum temperature had a longer lag and stronger correlation in hot weather counties, while maximum temperature's effect was more immediate (early lags), potentially influencing parasite maturation.
  • Explaining Inconsistent Findings: The study provides a compelling explanation for why previous studies have reported conflicting results: either they assumed a single, fixed lag period (which is biologically unrealistic) or they didn't account for the distinct climatic conditions of their study sites.

Key Results

  • Rainfall: Associated with malaria in both hot and cold weather counties with a lagged correlation. The lag range was longer in hot weather counties.
  • Relative Humidity: Correlated with malaria cases at early and late lags in hot weather counties, but showed a less significant association in cold weather counties.
  • Minimum Temperature: Exhibited a longer lag range and larger correlation coefficients in hot weather counties compared to cold weather counties.
  • Maximum Temperature: Was associated with malaria cases primarily at early lags.
  • Lag Range Differences: The lag range for meteorological factors was generally longer in hot weather counties compared to cold weather counties.
  • Magnitude of Effect: While the lag patterns varied, the magnitude of the effect for rainfall showed more variation across different rainfall levels compared to temperature and humidity.

Limitations and Future Work

The authors acknowledge several limitations:

  • Confounding Factors: As an observational study, it's possible that unmeasured confounders (e.g., differences in malaria prevention measures, mosquito control efforts, or human behavior) could influence the results. The use of county-specific random effects helps, but may not fully eliminate bias.
  • Data Quality: While efforts were made to ensure data quality, changes in data collection and reporting over the six-year period could introduce some variability.
  • Pre-defined Lag Ranges: The study used pre-defined lag ranges based on existing literature. While these were informed by biological reasoning, exploring a wider range of lags could potentially reveal further insights.
  • Parasite Species: The study did not differentiate between Plasmodium vivax and Plasmodium falciparum malaria, which have different incubation periods. Separate analyses for each species could provide more granular understanding.

Future work could focus on:

  • Epidemiological Research: Further research is warranted to explore the potential differences in lag patterns between parasite species.
  • Predictive Modeling: The findings suggest that traditional early warning systems (e.g., moving averages) need to be modified to incorporate the complex, lagged, and non-linear relationships identified. The paper poses the question of how to best integrate these lag patterns into predictive models.
  • Broader Geographic Scope: Expanding this analysis to other regions with different climatic conditions could enhance the generalizability of the findings.

Summary updated

Abstract

Using weekly malaria cases and meteorological information, this work studied the temporal lagged association pattern between malaria cases and meteorological information in south-west China. The results suggest that different meteorological factors show distinct patterns and magnitudes for the lagged correlation, and the patterns will depend on the climatic condition. Existing inconsistent findings for climatic factors' lags could be due to either the invalid assumption of a single fixed lag or the distinct temperature conditions from different study sites. The lag pattern for meteorological factors should be considered in the development of malaria early warning system.
BibTeX
@inproceedings{zhao2014temporal,
  author = {Xing Zhao and Fei Chen and Zijian Feng and Xiaosong Li and Xiao‐Hua Zhou},
  title = {{The temporal lagged association between meteorological factors and malaria in 30 counties in south-west China: a multilevel distributed lag non-linear analysis}},
  booktitle = {Malaria},
  year = {2014},
  doi = {10.1186/1475-2875-13-57},
}
Figure