Climatic parameters affect the survival and transmissibility of respiratory viruses causing influenza-like illness (ILI), like influenza, respiratory syncytial virus and rhinovirus. In temperate countries, ILI activity is inversely associated with humidity and temperature (which affect both the onset and size of ILI epidemics), and we hypothesized that changes in the average values of these climatic parameters (which have occurred in Europe in recent decades) may explain the declining trends in ILI incidence rates observed during the same period. Here, we tested this hypothesis by linking ILI and climatic time series in the Netherlands for the period 1970–2016.

Influenza-like illness surveillance data for the period 1970–2016 were obtained from the Netherlands Institute for Health Services Research (NIVEL), which relies on a network of 60 sentinel general practitioners in 40 practices that are distributed throughout the Netherlands (including urban and rural areas) and cover approximately 0.7% of the Dutch population. Data on daily average absolute humidity (g/m³) and temperature for the weather station of De Bilt (which is located in the middle of the country) for the period 1970–2016 were obtained from the Royal Netherlands Meteorological Institute. Linear regression models were used to assess whether the season average ILI incidence rate changed over time, and whether there was any change over time in daily average absolute humidity and temperature. We modelled the association between climate and ILI activity through negative binomial regression models with the weekly ILI count as dependent variable, absolute humidity (or temperature) with up to three lagged weeks as independent variables, and covariates accounting for contagiousness of ILI-causing agents and presence of immunity. Model-based predictions were made using a leave-one-out cross-validation procedure. The study was funded by Sanofi-Pasteur.

The season average ILI incidence rate declined significantly (P<0.001) from 1970 and reached a minimum (17.3 cases per 100,000 populations) in the season 2002–03; subsequently, the trend reversed and the rate increased significantly (P<0.001) from the season 2003–04 onwards. After three decades of rising absolute humidity and temperature (1970–2000), there was a trend-reversal point in the early 2000s for the climatic time series, with absolute humidity and temperature tending to decrease, especially in spring and summer. In multivariate negative binominal regression models, current (week 0) and 1-week lagged absolute humidity were inversely associated with ILI count: the IRR for a change by 1g/m³ were 0.965 (95% CI: 0.951–0.979, P-value<0.001) and, respectively, 0.960 (95% CI: 0.944–0.976, P-value<0.001). Similar results were obtained for the association between temperature and ILI count. Instead, there was no significant association between ILI count and longer lagged values (2- or 3-week) of absolute humidity and temperature. The model accurately predicted the weekly ILI incidence in each season (although it tended to underestimate it around the seasonal peak) and the changing trend of season average ILI incidence rate during the study period.

Changes in absolute humidity and temperature were able to satisfactorily explain the long-term trends of ILI incidence rate in the Netherlands during 1970–2016. Temperate countries should monitor climate-related change in infectious diseases epidemiology so that measures to mitigate their impact can be planned and implemented.