Daily mean temperature and HFMD: risk assessment and attributable fraction identification in Ningbo China

Hand, foot, and mouth disease (HFMD) is a common infectious disease, particularly affecting infants and children under 6 years old, though older children and adults can be infected. It is caused by various enteroviruses (e.g., Coxsackievirus A16, Enterovirus 71) and typically presents with fever, mouth ulcers, and vesicles on the hands and feet; a small proportion progress to severe neurological complications. Prior research has examined meteorological influences (temperature, humidity, altitude) on HFMD incidence, but associations vary by region and season, exhibiting single or bimodal peaks depending on local climatic and sociodemographic factors. In China, incidence patterns differ between northern and southern cities. While temperature’s contribution to HFMD burden is important for policy and prevention, it has not been quantified for Ningbo. This study investigates the relationship between daily mean temperature and HFMD incidence in Ningbo and quantifies the attributable fraction (AF) of HFMD due to temperature, to inform prevention and control strategies.

Previous studies identified meteorological factors (temperature, relative humidity, altitude) as important correlates of HFMD incidence, with pronounced geographic heterogeneity: single annual peaks in some countries (e.g., summer in Japan, autumn in Finland) and bimodal patterns in others (e.g., summer and autumn in the UK and Belgium). Within China, northern cities (e.g., Beijing, Hefei) often peak around June, whereas several southern cities (e.g., Kunming, Guangdong) show peaks in May and October–November. Discrepancies likely reflect climatic conditions, socioeconomic status, and demographic differences. Despite extensive research on associations, few studies quantify temperature-attributable HFMD risk, and none had done so for Ningbo prior to this work.

Study setting: Ningbo, southeastern China (28°51′–30°33′N, 120°55′–122°16′E), with humid subtropical climate: mean annual temperature 16.4 °C (January 4.7 °C; July 28.0 °C), average annual rainfall 1480 mm, and monsoonal plum rains in June. Population was ~8.05 million in 2017. Data: Daily HFMD cases (January 2014–December 2017) were obtained from the Data-center of China Public Health Science, diagnosed per National Health Commission guidelines. Cases were stratified by gender (male, female) and age (0–3, 4–5, ≥6 years). Meteorological data (daily mean temperature [°C], relative humidity [%], precipitation [mm]) were from the Yinzhou national meteorological station. Statistical analysis: A distributed lag non-linear model (DLNM) with quasi-Poisson regression quantified associations between daily mean temperature (Tmean) and HFMD incidence over lags 0–30 days. Model: log(E(Y_t)) = α + cb(Tmean_t, lag=30) + ns(time, df=7/year) + ns(RHmean_t, 3) + β·DOW. The cross-basis used a quadratic B-spline with 4 df for exposure-response and a natural cubic spline with 4 df for lag-response. Seasonality and long-term trends were controlled by a smooth time function (7 df/year). Relative humidity (RHmean) was modeled as ns with 3 df; day of week (DOW) entered as categorical. Reference temperature was chosen at the observed minimum risk (14 °C) identified from the exposure-response curve. Cumulative relative risks (CRRs) were computed for selected representative low (7 °C) and high (24 °C) temperatures across lag 0–30. Attributable fraction (AF): Temperature-attributable HFMD burden was estimated using the backward perspective within the DLNM framework: b−AF_xt = 1 − exp(− Σ_{l=lo}^L β_l x_{t−l}), with empirical 95% confidence intervals derived via Monte Carlo simulations from the multivariate normal distribution of model coefficients. Stratified analyses: Gender- and age-specific models were fitted; differences between subgroups were tested using (Q1 − Q2) ± 1.96·sqrt(SE1^2 + SE2^2). Sensitivity analyses: Checked collinearity between RHmean and precipitation (correlation 0.35); varied df for RHmean; compared models with/without RHmean and precipitation. Analyses used R 3.6.2 and the ‘dlnm’ package; two-tailed P < 0.05 indicated statistical significance.

• Dataset included 129,897 HFMD cases (2014–2017); 59.16% male, 40.84% female; age distribution: 0–3 years 76.43%, 4–5 years 17.19%, ≥6 years 6.38%.
• Time series showed strong bimodal seasonality with peaks in June and November; overall trend decreased (2014–2015), rose (2016), then decreased (2017).
• Temperature-HFMD association was significant over lag 0–30 days, with an M-shaped exposure-response and two peaks: low temperatures (5–11 °C) and high temperatures (16–29 °C). The minimum risk (reference) temperature was 14 °C.
• Representative low temperature (7 °C): highest single-lag RR 1.08 (95% CI: 1.06–1.10) at lag 3; CRR over lag 0–30 = 2.22 (95% CI: 1.61–3.07).
• Representative high temperature (24 °C): highest single-lag RR 1.06 (95% CI: 1.04–1.07) at lag 14; CRR over lag 0–30 = 3.54 (95% CI: 2.58–4.88).
• Sex-specific CRRs at 7 °C (lag 0–30): male 2.50 (1.77–3.55); female 1.89 (1.31–2.71). At 24 °C: male 3.12 (2.22–4.40); female 4.23 (2.94–6.07). Differences were not statistically significant.
• Age-specific CRRs at 7 °C: 0–3 years 2.04 (1.47–2.84); 4–5 years 3.64 (2.19–6.03); ≥6 years 1.84 (0.93–3.63). At 24 °C: 0–3 years 3.31 (2.40–4.59); 4–5 years 5.08 (3.08–8.36); ≥6 years 3.88 (2.00–7.55). Differences were not statistically significant.
• Attributable fraction (AF) of HFMD due to temperature (overall): 45.50% (95% CI: 36.21–52.48%). By temperature range: cold (5–11 °C) AF 5.23% (3.10–7.14%); hot (16–29 °C) AF 39.55% (30.91–45.51%).
• Sex-specific AFs (overall temperature): male 42.98% (31.51–51.23%); female 48.26% (39.09–54.77%).
• Age-specific AFs (overall temperature): 0–3 years 43.44% (33.14–51.41%); 4–5 years 53.34% (40.91–60.96%); ≥6 years 50.59% (26.93–62.89%). AFs were higher for hot than cold temperatures across subgroups.
• Sensitivity analyses indicated stable CRR estimates and no serious collinearity between RHmean and precipitation (r = 0.35).

The study addressed its central question by demonstrating a clear, non-linear association between daily mean temperature and HFMD incidence in Ningbo across lags up to 30 days, with a bimodal temperature-response consistent with observed seasonal patterns. Both cold and hot temperatures increased risk relative to the minimum-risk temperature (14 °C), but high temperatures contributed a substantially larger attributable burden, indicating heat plays a dominant role in HFMD transmission dynamics in this subtropical setting. Subgroup analyses suggested broadly consistent temperature effects across sexes and age groups, with slightly higher burdens among preschool/school-age children and females, though differences were not statistically significant. These findings underscore the public health importance of temperature in HFMD epidemiology and support incorporating meteorological surveillance into early warning systems and targeted prevention (e.g., hygiene promotion, childcare/school interventions) during high-risk temperature periods.

Both low and high daily mean temperatures are associated with increased HFMD incidence in Ningbo, with high temperatures accounting for a much larger share of the attributable burden. Using DLNM, the study quantified cumulative risks and temperature-attributable fractions overall and by demographic subgroups, providing evidence to inform timing and targeting of HFMD control measures. Future research could further investigate additional meteorological and environmental drivers, potential viral subtype differences, and multi-city comparisons to enhance generalizability and guide climate-informed public health strategies.