Spatial analysis of malaria incidence and environmental determinants in Hadiya Zone, Ethiopia

Abstract

Malaria remains a leading cause of morbidity and mortality in the developing world, particularly in sub-Saharan Africa. Spatial variability significantly influences efforts to control malaria and its incidence, which remains a serious public health concern in Ethiopia. Using geostatistical methods, this study investigates the environmental factors and spatial distribution of malaria incidence in the Hadiya Zone across woredas in 2022 and 2023. Descriptive analyses revealed consistent spatial heterogeneity, with high incidence rates in Shashogo, Soro, and Misrak Badawacho. Global spatial autocorrelation measures Moran’s I (0.558 in 2022 and 0.483 in 2023; p < 0.01) and Geary’s C (0.63 and 0.69, respectively) confirmed statistically significant clustering of malaria cases. Local Moran’s I analysis identified hot spots in Shashogo, Soro, and Misrak Badawacho, and cold spots in Misha, Duna, and Gombora, indicating localized spatial dependence. Spatial regression analysis, comparing Ordinary Least Squares (OLS) and Spatial Autoregressive (SAR) models, highlighted average maximum temperature (β = 0.945, p = 0.017) and proportion of highland terrain (β = 0.543, p = 0.040) as key predictors of malaria incidence. The SAR model showed superior fit, evidenced by lower AIC and higher log-likelihood values, confirming the influence of spatial dependence. These findings support geographically targeted malaria interventions in high-risk woredas. Limitations include the short study period (2022–2023) and the absence of socioeconomic variables due to lack of household survey and secondary data.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-33236-8.

Introduction

Malaria is a life-threatening disease caused by Plasmodium parasites transmitted through the bites of infected Anopheles mosquitoes. It remains one of the most prevalent tropical diseases worldwide. In 2018, nearly half of the global population was at risk, with sub-Saharan Africa bearing the highest burden. Six countries Nigeria, the Democratic Republic of the Congo, Uganda, Côte d’Ivoire, Mozambique, and Niger accounted for more than half of global malaria cases, and children under five years represented about two-thirds of malaria-related deaths¹. Despite global declines in malaria over the past two decades, recent evidence shows a resurgence in several high-burden countries, driven by climate variability, population displacement, and insecticide resistance.

Ethiopia continues to face substantial malaria challenges, with approximately 75% of the landmass considered malaria-endemic and nearly 70% of the population at risk of infection^2,3. In 2024 alone, over 7.3 million malaria cases and 1,157 deaths were reported, with periodic outbreaks accounting for up to 20% of under-five mortality. Transmission remains highly heterogeneous, shaped by rainfall, temperature, altitude, population movement, and ecological variations⁵. Importantly, 222 high-burden districts (20% of the national total) accounted for more than 75% of malaria cases in 2023, including 50 districts partially inaccessible due to conflict². Recent reports also highlight an expansion of malaria transmission beyond the typical primary (June–September) and secondary (February–May) rainy seasons, complicating control efforts and increasing vulnerability³.

Although national malaria control strategies such as insecticide-treated nets (ITNs), indoor residual spraying (IRS), and improved surveillance have reduced overall morbidity and mortality, geographic variability in transmission remains poorly addressed. Traditional epidemiological approaches often overlook spatial dependence, despite clear evidence that malaria incidence tends to cluster geographically due to shared environmental and demographic conditions⁴. Neglecting these spatial dynamics risks underestimating local hotspots and misdirecting resources.

To address this gap, the present study applies spatial autocorrelation and spatial autoregressive models to malaria incidence data from Hadiya Zone, Central Ethiopia. By integrating meteorological and environmental predictors, we aim to detect clustering patterns of malaria incidence, identify key environmental drivers of transmission, and provide evidence for geographically targeted interventions. To our knowledge, this is the first spatial epidemiological study of malaria in the Hadiya Zone, which has experienced recurrent epidemics and exhibits diverse ecological settings conducive to malaria transmission.

Data and methods

Description of study area and population

The study was conducted in the Hadiya Zone, located in Central Ethiopia, approximately 232 km south of Addis Ababa. The zone consists of several administrative woredas, which are illustrated in Fig. 1 as an administrative boundary map of the study area. The map is intended to show the spatial extent and administrative divisions of the woredas included in the analysis rather than to provide a full cartographic representation. Hadiya Zone is bordered by the Kembata, Gurage, Silte, and Wolayita zones and covers approximately 3,593 km², with elevations ranging from 1,500 to 2,500 m above sea level, placing it within both highland fringe and lowland malaria transmission strata. The climate is subtropical, characterized by bimodal rainfall patterns (June–September and February–April), with annual rainfall ranging from 800 to 1,200 mm and average temperatures between 15 °C and 25 °C^6,7. The zone’s hydrology includes tributaries of the Omo and Bilate Rivers, which create favorable breeding conditions for Anopheles mosquitoes, according to the Ministry of Health (2024). The population is predominantly rural and relies mainly on subsistence agriculture. Combined with ecological diversity, these factors sustain malaria transmission in both highland fringe and lowland areas.

Fig. 1.

Administrative boundary map of the study area (Hadiya Zone, Central Ethiopia) showing the woredas included in the spatial analysis. The map was created by the authors using ArcGIS Pro version 3.1 (Esri, Redlands, CA, USA; https://www.esri.com/arcgis), based on administrative boundary shapefiles obtained from the Ethiopian Central Statistical Agency (CSA). The map is intended to illustrate administrative boundaries only and does not represent a full cartographic map.

Malaria remains a major public health problem in the Hadiya Zone, particularly in lowland and highland fringe areas, with recurrent epidemics causing high morbidity and mortality, especially among children under five and pregnant women. Despite the implementation of control measures such as insecticide-treated nets (ITNs) and indoor residual spraying (IRS), malaria remains endemic, particularly in Misrak Badawacho woreda, where transmission is strongly influenced by climatic and environmental factors. The predominantly rural, agriculture-based livelihood and ecological conditions further contribute to sustained malaria transmission, highlighting the need for locally tailored and geographically targeted interventions⁶.

Source of data and analysis

Study setting and period

This study covered eight woredas in the Hadiyya Zone Lemo, Duna, Anilemo, Misha, Soro, Misrak Badawacho, Shashogo, and Gombora using secondary data for 2022 and 2023.

Health data

Confirmed malaria case counts were obtained from woreda-level health centers in the Hadiyya Zone. Records were screened for completeness, duplicate entries were removed, and missing values were cross-checked against facility logbooks and, where possible, corrected in consultation with woreda health offices.

Meteorological and environmental data

Daily rainfall, minimum temperature, maximum temperature, and humidity were sourced from the Southern Nations, Nationalities, and Peoples’ Region (SNNPR) Meteorological Center (Hawassa). These variables were aggregated to woreda-level averages for the study years. Where direct measurements were unavailable for a woreda, inverse distance weighting (IDW) interpolation from nearby stations was applied to estimate values. Environmental variables proportions of highland, midland, and lowland terrain were provided by the Hadiyya Zone Agricultural Bureau.

Population data

Projected woreda-level populations for 2022 and 2023 were obtained from the Hadiyya Zone Finance Bureau. These projections followed the Ethiopian Central Statistical Agency (CSA) 2017–2037 series using the exponential growth method. No woreda merges or splits occurred during the study period; administrative boundaries remained stable.

Outcome and predictors

The outcome was malaria incidence per 1,000 population in each woreda, calculated as confirmed cases divided by the woreda population and multiplied by 1,000. Predictor variables were average rainfall (RF), average minimum temperature (MIT), average maximum temperature (MAT), and the proportions of highland, midland, and lowland terrain.

Spatial analysis and model selection

Spatial dependence was represented with a row-standardized, first-order queen contiguity spatial weights matrix (8 × 8). Ordinary Least Squares (OLS) models were estimated first; spatial dependence in OLS residuals was assessed using Global Moran’s I and Lagrange Multiplier (LM) diagnostics. The diagnostics indicated significant spatial lag dependence and no evidence of spatial error dependence. Therefore, a Spatial Autoregressive (SAR-lag) model was selected as the preferred specification, capturing spatial spillovers through the weighted incidence of neighboring woredas consistent with ecological and epidemiological processes that influence malaria transmission across adjacent areas.

Software and reproducibility

Spatial processing and mapping were performed in ArcGIS Pro version 3.1 (Esri, USA). Statistical analyses were conducted in R version 4.3.2²⁷ using the packages spdep and spatialreg. Key functions included moran.test() for Global Moran’s I, localmoran() for Local Indicators of Spatial Association (LISA), and lagsarlm() for fitting the SAR-lag model.

Spatial statistical analysis

The nature, concept and definition of Spatial autocorrelation

Spatial analysis is grounded in the principle that “space matters,” meaning that phenomena in one region are often influenced by conditions in neighboring areas¹¹. In this study, we analyze the spatial distribution of malaria incidence across the Hadiya Zone, Ethiopia, to understand how neighboring areas influence local disease patterns.

Spatial autocorrelation, broadly defined, refers to the correlation of a variable with itself across space. Following Anselin and Bera (1998), positive spatial autocorrelation occurs when similar values cluster geographically, while negative spatial autocorrelation arises when dissimilar values are adjacent¹⁰. Measuring spatial autocorrelation helps to reveal the degree to which malaria incidence in one location is related to incidence in nearby locations, reflecting spatial interdependence in disease patterns.

In this study, we quantify spatial dependency using a spatial weights matrix that accounts for adjacency between woredas (districts) within Hadiya Zone. Classical measures, including Moran’s I and Geary’s C, are applied to identify clusters of high or low malaria incidence and to evaluate the strength and significance of spatial relationships⁸. The results guide further spatial modeling, including spatial autoregressive analyses, to assess the influence of environmental and meteorological factors on malaria distribution.

Fitting Spatial autoregressive models for malaria incidence

This concept underlies spatial autocorrelation, which refers to the correlation of a variable with itself across space. According to Anselin and Bera (1998), spatial autocorrelation is positive when similar values cluster and negative when dissimilar values are adjacent⁹.

Spatial autocorrelation captures how values of a variable relate to their spatial arrangement, reflecting the strength and direction of spatial dependence¹⁰. It is typically quantified using a spatial weights matrix, which defines the spatial relationships between observations, such as adjacency or proximity. Classical measures, including Moran’s I and Geary’s C, are used to evaluate the significance of spatial dependence.

When spatial effects are significant, they can be modeled with a spatial lag model:

1

Where Y is the outcome variable (malaria incidence), ρ is the spatial autoregressive coefficient, W is the spatial weights matrix, X represents covariates, β is a vector of regression coefficients, and ϵ is the random error term.

Methods for testing Spatial autocorrelation

To assess spatial dependence in malaria incidence, both global and local measures of spatial autocorrelation were employed. Initially, univariate spatial autocorrelation was diagnosed in the absence of covariates using global and local statistics. Subsequently, a standard regression model was estimated, and diagnostic tests were conducted to determine whether the included covariates adequately captured the spatial dependence in the dependent variable. If significant spatial dependence remained, a spatial autoregressive model, as indicated by the diagnostics, was fitted¹².

Measures of global and local Spatial autocorrelation

Global spatial autocorrelation was evaluated using Moran’s I and Geary’s C statistics. Moran’s I measures overall clustering or dispersion, whereas Geary’s C is more sensitive to local differences. Both statistics require a row-standardized spatial weights matrix (W). To capture spatial heterogeneity and identify specific regions exhibiting high or low spatial autocorrelation, Local Indicators of Spatial Association (LISA) were also computed^12,13.

Results

The distribution pattern of malaria incidence rates and environmental factors was investigated in this study using spatial autocorrelation tests. A spatial autoregressive model was then used to find spatial dependencies and influencing factors. In 2022 and 2023, 9,231 cases of malaria were reported in the study area, resulting in an average malaria incidence rate of 6.44 cases per 1,000 individuals (Tables 1 and 2).

Table 1.

Malaria incidence rate for 2022.

Woreda	Projected population size 2022 (X)	Malaria case in 2022 (Y)	Malaria incidence rate = Y/X*1000
Soro	261,295	593	(593/261,295)*1000 = 2.3
Misrak Badawacho	350,633	1879	(1879/350,633)*1000 = 5.35
Shashogo	143,168	2,083	(2083/143168)*1000 = 14.5
Misha	174,410	127	(127/174,410)*1000 = 0.73
Lemo	117,231	850	(850/117,231)*1000 = 7.3
Gombora	126,202	325	(325/126,202)*1000 = 2.6
Duna	167,270	32	(32/167270)*1000 = 0.19
Analemo	92,371	488	(488/92,371)*1000 = 5.28
Total in Hadiya	1,432,580	6377	(6377/1432580)*1000 = 4.45

Table 2.

Malaria incidence rate for 2023.

Woreda	Projected population size 2023 (X)	Malaria case in 2023 (Y)	Malaria incidence rate = Y/X*1000
Soro	261,295	420	(420/261295)*1000 = 1.61
Misrak Badawacho	350,633	1158	(1158/350,633)*1000 = 3.30
Shashogo	143,168	492	(492/143,168)*1000 = 3.44
Misha	178,742	93	(93/178742)*1000 = 0.52
Lemo	117,231	194	(194/117,231)*1000 = 1.65
Gombora	126,202	160	(160/126,202)*1000 = 1.27
Duna	167,270	194	(194/167270)*1000 = 1.15
Anilemo	92,371	143	(143/92,371)*1000 = 1.55
Total in Hadiya	1,432,912	2854	(2854/1,432,69)*1000 = 1.99

Spatial distribution of malaria

In 2022, Table 1 presents the spatial distribution of malaria incidence rates across the study area. The highest incidence rate was recorded in Shashogo woreda (14.5 per 1,000 population), while the lowest was in Duna woreda (0.19 per 1,000 population). Overall, higher malaria incidence rates were concentrated in the western part of the study area, whereas the eastern part experienced relatively lower rates. The zonal malaria incidence rate for the year was 5 per 1,000 population.

In 2023, as shown in Table 2, the spatial distribution pattern of malaria remained similar, with the highest incidence again observed in Shashogo (3.44 per 1,000 population) and the lowest in Misha (0.52 per 1,000 population). The western part of the zone continued to show higher incidence rates compared to the eastern areas. The overall zonal malaria incidence rate in 2023 decreased to 2 per 1,000 population.

In general, the spatial pattern indicates that malaria incidence decreased from 2022 to 2023, suggesting a downward trend in malaria cases across the study area.

Based on the p-values for both Moran’s I and Geary’s C under normality and randomization assumptions, the null hypothesis of no spatial autocorrelation was rejected. The Z-scores for Moran’s I (6.11 and 2.067) were positive and significant, while those for Geary’s C (–1.93 and − 4.56) were negative and significant, together indicating positive spatial autocorrelation that is, clustering of similar malaria incidence rates among neighboring woredas. The coefficients further support this pattern: Global Moran’s I was 0.558 (normality) and 0.483 (randomization), both significant at p < 0.01, while Global Geary’s C was 0.63 in 2022 and 0.69 in 2023, also significant. Thus, the test statistics in Table 3 confirm that malaria incidence was not randomly distributed but significantly clustered across the study area.

Table 3.

Results of global moran’s I and geary’s C Statistics.

Assumption	Coefficient	observed	Expected	Devstd	Z	Pr > Z
Normality	Moran’s I	0.558	−0.0389	0.100	6.11	< 0.00028
Randomization	Geary’s C	0.00219	1.000	0.468	−1.93	< 0.0004
Normality	Geary’s C	0.0567	1.007	0.1643	−4.56	< 0.0001
Randomization	Moran’s I	0.4833	−0.0383	0.248	2.067	< 0.0006

Note: p-value < 0.05 indicates statistical significance at α = 0.05.

Local moran’s I test for spatial autocorrelation

The Local Moran’s I test (Table 4; Fig. 2) revealed statistically significant malaria incidence clustering in six of the eight Woredas, with high–high clusters in Shashogo, Soro, and Misrak Badawacho and low–low clusters in Misha, Duna, and Gombora, patterns positively associated with average maximum temperature, rainfall, and lowland/midland coverage but negatively associated with minimum temperature and highland area, confirming that malaria incidence in Hadiya Zone exhibits localized spatial dependence rather than uniform distribution.

Table 4.

Results of local moran’s I test.

Id	Woreda	Observed	Expected	Devstd	Z	P
1	Lemo	0.378	0.110	0.034	1.340	0.0011
2	Duna	−0.190	−0.172	0.525	0.456	0.0023
3	Anilemo	0.843	0.733	0.125	4.670	0.750
4	Misha	−0.230	−0.219	0.650	12.02	0.0065
5	Soro	0.346	0.233	0.450	9.03	0.0043
6	Misrak badawacho	0.122	−0.333	0.330	−4.50	0.0020
7	Shashogo	−0.123	−0.230	0.034	0.120	0.0040
8	Gombora	−0.523	−0.433	0.003	0.560	0.760

Fig. 2.

Moran’s I Scatter plot for local spatial autocorrelation of malaria incidence in Hadiya Zone.

Diagnostics for spatial dependence in malaria incidence

The diagnostic results (Table 5) indicated significant spatial lag dependence in malaria incidence across woredas, as both the Lagrange Multiplier (LM) lag and Robust LM (lag) tests were statistically significant (p < 0.05), while the LM error and Robust LM (error) tests were not, suggesting no substantial spatial error dependence. The significant Moran’s I on OLS residuals further confirmed the presence of spatial autocorrelation, violating the independence assumption of Ordinary Least Squares (OLS) regression and justifying the use of a spatial autoregressive (SAR) model. After establishing spatial autocorrelation, LM diagnostics developed by Anselin and Rey (1991) were applied to distinguish between spatial lag and spatial error processes, again showing significance for LM-lag tests (p < 0.001) but not for LM-error tests, indicating the spatial lag model as the appropriate specification. Residual spatial autocorrelation was substantially reduced in the SAR model compared to OLS (Moran’s I decreased from 0.362 to 0.1332), demonstrating improved model fit. Beyond statistical diagnostics, the SAR model is theoretically appropriate because malaria transmission exhibits spatial spillover between neighboring woredas due to shared mosquito habitats, climatic conditions, and human mobility across administrative boundaries^14,15.

Table 5.

Diagnostics for Spatial Dependence.

Test	Moran’s I	Value	prob
Moran’s I (Error)	0.362	2.730	0.000312
Lagrange Multiplier	1	11.034	0.000178
Robust LM (lag)	1	8.110	0.00045
Lagrange Multiplier(error)	1	0.534	0.17260
Robust LM (error)	1	1.230	0.15620

Spatial autoregressive models

Diagnostic tests confirmed the presence of spatial autocorrelation in malaria incidence, indicating that values at one location were correlated with those of nearby areas. This dependence violated the independence assumption of OLS regression, making it unsuitable for analysis. To address this, a spatial lag model was applied using maximum likelihood estimation, with results (Table 6) showing that average rainfall, maximum temperature, and the proportions of highland, midland, and lowland areas were significant positive predictors of malaria incidence, while minimum temperature had a negative but non-significant effect. Model validation was performed using multiple diagnostic approaches. The fit of the SAR model was compared to the OLS model using Akaike Information Criterion (AIC) and log-likelihood values, with SAR showing superior performance. Residual spatial autocorrelation was substantially reduced in the SAR model (Moran’s I = 0.1332) compared to OLS (Moran’s I = 0.362), confirming improved model adequacy. These internal validation steps provided evidence of model robustness, although independent validation with external datasets was not possible due to data limitations.

Table 6.

Results of Spatial lag model estimation.

Variable	Coefficient	Std. Error	T-statistics	probability
W-Malaria	0.3302	0.7532	0.438	0.030
Constant	0.0956	0.3412	0.28	0.0002
Av-Rf	0.5230	0.9810	0.533	0.050
AV-MIT	0.4320	0.7123	0.606	0.050
AV-MAT	0.9450	0.6721	1.406	0.017
Perc-of Hl Area	0.5430	0.8314	0.653	0.040
Perc-of ML area	0.0320	0.6945	0.046	0.960
Perc-of LL Area	0.0254	0.3490	0.073	0.940

Model performance was validated through the Moran’s I test on residuals, which showed reduced spatial autocorrelation from 0.362 in the OLS model to 0.1332 in the spatial lag model (Appendix Table 4), demonstrating an improved fit. Furthermore, the significance of LM-lag and robust LM-lag tests, coupled with the insignificance of LM-error tests, confirmed that the spatial lag model provided the most appropriate framework for capturing the spatial dependence between malaria incidence and environmental factors.

From Table 6, the spatial lag model indicated that average maximum temperature (β = 0.945, p = 0.017), average rainfall (β = 0.523, p = 0.050), and the proportion of highland areas (β = 0.543, p = 0.040) significantly influenced malaria incidence, shaping its spatial distribution across neighboring woredas. In contrast, the proportions of midland (β = 0.032, p = 0.960) and lowland areas (β = 0.025, p = 0.940) were not statistically significant. Among the significant predictors, average maximum temperature exerted the strongest positive effect, indicating that a 1 °C increase was associated with nearly one additional malaria case per 1,000 population, highlighting its critical role in promoting mosquito emergence and survival, particularly when combined with rainfall. The positive effect of highland areas further emphasizes how environmental heterogeneity, coupled with spatial spillover captured through the standardized spatial weights matrix, drives localized malaria clustering.

Discussion

This study revealed significant spatial clustering of malaria incidence in the Hadiya Zone, using both global and local spatial autocorrelation techniques. The Global Moran’s I and Geary’s C statistics confirmed strong positive spatial autocorrelation, indicating that malaria incidence rates are not randomly distributed but tend to cluster geographically.

The spatial patterns revealed that malaria incidence was consistently higher in the western part of the zone, particularly in woredas like Shashogo and Soro, across both 2022 and 2023. This finding aligns with previous studies in Ethiopia and other Sub-Saharan African contexts that have identified spatial heterogeneity in malaria distribution due to varying environmental and topographical conditions¹⁶.

The spatial lag model, validated by Lagrange Multiplier (LM) diagnostics, was the most appropriate approach due to spatial dependence in the data. The model demonstrated that the land classification (highland, midland, and lowland), average rainfall, and average maximum temperature all had a significant impact on the incidence of malaria. Of these, the average maximum temperature had the most beneficial impact, which is consistent with recent studies showing that higher temperatures are a major factor in mosquito development and disease transmission^17,20. Conversely, the average minimum temperature displayed a negative and statistically insignificant correlation, which might be due to lower nighttime temperatures in highland regions that are less favorable for mosquito breeding²².

Additionally, the percentage of lowland and midland regions showed robust positive correlations with the incidence of malaria, confirming the importance of topography and altitude in malaria risk. Mosquito breeding is facilitated by the stagnant water sources and generally warmer climates found in lower elevation zones^21,22. These results demonstrate how crucial it is for malaria control initiatives to take environmental heterogeneity into account. The selection of the SAR model was consistent with both statistical diagnostics and epidemiological reasoning. Statistically, spatial lag dependence dominated, while spatial error dependence was negligible, confirming SAR as the most appropriate specification. Epidemiologically, malaria transmission is influenced by ecological spillover across administrative boundaries, as vectors and human mobility do not respect woreda borders. Therefore, SAR better captures the diffusion of malaria risk between adjacent areas compared to traditional OLS models.

In comparison with previous spatial studies such as that by Oluyemi, A¹⁷., who used Moran’s I and Gi* in East Wollega, and Yeshiwondim et al. (2009), who analyzed village-level spatial patterns of malaria in Ethiopia the current study reaffirms the value of spatial statistical tools in detecting disease clusters and understanding their drivers. However, unlike Yeshiwondim’s findings, which emphasized demographic variables such as age and gender, this study focused on geospatial and environmental variables, revealing different but complementary perceptions into malaria epidemiology^18,19.

In addition to environmental determinants, socioeconomic factors are widely recognized as critical drivers of malaria transmission. Housing conditions influence mosquito entry and resting behavior, with improved housing reducing malaria risk in sub-Saharan Africa²⁵. Inequities in access to healthcare and malaria interventions contribute to spatial heterogeneity in disease burden²⁴. Although our study did not incorporate socioeconomic variables due to data limitations, integrating these factors alongside environmental predictors would provide a more comprehensive understanding of malaria risk in the Hadiya Zone²⁵.

Furthermore, the consistency of the malaria hotspots found in Shashogo and Soro woredas over two years points to enduring environmental or infrastructure elements that support long-term transmission. Irrigation techniques, stagnant water sources that are not addressed by seasonal interventions, or restricted access to healthcare could all have an impact on this persistence. In southern Ethiopia and portions of Uganda, where socio-ecological factors such as agricultural growth and insufficient vector control coverage sustain local transmission, similar persistent clusters have been documented^23,24. These spatially stable hotspots emphasize the need for year-round, geographically focused malaria interventions as opposed to general, uniform strategies.

Conclusion

This study aimed to identify the spatial distribution and environmental determinants of malaria incidence in the Hadiya Zone of Central Ethiopia. The results demonstrated that malaria incidence was not randomly distributed but significantly clustered, with persistent hotspots in western woredas such as Shashogo, Soro, and Misrak Badawacho. Spatial dependence, confirmed through Moran’s I and spatial autoregressive modeling, indicated that malaria transmission in one woreda is influenced by conditions in neighboring areas. Among the environmental predictors, average maximum temperature, rainfall (marginally significant), and the proportion of highland terrain were identified as significant drivers of malaria incidence.

The findings suggest that malaria interventions should move beyond uniform strategies and instead adopt geographically targeted approaches, prioritizing high-incidence woredas for enhanced surveillance and control. Incorporating geospatial tools and climate-sensitive early warning systems can improve outbreak detection and response, while inter-woreda coordination may optimize resource use. These strategies are particularly relevant in ecologically diverse zones that sustain malaria transmission. However, the study relied on secondary health records and environmental data, without incorporating socioeconomic, behavioral, or entomological factors that also influence malaria dynamics. In addition, independent validation with external datasets was not performed, which may limit the generalizability of the results. Future research should integrate these dimensions and apply cross-validation or external datasets to strengthen the robustness and applicability of spatial regression models in malaria epidemiology.

Limitations

This study did not include socioeconomic variables such as housing conditions, access to health facilities, or human mobility, which are critical determinants of malaria risk, due to the lack of reliable woreda-level data. Population projections were used to account for demographic variation, but future research should integrate both environmental and socioeconomic drivers for a more comprehensive understanding of malaria transmission in the Hadiya Zone. In addition, the analysis was limited to two years (2022–2023) because only these datasets provided complete and high-quality malaria and meteorological information. While these enabled detection of spatial dependence, it does not fully reflect interannual variability in incidence and climate factors. Future studies should therefore incorporate 5–10 years of data to improve temporal representativeness and support the design of spatiotemporal early warning systems.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

SAR

Spatial autoregressive model

OLS

Ordinary least squares

AIC

Akaike information criterion

LM

Lagrange multiplier

SNNPR

Southern nations, nationalities, and peoples’ region

WHO

World health organization

GIS

Geographic information system

ARC GIS

A specific GIS software platform

ITN

Insecticide-treated nets

IRS

Indoor residual spraying

RF

Rainfall

MIT

Minimum temperature

MAT

Maximum temperature

LL

Log-likelihood

GMI

Global Moran’s I

LCI

Local Moran’s I

Geary’s C

Geary’s contiguity ratio

API

Annual parasite incidence

MoH

Ministry of health

Author contributions

S.S.A. conceived and designed the study, acquired and analyzed the data, interpreted the results, prepared all figures and tables, and wrote the entire manuscript. S.S.A. also reviewed and approved the final version of the manuscript.

Funding

This research was conducted without any external funding support.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available and must be obtained from the corresponding author upon reasonable request. The data were sourced from the Hadiyya Zone Health Bureau, the SNNPR Meteorological Center, and the Hadiyya Zone Agricultural and Finance Bureaus, and access requires official permission from the respective agencies.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

Ethical clearance for this study was obtained from the Institutional Review Board (IRB) of Wachemo University (WCU 17.2023). The study was based on secondary data without personal identifiers. All methods were performed in accordance with the relevant guidelines and regulations of Wachemo University and national research ethics standards. All necessary permissions were secured from relevant health and administrative offices in the Hadiyya Zone to ensure the ethical use of the data.

Informed consent

As the study utilized secondary, de-identified data and did not involve direct interaction with human participants, informed consent was not applicable. However, permission to use the data was formally obtained from the appropriate authorities.