Urged by global warming and evidence of prolonged pollen seasons, we performed the first clinical study investigating the relation between air temperature and symptomatic SAR in general practice. In an exploratory longitudinal study based on 25 years of robust general practice data we examined trends over time of the start and length of peaks in presentation of SAR symptoms. We found an earlier start (1.7 days/year) and prolonged duration (+1.3 days/year) of the SAR season over time, and a modest increase in contact frequency of SAR patients with their GP over time (+0.25 encounters over 25 years). Statistical testing showed a significant association between temperature from February until July and the duration of the SAR peak, although direct effects of a warmer start of distinct years on the timing of the start of the peak in this particular year were not observed (P = 0.06 for a warm March).

A major strength of this study is the data source: FaMe-Net is a reliable practice-based research network representative for the Dutch population regarding age, sex and social class21,22. It is the world’s longest uninterrupted primary care morbidity registration, unique in using an unaltered disease classification system over this long period of time21,22, which enabled us to study 25 years of general practice data in retrospect—starting long before climate change awareness raised. Although FaMe-Net’s population size is substantial, only large effects of temperature on the outcome measures could be shown conclusively, since each calendar year included counted as a single observation in the statistical models. Despite the robust dataset, this resulted in limited statistical power for our study. It’s questionable if the three practices we included are representative to all Dutch general practices. We tried to include more practices, but this introduced bias due to different follow-up periods between the practices.

SAR is considered a chronic disease, but the burden patients experience can alter over time23. We defined ‘SAR patients’ as patients contacting the GP for SAR within the particular year (contact prevalence proportion)24. This may have led to fluctuations in our recorded prevalence of SAR. The logical alternative, considering SAR as chronic disease after a patient’s initial diagnosis onwards, likely creates overestimation of clinically relevant SAR in an ageing population within our long study period. To increase internal validity, we included only the practices that continuously and uniformly registered during our entire study period of 25 years. To minimise potential bias due to a changing prevalence over the years we corrected our outcome measures by dividing by the number of SAR ‘patient years’. Obviously, only morbidity that patients present to their GP can be recorded so that it reflects a proportion of perceived symptoms. Other potential biases such as changes in consultation behaviour may also remain. Moreover, increased accessibility of SAR medication during our study period may have resulted in an underestimation of the increased health care demand (+0.25 encounters per SAR patient per year over 25 years)25.

Finally, misclassification might have played a role. Perennial and seasonal allergic rhinitis may co-occur as do allergic and non-allergic rhinitis26. Thanks to the ICD-10 subcoding in FaMe-Net we were able to exclude isolated registered ‘perennial AR’. However, it is not possible to register multiple specific subcodes of ‘allergic rhinitis’. This might have resulted in excluding SAR patients with predominant ‘other specific allergy’. On the other hand, presented ‘other’ allergic symptoms (other than specific ‘SAR’) of included ‘SAR patients’ may have been counted as ‘SAR’ encounters. This potential misclassification might have introduced bias, in either direction. We expect that such bias in two possible directions will mediate and has not had relevant impact on our results.

We explored a new and reproducible method to identify SAR peak-periods, but it might be influenced by extreme values. We limited the potential effects of extreme values by basing the cut-off point on the mean of all 25 years. Besides temperature, other factors, e.g. rainfall9, traffic-related pollutants27,28 and level of urbanisation27,29 may also influence the timing of the start and length of the SAR season and its severity, since they influence local pollen concentrations. We had no data on these potential confounders. In some years, we found more than one peak leading to a possible overestimation of the whole peak period within these particular years. It could have been helpful to couple these peaks to pollen concentrations to distinguish between peaks caused by tree and/or grass pollen. Unfortunately, data on pollen concentrations were not available for our study. Finally, local variations in contributing factors (e.g. temperature) might have played a role, but we were unable to determine them since our calculations were based on the CNT. Further research could relate clinical data to local factors (local temperature, pollen concentrations, level of urbanisation/pollution) and strengthen our results.

We interpret the observed increased frequency of SAR over time as increased severity of SAR. This corresponds with increasing prevalence of SAR after heat exposure17. A previous study reported more GP visits among patients with moderate/severe SAR compared to patients with mild disease28. Our findings are in line with the only previous study we identified that related temperature and timing of SAR symptoms. Kim and colleagues studied hospital visits as primary outcome and found that a high minimum temperature in March was positively correlated with the number of hospital visits by SAR patients from April to July18. Our primary care results suggest that a warm month of March could lead to an earlier start of the SAR season. The association we observed, however, did not reach statistical significance (P = 0.06).

This study derived presence of SAR symptoms from longitudinally registered GP encounters for SAR. Although GPs only see 1–2% of all patients with SAR symptoms30, these patients still lead to a substantial demand of health care. The recorded prevalence and the contact frequency of SAR patients with their GP we observed in our study correspond with those reported in another large Dutch primary care dataset31.

Apart from SAR, other respiratory conditions also seem to be influenced by climate change. Air pollution increases the incidence and severity of upper respiratory tract infections, and aggravates asthma. Interior warming of houses stimulates proliferation of allergens inside homes such as dust mites and fung32. As goes for SAR, the evidence could be strengthened to demonstrate that climate change indeed negatively impacts these other conditions.

The increased contact frequency for SAR we found over time [hypothesis 3] may reflect increased severity of SAR. We consider the increased mean consultation from 2.0 (1995–1999) to 2.2 encounters per year (2015–2019) clinically relevant, especially given the increased availability of SAR medication without consulting a doctor, and due to increased access to health education, e.g. thanks to patient education website GPinfo (www.thuisarts.nl) launched in 2011.

A low or high average temperature from February till July resulted in a significantly longer duration of the SAR season. This result can be partly explained: a higher temperature leads to earlier spring, and thereby earlier start of the (tree) pollen season and probably a longer SAR season33. A lower temperature results in more simultaneous blooming, which seems to affect pollen concentrations, causing more severe SAR complaints34. Polysensitisation probably contributes4. We can only speculate if a strong initial response somehow triggers sustained complaints, explaining the longer duration of the SAR peak.

These signs of increased severity of SAR (more contacts per year per SAR patient), and the prolongation of the period of the year in which patients present to GPs with symptoms of SAR—including a tendency towards earlier start of SAR seasons—are likely caused by climate change. They result in a higher burden for patients and a higher workload for GPs, and hence are important to patients, primary care clinicians and policy makers.

However, our study had limited statistical power. Therefore additional and larger scale research is needed. The methods we applied could be easily reproduced by other Practice-Based Research Networks that record data on SAR in different regions so that data could be combined.

Data from weather stations from different countries and regions could help to study the association between SAR symptoms and temperature more precisely.

Collaboration between clinicians and climate scientists or biologists is needed to make more direct links between clinical and environmental data, by addition of pollen concentrations, frost-free days and cumulative degree days. For patients to self-manage SAR, and to predict impact on medical services, it could be helpful to provide accurate forecasts. This requires local instead of regional data which is not always available, let alone data of all factors that can influence the pollen season35. Our current study might help to find a method to eventually forecast the pollen season without acquiring to much environmental data.

Climate change has other health effects besides respiratory problems. Heat waves result in heat strokes, renal function problems and cardiovascular disease. Ultraviolet radiation increases skin cancer risk. Additionally, it results in more infections transmitted by water, food and vectors32.

Globally, awareness of the climate crisis and the need for urgent action has massively increased over the last few years. Linking climate change to adverse health outcomes probably helps to make the need to respond to this crisis urgently recognisable for more individual people. The health profession has an obvious role in identifying such associations. Just as happened in the past regarding advancements in sanitation, hygiene and tobacco control, health professionals should and are indeed showing responsibility and leadership19,20.

In this first study linking air temperature to clinical presentation of SAR to GPs, our hypotheses that increased temperature led to a greater impact of SAR on general practice were confirmed: we observed a longer duration of the period in which SAR was presented, with a tendency towards an earlier start of SAR presentation in warmer years, and increased frequency of GP contacts for SAR, suggesting increased severity. Statistical power of this study was limited, justifying careful interpretation. Evidence directly linking climate change to adverse health outcomes is still limited but steadily growing in the past few years. Our results call for additional research to confirm and strengthen our findings, for which our exploratory methods can serve as an example for other PBRNs.