Seasonal and meteorological determinants of district level acute watery diarrhea incidence in Fiji, 2016–2023: a time-series analysis

Abstract

Background

Pacific island countries (PICs), including Fiji, are among the most vulnerable to the health impacts of climate change. In Fiji, acute watery diarrhea (AWD) remains a common communicable disease of public health concern, yet evidence regarding direct meteorological drivers remains equivocal, especially in Pacific settings. This study examined the associations between rainfall, temperature, and humidity and AWD incidence across districts, while also assessing seasonal and autoregressive influences.

Methods

An ecological time-series analysis was conducted using eight years (2016–2023) of surveillance data from the Early Warning Alert and Response System (EWARS). Meteorological predictors included monthly averages of rainfall, temperature, and humidity. Negative binomial mixed-effects regression models were used to examine associations between AWD incidence and meteorological variables, adjusting for seasonal indicators, autoregressive terms, and district population. The analyses were stratified by age group (under five years [U5] and individual five years and older [O5]) and by division.

Results

Strong seasonal trends were observed: AWD incidence peaked in Quarter 1 (January–March) for both the U5 (RR 1.20, 95% CI 1.06–1.37) and O5 (RR 1.25, 95% CI 1.12–1.39) populations, compared to Quarter 4 (October–December). Quarter 3 (July–September) showed reduced risk (U5: RR 0.89, 95% CI 0.80–0.99; O5: RR 0.91, 95% CI 0.83–0.99). Prior-month AWD incidence consistently predicted current-month incidence (U5: RR 1.006, 95% CI 1.004–1.007; O5: RR 1.002, 95% CI 1.002–1.003). No significant associations between AWD incidence and average monthly rainfall, temperature, or humidity. Regional heterogeneity emerged: in the Central Division, elevated temperatures increased AWD risk for both age groups (U5: RR 1.03, 95% CI 1.00–1.07; O5: RR 1.05, 95% CI 1.02–1.07) and high humidity increased risk in U5 (RR 1.02, 95% CI 1.01–1.03); In the Western Division, elevated temperature was protective for U5 (RR 0.98, 95% CI 0.96–1).

Conclusion

AWD incidence in Fiji's Central and Western Divisions is driven primarily by seasonal and autoregressive patterns rather than average monthly meteorological factors. Strong Quarter 1 peaks (20–25% elevated risk) provide a predictable signal for proactive intervention. Context-dependent climate-disease relationships emerged, with Central Division showing temperature and humidity associations absent in Western region. These findings support calendar-based preparedness, with district-specific vulnerability assessments guiding pre-positioned WASH and medical resources in high-risk areas before the annual Quarter 1 surge.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26496-0.

Introduction

Pacific island countries (PICs), including Fiji, face increased risks from climate change due to their small land areas, geographic isolation, and reliance on climate-sensitive sectors [1]. These structural factors, combined with their limited adaptive capacity, place them among the most vulnerable climate regions worldwide [2]. According to the World Risk Index 2021, Oceania was identified as the most risk-exposed region globally and ranks Fiji as the 14th out of 181 countries [3]. Climate change in this region increasingly manifests through intense storms, prolonged droughts, erratic precipitation patterns, and rising sea levels, which amplify the effects of storm surges and king tides [4]. Changes in meteorological conditions including temperature, precipitation, and humidity have the potential to alter pathogen survival, modify vector and host behaviours, and shift patterns of human exposure pathways [5]. These effects are particularly pronounced for waterborne diseases, which are highly sensitive to short-term weather variability and extreme hydrological events such as floods and droughts [6]. As a result, PICs are at the forefront of global health and climate challenges, facing a spectrum of direct and indirect health risks that threaten population well-being.

Fiji provides an important case study for understanding climate-health dynamics within the Pacific region. The country’s climate is best described as tropical maritime, characterised by a wet season (November–April) that is warm and humid, and a dry season (May–October) [7]. Spatial climatic variation exists between divisions, which is well documented in both observational and modelling studies of Fiji’s climate [8]. Climate projections indicate that while tropical cyclones may become less frequent, their intensity is likely to increase, raising concerns for both health and economic resilience [9, 10].

These climatic patterns affect public health through multiple pathways. Episodes of extreme rainfall can overwhelm water supply and sanitation infrastructure, leading to contamination of drinking water and increased risk of waterborne diseases [1, 11]. Elevated temperatures and humidity influence pathogen viability and transmission, while affecting human behaviours related to water storage and hygiene practices. These factors contribute to seasonal and inter-annual variability in infectious diseases, particularly diarrhea [12]. Vulnerable subpopulations including Fiji’s economically disadvantaged, marginalised, and remote communities bear unequal burdens from these effects [13].

Diarrheal diseases, especially acute watery diarrhea (AWD), are sensitive to climatic and environmental changes. The relationships between climate variables and diarrheal disease are well documented worldwide [14], but are complex and context-specific. AWD remains a public health concern in the Pacific, disproportionately affecting children U5 [15]. Multiple studies have demonstrated positive associations between diarrheal disease incidence and both elevated temperatures and extreme rainfall in Fiji, with outbreaks often occurring after flooding or drought [2, 12, 15]. However, findings from other contexts are heterogeneous and strongly influenced by local behavioural, infrastructural, and environmental factors [14, 16, 17]. Evidence from Pacific Island settings remains limited.

To date, most analyses in Fiji have aggregated data at the national or divisional level, or focused on single localities, with limited exploration at the district scale. The EWARS is an automated surveillance platform supporting both indicator and event-based reporting [18]. Introduced in Fiji in 2016, EWARS has reliably detected outbreaks while reducing the reporting burden and human error. Automated data collation and threshold-based alerts, as outlined in national SOPs, enable health staff to focus on timely response [9].

The availability of district-level EWARS surveillance data, combined with meteorological records, provides an opportunity to examine climate health relationships at a granular scale. The objectives of this study are to: (I) assess whether meteorological factors (temperature, rainfall, and humidity) are associated with AWD incidence across U5 and O5 populations, and (II) examine seasonal variation and autoregressive effects on AWD incidence at the district and divisional scales, with comparisons between the U5 and O5 populations.

Methods

Study design

We conducted an ecological time series analysis to evaluate the relationships between meteorological variables and acute watery diarrhea (AWD) incidence in Fiji’s Central and Western Divisions from Week 1, 2016 (January 4–10, 2016) through Week 52, 2023 (December 25–31, 2023).

Research location

Fiji comprises 332 islands divided into four administrative divisions: Central, Western, Eastern, and Northern [19]. The Central and Western Divisions together account for most of Fiji’s population and include its largest urban centres (Suva, Nadi, Lautoka) alongside extensive rural catchments [20]. The Central Division, on the windward side, receives more than 4,000 mm of annual rainfall, whereas the leeward Western Division receives less than 2,000 mm, making it more prone to drought and water scarcity [21, 22].

Seasonal and interannual variability is strongly influenced by the South Pacific Convergence Zone (SPCZ), which shifts position with the El Niño–Southern Oscillation (ENSO) [7]. When the SPCZ is positioned closer to Fiji, the Central Division experiences prolonged wet periods and increased flood risk. Conversely, displacement of the SPCZ north or east can result in reduced rainfall, especially in the western division [23, 24]. Fiji’s location within the South Pacific cyclone belt compounds its vulnerability, alongside the broader climate-sensitive exposures described above [25, 26].

This study focused on 14 districts within Fiji’s Central (n = 4) and Western (n = 8) divisions (Fig. 1). Among the 39 districts in these two divisions, only those linked to a weather station with complete rainfall, temperature, and humidity records and with established EWARS sites were included. The district level was chosen because it is the lowest geographic unit where EWARS surveillance data, population estimates, and meteorological observations can be consistently matched. Weekly AWD notifications were often sparse in small catchments. To address this, the data were aggregated to monthly counts, which reduced random variability, and supported robust assessment of the climatic associations. These 14 districts span both rural and urban catchments within the Central and Western Divisions, ensuring coverage across diverse settlement types.

Fig. 1.

Geographic distribution of the 14 districts included in the analysis, located within Fiji’s Central (purple) and Western (yellow) Divisions. Base map adapted from Fiji Bureau of Statistics geographic dashboard (2025) [27]

Data source

Acute watery diarrhea

AWD notifications were recorded at the early warning alert and response system (EWARS) sites (four in the Central Division and eight in the Western Division). AWD is defined in EWARS as the passage of ≥ 3 loose or watery stools within 24 hours without visible blood [9, 28]. Patient age was pre-categorised within the EWARS system into two groups: under five years (U5) and individuals aged five years and older (O5), with cases recorded to their EWARS site location.

Meteorological data

Daily climatic data were sourced from 12 weather stations managed by the Fiji Meteorological Service, ensuring representation across high and low rainfall zones. Two stations provided coverage for multiple districts due to close geographic proximity: Nadi station covered the Nadi and Nawaka districts, whereas Navua station covered the Serua and Veivatuloa districts. The remaining 10 stations were mapped to a single district within its geographic catchment. The stations provided consistent measurements of daily precipitation, minimum temperature (°C), maximum temperature (°C), and relative humidity (%). The daily mean temperature was calculated as (Tmax + Tmin)/2. Daily records were aggregated to monthly averages for rainfall (mm), temperature (°C) and humidity (%) for analysis.

Population data

District-level population estimates were sourced from the 2017 Fiji Bureau of Statistics census. Age-disaggregated denominators were used to calculate incidence rates and as offsets in regression models.

Data preparation

The dataset was pre-processed to support time-series analysis and included the following:

• Outcome variables: Monthly counts of reported AWD cases in children under five (U5) years and individuals five years and older (O5).

• Meteorological predictors: monthly averages of rainfall, temperature, and humidity.

• Covariates: District, year, and seasonal indicators Quarter 1: January–March; Quarter 2: April–June; Quarter 3: July–September).

Categorical variables such as year, weather station, province, and district were converted into factors, and the data were sorted chronologically by year and month. Missing meteorological values were rare; when present, they were imputed using station-level monthly means [29]. In sensitivity checks, the exclusion of months with > 20% missing values did not change the direction of effect estimates. The outcome variables were converted to numeric formats and inspected for anomalies.

Time series stationarity was confirmed via the augmented Dickey–Fuller test, with all the variables were being stationary (p < 0.01). Lagged terms up to two months for exposure and outcome variables were selected based on lag structures in previous climate-health studies and environmental persistence of AWD-associated pathogens, including E. coli, Shigella, Rotavirus, and S. typhi [30, 31]. Quarterly indicators (Q1–3) were created to represent seasonal patterns, with Quarter 4 (October–December) as the reference category because descriptive analyses indicated it had the lowest mean monthly AWD incidence over the study period. District–level variation was modelled using random intercepts for the district-year, and combinations and log-transformed population sizes were included as offsets.

Statistical analysis

Preliminary models used generalised linear models (GLMs) with a Poisson distribution to estimate the monthly AWD incidence. Diagnostic checks indicated overdispersion, prompting the use of negative binomial mixed-effects regression as the final modelling approach. Separate models were fitted for AWD cases in U5 and O5. Each model included lagged dependent variables (autoregressive terms), as well as current and lagged values (0–2 months) of the key meteorological variables: rainfall, temperature, and humidity.

Seasonality was controlled via quarterly indicators aligned with Fiji’s climate: Q1 (wet/cyclone season), Q2 (transition), Q3 (dry), and Q4 (late wet/cyclone season; reference). This approach reflects Fiji's distinct wet/dry regime and aligns with public health planning cycles [32], rather than assuming smooth sinusoidal transitions. Alternative specifications using Fourier terms and natural cubic splines were assessed via sensitivity analyses (see Supplemental Table 1), with results directionally consistent across approaches. Random intercepts were specified to account for clustering at the district-year level. Model performance was evaluated using Akaike information criterion (AIC), likelihood ratio tests, and residual diagnostics conducted via the DHARMa package. Monthly aggregation resulted in minimal zero counts (U5: 6.1%; O5: 2.9%), and DHARMa diagnostics confirmed appropriate model specification, including tests for overdispersion and zero-inflation. Model coefficients were exponentiated to provide relative risks (RRs) with 95% confidence intervals (CIs).

In addition to regression modelling, we conducted a district-level prioritisation to identify districts at high risk during seasonal surge. For each district, we calculated: (1) the mean monthly AWD incidence in both U5 and O5 groups, (2) the Quarter 1 seasonal surge factor (the ratio of mean Q1 incidence to annual mean), and (3) variability in case counts (standard deviation divided by the mean). Each metric was standardised using z-scores and summed to generate a composite vulnerability score for each district. Districts were then grouped into high, moderate, or standard vulnerability using tertiles of the composite score distribution (see Supplemental Material 1). All analyses were conducted using R Statistical Software (v 4.4.2; (2024–10–31).

Results

The final dataset comprised 1,125 monthly observations across 14 districts. Prior to imputation, meteorological variables (monthly average rainfall, temperature, and humidity) had ≤ 1.2% missing data, which were imputed using station-level monthly means. After imputation, all key variables were complete (Table 1). From January 2016 to December 2023, AWD cases were reported across Fiji’s Central and Western Divisions. In absolute numbers, more cases were recorded in the O5 cohort, but when adjusted for population size, the incidence was higher in U5. The median number of monthly U5 cases ranged from 2 in Navosa to 99 in Nadi & Nawaka, while the number of O5 cases ranged from 13 in Navosa to over 265 in Suva. The highest U5 incidence rates were observed in Bau (19.6 per 1,000) and Rakiraki (13.9 per 1,000), whereas the O5 incidence rates were lower, ranging from 0.82 per 1,000 in Serua & Veivatuloa to 4.05 per 1,000 in Lomaivuna. The Central Division districts recorded higher rainfall and humidity levels (~ 187 mm, 81–84%) relative to Western Division districts (~ 147 mm, 73–78%), whereas the maximum temperatures occurred in the Western Division (27.3 °C in Tavua).

Table 1.

Median monthly acute watery diarrhea (AWD) case counts, incidence rates per 1000 population, and mean meteorological variables by district and age group, Fiji, 2016–2023

District	Division	Median monthly cases (U5)	Median monthly cases (O5)	Median incidence per 1,000 (U5)	Median incidence per 1,000 (O5)	Mean Rainfall (mm)	SD Rainfall	Mean Temp (°C)	SD Temp	Mean Humidity (%)	SD Humidity
Ba	western	33	98.5	9.87	2.73	147.1	59.04	23.3	4.57	75.5	6.98
Baravi	western	5	19	5.43	2.56	146.8	53.58	25.8	1.72	76.4	7.35
Bau	central	66	81.5	19.63	2.95	186.6	60.21	25.7	1.56	81.5	3.45
Lomaivuna	central	8	19	11.03	4.05	187.7	60.89	25.1	2.37	84.4	15.73
Nadi & Nawaka	western	98.5	110.5	13.27	1.62	148.1	55.24	26.3	1.27	72.8	5.62
Nasigatoka	western	16	51	10.23	3.92	146.8	54.12	25.1	1.84	81.4	5.05
Navosa	western	2	13	3.05	2.92	148.9	54.93	24.7	3.16	75	23.07
Rakiraki	western	19	40.5	13.93	3.23	148.6	55.47	25.8	1.56	76.8	4.04
Serua & Veivatuloa	central	7	16	2.85	0.82	186.2	59.65	22	4.73	81.6	4.82
Suva	central	59.5	265.5	6.79	3.11	186.6	59.99	26	1.77	79.4	2.94
Tavua	western	27	27	10.57	1.23	147.1	56.21	27.3	1.17	78.7	9.64
Vuda	western	52	205	5.48	2.27	147.8	55.27	26	2.17	74.2	7.22

Median monthly case counts are based on reported AWD cases aggregated by district and age group over the study period. Incidence rates were calculated per 1,000 population using 2017 census denominators. Meteorological variables represent district-level monthly means of rainfall (mm), temperature (°C), and relative humidity (%). U5 = children under 5 years; O5 = individuals 5 years and older; SD = standard deviation

Model selection

The negative binomial models provided substantially better fit than the Poisson models (AIC for U5: 9,001.2 vs. 15,323.5; O5: 10,625.9 vs. 22,897.3), with significant overdispersion detected in Poisson residuals (p < 0.001) (see Supplemental Table 2). DHARMa diagnostic plots showed no evidence of systematic bias.

Age-stratified analysis of AWD temporal patterns

The incidence of U5 AWD in the previous month was associated with higher incidence in the current month (RR: 1.006, 95% CI: 1.004–1.007) (Table 2). The risk of AWD was higher in Quarter 1 than in the reference quarter (RR: 1.20, 95% CI: 1.06–1.37), whereas Quarter 3 (July–September) had a lower risk (RR: 0.89, 95% CI: 0.80–0.99). No significant association between Quarter 2 and AWD incidence was observed (RR: 0.93, 95% CI: 0.83–1.05).

Table 2.

Associations between meteorological factors, autoregressive terms, and acute watery diarrhea (AWD) incidence, Fiji, 2016–2023, stratified by under five years (U5) and individuals aged five years and older (O5)

Variable	Under 5 years (U5) RR (95% CI)	Over 5 years (O5) RR (95% CI)
Autoregressive terms
Cases in previous month (same age group)	1.006 (1.004–1.007)	1.002 (1.002–1.003)
Cases two months prior (same age group)	1.000 (0.998–1.001)	1.000 (0.999–1.000)
Meteorological variables
Average monthly Rainfall, current month (mm)	0.999 (0.992–1.006)	0.997 (0.991–1.003)
Average monthly Rainfall, lag 1 (mm)	0.997 (0.990–1.004)	0.996 (0.990–1.002)
Average monthly Rainfall, lag 2 (mm)	0.996 (0.989–1.003)	0.994 (0.988–1.000)
Average monthly Temperature current month (°C)	0.989 (0.974–1.004)	1.001 (0.988–1.014)
Average monthly Temperature, lag 1 (°C)	1.006 (0.990–1.023)	0.997 (0.983–1.010)
Average monthly Temperature, lag 2 (°C)	1.006 (0.994–1.018)	1.005 (0.994–1.015)
Average monthly Humidity current month (%)	1.005 (0.999–1.012)	1.000 (0.995–1.006)
Average monthly Humidity, lag 1 (%)	0.998 (0.991–1.005)	1.001 (0.995–1.007)
Average monthly Humidity, lag 2 (%)	0.998 (0.991–1.004)	1.000 (0.995–1.005)
Seasonal indicators
Quarter 1 (Jan–Mar)	1.201 (1.057–1.365)	1.249 (1.122–1.391)
Quarter 2 (Apr–Jun)	0.934 (0.827–1.054)	1.004 (0.909–1.110)
Quarter 3 (Jul–Sep)	0.889 (0.796–0.994)	0.905 (0.825–0.992)

Values are relative risks (RRs) with 95% confidence intervals (CIs) from negative binomial mixed-effects regression models. Autoregressive terms represent AWD cases from the previous one and two months. Meteorological variables include current month and lagged values (1–2 months) for rainfall, temperature, and humidity. Quarter 4 (October–December) served as the reference category for seasonal indicators. Models included random intercepts for district-year clusters and log-population offsets to account for varying district population sizes

In the O5 cohort, a similar pattern was observed. Previous reported AWD cases significantly predicted current incidence (RR: 1.002, 95% CI: 1.002–1.003) (see Supplemental Fig. 2). Seasonal effects paralleled the U5 cohort: the risk of AWD was elevated in Quarter 1 compared with the reference quarter (RR: 1.25, 95% CI: 1.12–1.39) and reduced in Quarter 3 (RR: 0.905, 95% CI: 0.83–0.99). No significant association between Quarter 2 and AWD incidence was observed (RR: 1.00, 95% CI: 0.91–1.11). The incidence of AWD was consistently lowest during Quarter 3 in both age cohorts (Fig. 2).

Fig. 2.

Relative risks (RR) of acute watery diarrhea (AWD) incidence by calendar quarter, Fiji, 2016–2023, stratified by age group. Estimates are from negative binomial regression models for children under five years (U5, blue) and individuals five years and older (O5, red), with Quarter 4 as the reference category (RR = 1.0)

Climate variables and AWD associations

No significant associations were detected between AWD incidence and rainfall, temperature, or humidity variables, current or lagged, in the adjusted model. Across both age groups, the effect sizes for meteorological variables were close to null after controlling for temporal correlation and seasonal factors.

Division-stratified analysis

In the Western Division, reported cases of current month AWD were associated with reported cases in the previous month in the U5 (RR: 1.006, 95% CI: 1.004–1.008) and O5 populations (RR: 1.002, 95% CI: 1.001–1.003) (Table 3). AWD incidence was associated with increased risk in Quarter 1 (January–March) for both age groups (U5: RR: 1.28, 95% CI: 1.08–1.52, p = 0.0054; O5: RR: 1.30, 95% CI: 1.13–1.50). Elevated temperature in the current month was associated with reduced risk in the U5 population only (RR: 0.979, 95% CI: 0.961–0.997). No significant associations were detected between rainfall or humidity variables and AWD incidence in the western division.

Table 3.

Associations between meteorological factors, autoregressive terms, and reported acute watery diarrhea (AWD) incidence, comparing Western and Central Divisions of Fiji, 2016–2023, stratified by under five years (U5) and individuals aged five years and older (O5)

Variable	Western		Central
	U5 RR (95% CI)	U5 RR (95% CI)	O5 RR (95% CI)	O5 RR (95% CI)
Autoregressive terms
Cases in previous month (same age group)	1.006 (1.004–1.008)	1.005 (1.003–1.008)	1.002 (1.001–1.003)	1.002 (1.001–1.003)
Cases two months prior (same age group)	1.000 (0.999–1.002)	0.999 (0.997–1.001)	1.000 (0.999–1.000)	1.000 (0.999–1.001)
Rainfall
Average monthly Rainfall, current month (mm)	1.000 (0.990–1.010)	0.995 (0.984–1.006)	0.997 (0.990–1.005)	0.991 (0.982–1.001)
Average monthly Rainfall, lag 1 (mm)	0.996 (0.986–1.006)	0.998 (0.987–1.009)	0.997 (0.988–1.005)	0.994 (0.984–1.003)
Average monthly Rainfall, lag 2 (mm)	0.996 (0.986–1.006)	0.994 (0.983–1.005)	0.997 (0.989–1.005)	0.988 (0.979–0.997)
Temperature
Average monthly Temperature, current month (°C)	0.979 (0.961–0.997)	1.033 (1.002–1.065)	0.989 (0.974–1.004)	1.045 (1.019–1.072)
Average monthly Temperature, lag 1 (°C)	1.010 (0.989–1.032)	1.009 (0.979–1.040)	1.003 (0.986–1.021)	1.001 (0.979–1.024)
Average monthly Temperature, lag 2 (°C)	1.002 (0.983–1.020)	1.004 (0.987–1.021)	0.997 (0.981–1.012)	1.013 (0.998–1.029)
Humidity
Average monthly Humidity, current month (%)	1.000 (0.992–1.008)	1.019 (1.006–1.032)	0.998 (0.992–1.005)	1.006 (0.996–1.017)
Average monthly Humidity, lag 1 (%)	0.998 (0.990–1.007)	0.997 (0.983–1.010)	1.002 (0.995–1.009)	0.998 (0.988–1.009)
Average monthly Humidity, lag 2 (%)	0.997 (0.989–1.005)	0.997 (0.986–1.009)	0.999 (0.993–1.006)	1.003 (0.994–1.012)
Seasonal Indicators
Quarter 1 (January–March)	1.280 (1.076–1.523)	1.060 (0.871–1.289)	1.302 (1.129–1.501)	1.102 (0.934–1.299)
Quarter 2 (April–June)	0.988 (0.844–1.157)	0.834 (0.689–1.009)	1.068 (0.940–1.215)	0.841 (0.717–0.986)
Quarter 3 (July–September)	0.889 (0.776–1.017)	0.905 (0.741–1.106)	0.971 (0.869–1.084)	0.783 (0.662–0.927)

The values are relative risks (RRs) with 95% confidence intervals (CIs) from negative binomial regression models. The models were stratified by age group (U5 and O5) and by division (Western, Central). Quarter 4 (October–December) served as the reference category. RR > 1 indicates increased AWD risk per unit increase in predictor, RR < 1 indicates reduced AWD risk. All the models were adjusted for district–year clustering (random intercepts) and population size (log-population offset)

For the Central Division, reported AWD cases in the current month were significantly associated with reported cases in the previous month across both age groups (U5: RR: 1.005, 95% CI: 1.003–1.008, O5: RR: 1.002, 95% CI: 1.001–1.003) (see Table 3). Elevated temperature in the current month was associated with increased AWD incidence in both the U5 (RR: 1.033, 95% CI: 1.002–1.065) and O5 populations (RR: 1.045, 95% CI: 1.019–1.072). Elevated humidity in the current month was also associated with increased AWD incidence in the U5 group (RR: 1.019, 95% CI: 1.006–1.032, p = 0.0038). In the O5 population, AWD risk decreased with increasing rainfall two months prior (RR: 0.988, 95% CI: 0.979–0.997), Quarter 2 (RR: 0.841, 95% CI: 0.717–0.986), and Quarter 3 (RR: 0.783, 95% CI: 0.662–0.927).

The magnitude and timing of the peak incidence of reported AWD cases varied across administrative divisions. Seasonal variation was more pronounced in the western division, with sharp increases during Quarter 1, whereas the central division showed less seasonal fluctuation as shown in Fig. 3.

Fig. 3.

Seasonal variation of acute watery diarrhea (AWD) incidence (per 1,000 population) in the Central and Western Divisions of Fiji, 2016–2023. Polar plots illustrate AWD incidence across calendar quarters (Q1–Q4). Data represent combined cases in U5 and O5. Peak incidence occurred in Quarter 1 (January–March), with sharper seasonal variation in the Western Division compared to the Central Division, where fluctuations were less pronounced. Shaded areas represent 95% confidence intervals around quarterly AWD rates

District-level prioritisation during Quarter 1

District vulnerability assessment identified three high-priority districts for Q1 preparedness: Suva, Vuda, and Bau (Fig. 4, Supplemental Fig. 1). These districts combined elevated baseline AWD burden with Q1 seasonal surge factors exceeding the annual mean (≥ 1.04), with Bau recording the highest surge factor at 1.11. Notably, Suva and Vuda contain two of Fiji's three largest urban centres, emphasising the implications of seasonal surges for densely populated areas. Nadi and Nawaka, despite housing Fiji's third major urban centre, was classified as moderate priority due to Q1 incidence below the annual baseline (surge factor 0.94). Rakiraki and Nasigatoka showed similar below-baseline Q1 patterns (surge factors ≤ 0.94). Districts classified as low priority (Baravi, Lomaivuna, Serua & Veivatuloa and Tavua) demonstrated both lower baseline burden and modest seasonal variation. This prioritisation highlights where preparedness resources should be deployed ahead of predictable seasonal surge.

Fig. 4.

Prioritisation matrix for AWD intervention across districts, based on a combined index of vulnerability assessment and Q1 seasonal surge factor. The x-axis indicates the Quarter 1 seasonal surge factor, while colours represent intervention priority (red = high, yellow = medium, green = standard). Values on the bars denote the surge factor for each district. Classification was based on a composite of baseline incidence, seasonal surge, and population size

Discussion

This study represents the first district-level time-series analysis of acute watery diarrhea (AWD) in Fiji, leveraging EWARS surveillance data from 14 districts in the Central and Western Divisions linked with meteorological and census datasets. Using routinely collected data, we provide new insights into climate–health relationships that directly apply to public health practice in small island settings. The principal finding is a pronounced and predictable seasonal pattern in AWD, with peak incidence of reported AWD cases in Quarter 1 (January-March) showing elevated risk for both U5 (RR: 1.20, 95% CI: 1.06–1.37) and O5 (RR: 1.25, 95% CI: 1.122–1.391). This pattern mirrors the seasonal pattern reported in Fiji [9] and in Iran [33], Australia [34] and Nepal [35], though direct comparisons are influenced by differences in season classification. Conversely, AWD incidence was consistently lowest in Quarter 3 (July–September), a period characterised by a decreased in rainfall, increased temperatures and lower humidity 26, 36]. This consistent Q1 surge, which persists after adjustment for meteorological variables and temporal autocorrelation, represents a reliable epidemiological signal that can form the basis of a proactive public health calendar and indicates drivers other than climate variables [17]. These include: (1) coincident climate extremes such as intense rainfall extending from December [26] and the peak cyclone season [7]; and socio–behavioural factors such as increased population mobility from rural to urban areas and vice versa during the festive period and school-year commencement. However, this remains a plausible hypothesis, as definitive evidence linking these socio-behavioural patterns to AWD seasonality in Fiji is currently lacking.

The observation that AWD seasonal trends were stronger in the Western Division, with sharp Quarter 1 peaks in both age groups, underscores the importance of local context. This likely reflects the division’s drier leeward climate, which is more prone to flooding and orographic rainfall patterns [4, 7, 15]. Elevated temperatures showed opposite associations across divisions: moderately protective for Western Division children U5 but risk-enhancing in the Central Division across both age groups. In the Central Division, humidity was also associated with increased risk in U5. These contrasting patterns may reflect differences in water security and adaptive behaviours. Understanding these requires examining whether temperature-AWD associations vary by water supply reliability and whether rainfall-related risk differs between gravity-fed and groundwater systems.

Prior-month AWD cases strongly predict current-month incidence across all subgroups. This pattern may reflect multiple mechanisms. Environmental pathogen persistence in contaminated water, soil, and surfaces may enable sustained exposure [37]. Additionally, socioeconomic barriers and delayed healthcare-seeking may prolong transmission by delaying case detection, while intrahousehold transmission can sustain secondary infection cycles [38, 39]. However, our ecological study design cannot definitively distinguish among these pathways. The observed temporal autocorrelation may also reflect unmeasured time-varying factors such as seasonal behaviours, healthcare utilization, or reporting bias [40]. This autoregressive pattern indicates that current case burden predicts subsequent risk, enabling rapid, localized response before clusters escalate. Importantly, AWD overlaps symptomatically with leptospirosis and typhoid fever, which share similar transmission pathways [9], suggesting that seasonally timed WASH interventions could benefit prevention of multiple communicable diseases in Fiji [41–43].

The absence of significant associations between monthly meteorological variables (rainfall, temperature, humidity) and AWD incidence likely reflects several contextual and methodological factors. First, extreme weather events such as cyclones, flooding, and prolonged dry spells are critical drivers of AWD risk in Fiji [4]. These discrete events disrupt water and sanitation systems abruptly and are poorly captured by monthly averages. Second, AWD incidence exhibits strong seasonal and autoregressive structure, explaining substantial temporal variation. Once these dominant patterns are modelled, residual meteorological contributions appear limited. Third, Fiji's relatively good water supply network and structured healthcare access may buffer populations from gradual climate fluctuations, except during major system disruptions. Levy et al. (2018) caution against treating AWD as universally climate-sensitive, emphasizing that local socioenvironmental conditions can modulate or override direct meteorological effects [17]

A pivotal finding is the lack of a significant association between average monthly rainfall and AWD incidence at the national level. This should not be interpreted as evidence that climate is irrelevant; rather, the predictable seasonal baseline burden of AWD is driven by factors beyond short-term meteorological fluctuations. This aligns with studies from Indonesia [44], Federated States of Micronesia [45] and Afghanistan [46]. Studies in Taiwan [47], Ethiopia [48] and two previous studies in Fiji documented positive associations, with one observing a U-shaped association at one-month lag [12, 15]. Our findings diverge from those earlier Fiji studies due to differences in study period, scale, and case definitions: both earlier studies preceded rotavirus vaccination, relied on national/sub- divisional aggregation, and included all diarrheal cases, whereas our analysis used recent EWARS data at district level. Nelson et al. (2022) similarly found rainfall contributed little explanatory power, consistent with our findings [9]. However, extreme rainfall events remain epidemiologically important, as demonstrated by AWD outbreaks following severe flooding in Solomon Islands [49]. This suggests rainfall-related AWD risk in Fiji may be driven more by extreme events than average monthly variation [42], an effect likely to intensify under climate change projections [4, 50].

The relationship between temperature and diarrhoeal disease remains context dependent. While global reviews and an earlier national-level study in Fiji reported positive associations [15, 51, 52], our overall analysis revealed no consistent effect of temperature on AWD incidence. This contrasts with previous Fijian studies that described uniformly positive associations [12, 15], demonstrating the value of finer geographic resolution for understanding health–climate relationships. In Fiji, the district is also the first administrative tier for health planning and resource allocation [53]. Similar heterogeneity has been reported elsewhere, including in India [54], Australia [34], and recent analyses in Fiji [9].

Similarly, we found no significant association between humidity and AWD incidence in either the U5 or O5 populations at the national level, which aligns with findings from Ethiopia [48] and Fiji [12]. In contrast, positive associations have been reported in Vietnam [14], Bangladesh [55], Indonesia [56] and Cape Town [57]. The influence of humidity may be obscured by its strong collinearity with rainfall and temperature [58]. In Fiji, large-scale climate phenomena such as the El Niño–Southern Oscillation (ENSO) strongly shape weather patterns, with El Niño often resulting in drought conditions and La Niña bringing excessive rainfall, both with clear implications for AWD [59].

Our findings have direct and practical implications for shifting towards a proactive public health response in Fiji. Integrating seasonal forecasts and ENSO indices into the EWARS can enhance its role as an anticipatory tool. With Fiji projected to experience more intense rainfall events, AWD risk may increase, especially in flood-prone areas with fragile water and sanitation infrastructure [50]. Strengthening the climate-adaptiveness of syndromic surveillance, including the integration of meteorological indicators (e.g., rainfall, temperature and ENSO indices), may improve the early detection of high-risk periods [60]. AI-driven models using these indicators have already shown promise for diarrheal disease prediction in the Asia–Pacific region [61]. These advances require coordinated action between the health and meteorological sectors, supported by sustainable climate–health financing [60].

Targeted interventions before high-risk Quarter 1 remain essential. These include: (1) prepositioning ORS and zinc in health care facilities in high-priority districts identified in our vulnerability assessment (Fig. 4); (2) launching hygiene promotion campaigns ahead of the December festive season and school reopening in January; and (3) prepositioning of WASH emergency supplies (i.e., WASH kits, water purification tablets and emergency sanitation) in flood-prone areas. District-specific and municipal councils preparedness allows the strategic deployment of resources before the annual Q1 peak and cyclone season. Embedding this localised intelligence into the Fiji WASH cluster's coordination platforms ensures that response capacity is amplified in advance.

We acknowledge several limitations. The ecological design precludes individual-level inference, and unmeasured confounders such as hygiene practices and health-seeking behaviour that may have influenced the results. For instance, if hygiene improvements (e.g., boiling of drinking water) during post extreme weather events, these could lower AWD incidence independently of climate. Discrete extreme weather events such as cyclones and floods were not modelled due to unavailable event data, potentially missing key climate-health relationships. Reliance on 2017 census denominators across the 2016–2023 period may not capture recent population growth in districts, though a recent sensitivity analysis demonstrated stable patterns under plausible measurement error, suggesting minimal bias from fixed denominators [62]. Our key findings, based on relative patterns (e.g., within-district seasonal surge factors), are minimally affected by population changes. Despite these constraints, the strengths of this study include the use of routine district-level EWARS data integrated with meteorological and census information, providing locally relevant evidence for public health planning. Stratification by age and division revealed heterogeneity that can guide targeted interventions. Future research should extend this work by incorporating finer-scale data on WASH infrastructure, sociocultural practices, and health-seeking behaviour, and by explicitly modelling extreme events such as cyclones and floods. Extension to Fiji's Eastern and Northern Divisions would provide comprehensive national coverage.

Conclusion

This study demonstrated that acute watery diarrhea (AWD) in Fiji is shaped primarily by seasonal and autoregressive patterns, with a consistently elevated incidence in Quarter 1, rather than in response to the direct effects of average monthly rainfall, temperature, or humidity. The prominent role of temporal patterns and the modest effect of aggregated climate exposures suggest public health preparedness must pivot to a proactive model. This involves institutionalizing a preparedness calendar that triggers targeted WASH investments, resource prepositioning, and communication campaigns in high-risk districts ahead of the annual Q1 surge and cyclone season. Strengthening EWARS through the integration of this simple seasonal forecast and our district-level vulnerability assessment can enhance early action and support the development of a climate-resilient health system that anticipates epidemics before they begin.

Abbreviations

AWD

Acute watery diarrhea

EWARS

Early warning alert and response system

FBoS

Fiji Bureau of Statistics

FCDC

Fiji Centre for Disease Control

FMS

Fiji Meteorological Service

MoHMS

Ministry of Health and Medical Services

U5

Children under five years

O5

Individuals five years and older

WASH

Water, Sanitation and Hygiene

Authors’ contributions

SMB conceived and designed the study, curated data, interpreted results, and drafted the manuscript. NO provided methodological oversight and critically revised the manuscript. PD contributed to study design, performed analyses, and interpreted results. LV and AV facilitated data access and contributed to interpretation. SAR contributed to study conception, interpretation, and critical revision of the manuscript. All authors read and approved the final manuscript and agree to be accountable for all aspects of the work.

Funding

The primary author (SB) is supported by the Australian Government – Research Training Program scholarship (RTP). The scholarship provider had no role in the study design, data analysis, or interpretation of the results.

Data availability

The datasets analysed in this study are not publicly available due to restrictions by the Fiji Ministry of Health and Medical Services (MoHMS), the Fiji Meteorological Service (FMS), and the Fiji Bureau of Statistics (FBoS). However, they are available from the corresponding author on reasonable request and with permission from MoHMS, FMS, and FBoS.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the University of Queensland Human Research Ethics Committee (Exemption No: 2023/HE000010). The study analysed de-identified datasets and did not involve human participants. It was conducted in accordance with the Declaration of Helsinki and the Australian National Statement on Ethical Conduct in Human Research. Approval to access data was granted by the Fiji Ministry of Health and Medical Services, the Fiji Meteorological Service, and the Fiji Bureau of Statistics.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.