Environmental Drivers of Respiratory Emergency Admissions: The Role of Tropospheric Ozone and Humidity in Lleida, Spain (2010–2019)

Abstract

Background: Tropospheric ozone (O₃) is a secondary air pollutant associated with respiratory morbidity. Lleida is an inland Mediterranean city with a continentalized climate, frequent winter thermal inversions and hot, dry summers, where ozone episodes and high humidity often co-occur under stagnant atmospheric conditions. This study explores the association between air pollutants, weather variables, and respiratory emergency admissions in Lleida, Spain. Methods: We conducted a time-series analysis using distributed lag non-linear models (DLNM) on hospital emergency room admissions for acute respiratory conditions in Lleida (2010–2019). Data on weather (temperature, humidity, solar radiation) and air pollution (O₃, NO₂, PM10, SO₂) were obtained from local monitoring stations. The primary outcome was the daily number of admissions for respiratory conditions (ICD-10 codes J09–J18, J20–J22, J44.1, J45.9). Results: A total of 19,428 respiratory admissions were recorded. High O₃ concentrations and elevated relative humidity were significantly associated with increased admissions, even after adjusting for temperature and solar radiation. The strongest effects were observed with delayed lags (up to 21 days). NO₂, PM10, CO and SO₂ levels did not show a significant association. Conclusions: Our findings support a significant and independent association between elevated ozone concentrations, high humidity, and respiratory emergencies. These results highlight the need for public health strategies and policy interventions focused on environmental risk forecasting and air quality management, particularly in vulnerable inland Mediterranean regions.

Background

Air pollution constitutes one of the most pressing environmental and public health issues, exerting direct effects on human health¹.Air pollutants, including particulate matter (PM2.5 and PM10), nitrogen oxides (NOx), sulfur dioxide (SO₂), carbon monoxide (CO), and volatile organic compounds (VOCs), originate from a variety of sources such as vehicular traffic, industrial activities, fossil fuel combustion, and deforestation². These pollutants have detrimental effects on respiratory health, contributing to the onset and exacerbation of various respiratory conditions, including asthma, chronic bronchitis, and chronic obstructive pulmonary disease (COPD)³.

Numerous studies have demonstrated that prolonged exposure to air pollution significantly increases the likelihood of developing severe respiratory diseases^4–6. A report by the World Health Organization (WHO)⁷. concluded that air pollution is responsible for 7 million deaths annually, many of which are linked to respiratory and cardiovascular conditions. The burden of respiratory diseases, in turn, leads to increased demand for medical care, including a rise in healthcare visits due to respiratory flare-ups or exacerbation episodes, particularly among vulnerable populations such as children, the elderly, and individuals with pre-existing conditions⁸.

Air pollution has also been associated with a higher number of emergency department admissions, particularly during periods of elevated pollutant concentrations⁹. Individuals with pre-existing respiratory conditions are especially vulnerable to exacerbations, often resulting in increased admissions due to asthma attacks or acute respiratory failure. According to various studies, including those published in The Lancet¹⁰. and Environmental International^11,12, hospitals in urban areas with high levels of pollution report admission spikes during these high-pollution episodes, placing additional strain on public health systems. Indeed, a study published in 2022 demonstrated that increases in sulfur dioxide (SO₂) levels represent a significant risk factor for emergency visits due to cardiovascular and respiratory causes, as well as for hospital admissions related to respiratory conditions¹³.

Most previous Mediterranean studies on ozone and respiratory morbidity have been conducted in coastal cities such as Valencia or Cartagena, where sea-breeze circulations facilitate the dispersion of pollutants and the seasonal pattern of humidity differs from that of inland basins.

The climate and geography of Lleida have a direct impact on air pollution. The city lies in an inland basin of the Ebro valley and is surrounded by mountainous topography, which acts as a natural barrier and hinders the dispersion of pollutants, particularly during the winter months. Low temperatures, persistent fog and the absence of wind contribute to thermal inversions that trap pollutants close to the ground and prevent their dispersion into the upper atmosphere¹⁴. In contrast, summers are hot and dry, with intense solar radiation that favours photochemical ozone formation. This combination of frequent winter inversions with high humidity and stagnation, together with summer ozone episodes, creates a distinctive pollutant–weather pattern that differs from coastal Mediterranean settings. These features make Lleida a particularly informative setting to investigate the joint and delayed effects of ozone and humidity on respiratory morbidity.

In addition, agricultural activity and vehicular traffic in the province of Lleida contribute significantly to the emission of atmospheric pollutants. The burning of agricultural residues, particularly during harvest season, releases large quantities of fine particulate matter into the air, increasing pollution levels and directly impacting the health of the local population. Agricultural practices and industrial activity also represent major sources of volatile organic compounds (VOCs) and other polluting gases that further degrade air quality¹⁵.

With regard to climate change, rising summer temperatures in the region may contribute to the formation of tropospheric ozone and higher concentrations of particulate matter in both urban and rural areas, thereby worsening respiratory conditions. The combination of high temperatures, low humidity, and air pollution increases the population’s vulnerability to respiratory problems, particularly among children, the elderly, and individuals with underlying respiratory diseases. The interrelationship between climate, geography, and air pollution in Lleida highlights the urgent need for policy measures aimed at reducing pollutant emissions and promoting public health in the region—especially in the context of climate change, which could further intensify these adverse effects.

This study analyzes the association between air pollution and the incidence of acute respiratory conditions treated at the Emergency Department of a referral hospital in the Lleida region, which serves a broad territory characterized by climatic variability. The aim is to contribute to the understanding of the environmental context’s impact on respiratory health and to provide evidence that may inform the planning of public policies in the fields of health and the environment¹⁶.

Against this background, we aimed to assess the association between air pollutants, meteorological variables and respiratory emergency admissions in Lleida, an inland Mediterranean city characterised by frequent winter thermal inversions and hot, dry summers. Specifically, we sought to disentangle the independent and delayed (up to 21 days) effects of tropospheric ozone (O₃) and relative humidity on respiratory morbidity in this highly susceptible, inversion-prone region, and to compare these findings with those reported in coastal Mediterranean cities.

Methods

Setting

Hospital admissions at emergency room for breathing issues and data on weather and pollutants for the Spanish province of Lleida were collected between 2010 and 2019.

Lleida is a Spanish province with 0.44 million inhabitants. Its so called capital, placed in the south of the province, accounts for around the 32% of the province population. Together with the municipalities of its health region accounts for approximately the 83% of the province population. The city of Lleida and nearby municipalities have a cold semi-arid climate (BSk) according to the Kóppen Climate Classificacion, while the north (more mountainous region) is characterized by temperate, no dry season, hot or warm summers (Cfa-b). At the towns of the Pyrenees north region, climate is cold, with no dry season, warm or cold summers (Dfb-c).

Data

The number of hospital admissions for breathing issues were obtained from the daily counts of non-scheduled hospital admissions registered in the Spanish province of Lleida during the study period according to the CMBD (“Conjunto Mínimo Básico de Datos”) register of the Spanish Ministry of Health. This register, with data available since 1997, is compulsory for all hospitals from the National Health System, even though private hospitals have been included since 2016. Hospitalizations caused by breathing issues included the ICD-10-CM codes: J09-J18; J20-J22; J44.1 and J45.9.

Weather variables included daily (24 h) mean, maximum and minimum of temperature, relative humidity and atmospheric pressure. Daily accumulated precipitation, daily global solar irradiation and daily mean and maximum wind velocity at 2 meters high were also included and analyzed. All these data were obtained from the automatic weather stations network of the Meteorological service of Catalonia. We computed the mean of each weather variable registered by the province capital weather stations (a total of 4) as representative for the whole province given the concentration of the population in the capital.

Air pollutants included the summary measures of the daily (24 h) mean of NOX, PM10, SO2, NO, and NO2, the daily maximum of NOX, O3, SO2, NO, and NO2 and the daily maximum 8 h moving average of CO and O3. All these data were obtained from the atmospheric pollution monitoring and forecasting network of Catalonia (XVPCA). Only air pollutants registed in the city station were taken into account since it is the only station with data on air NO and NO2 concentrations. The European air quality index limits for a poor air quality defined for CO, NO2, O3, PM10 and SO2 were applied. Specifically, a maximum of 10 mg/m3 per hour for CO, a maximum of 120 mcg/m3 per hour for NO2, a maximum of 130 mcg/m3 per hour for O3, a 24 h mean of 50 mcg/m3 for PM10 and a maximum of 350 mcg/m3 per hour for S02.

Statistical analysis

We used distributed lag non-linear models (DLNM) to estimate the relationship between hospital admissions caused by breathing issues and the exposure variables of weather and pollution in order to capture possible delayed effects. We fitted a time-series quasi-Poisson regression model following previous time-series studies (Martínez-Solanas & Basagaña, 2019)¹⁷. The exposure-response association was modelled using a quadratic B-spline of grade 2 with 3 internal knots placed at the 10th, 50th and 90th percentiles of location-specific exposure distribution. Each model took into account seasonality by including a natural cubic spline of the day of the year with 4 degrees of freedom and an interaction between this spline function and the indicator of year to relax the assumption of a constant seasonal trend. Long-term trends were controlled by including a linear term for the day of the study period. In addition, we included indicator variables for the day of the week and for holidays. The lag-response association was modelled by a natural cubic spline with an intercept and up to three internal knots placed at equally spaced values in the log scale. To explore possible short as well as long delays in the effects of exposure variables, the lag period was tested at 3, 7 and extended up to 21 days to capture long lags related to weather or pollution variables.

The results were aggregated over all lags to obtain the overall exposure-response association curves, and were reported as relative risks in reference to the percentile of minimum hospitalizations (minY) estimated from each curve. This minY was replaced by the median or 50^th percentile for minY values below 20 or above 80. These estimated exposure-response curves were used to quantify the percent relative effect of very extreme exposure values, defined as percentiles 1, 2.5, 10, 90, 97.5 and 99). We first fitted single-pollutant time-series models for each air pollutant to explore their association with daily respiratory emergency admissions. Pollutants that did not show evidence of a significant association (NO₂, PM₁₀, SO₂) were not retained as exposure terms in subsequent models. The final distributed lag models therefore focused on tropospheric O₃ and relative humidity, in line with our primary study objective.

In case of significant associations with both weather and pollution variables, we used DLNM to assess if, once adjusted by the weather variable inside the model through an additional crossbasis of the same lag period to allow for effect delays, the exposure to the pollutant keep its significant association with hospital emergency room admissions.

All analyses were performed with R software¹⁸. The package dlnm was used¹⁹.

Ethics

The project was approved by the Research Committee of Lleida with ID 2323. All the research was conducted following Helsinki Declaration. All data was handled according to current European legislation. The researcher accessed it for research purposes on the 13th of September, 2022. The authors had no access to information that could identify individual participants during or after data collection. The research committee waived the requirement for informed consent.

Results

A total of 19428 admissions to hospital emergency rooms due to breathing issues were registered from 2010 to 2019, with annual counts ranging from 1507 to 2494, representing the 2.14% of the total 908.003 admissions. The distribution of daily admissions to hospital emergency rooms showed a mean (standard deviation SD) of 5.3 (3.5), with a median of 5 and ranging from 0 to 46, while the cumulative number by month showed a mean (SD) of 161.9 (63.9), ranging from 30 to 464, and exhibiting the highest median cumulative number of admissions in January (Fig. 1).

Fig. 1. Evolution of Emergency Department visits for respiratory reasons by month of the year, with a decrease during the summer months.

Distribution of daily (A) and monthly (B) hospital emergency room admissions for breathing issues per month, 2010–2019.

The most variable daily weather measure along the study period was the daily cumulative precipitation, with a coefficient of variation of 3.94 and a mean (SD) of 0.973 (3.83) and a 0 median. The second most variable daily weather measure was the 24 h solar irradiation, with a coefficient of variation of 0.5091 and a mean (SD) of 17.05 (8.68). In contrast, the daily 24h-mean temperature showed a coefficient of variation of 0.03 (assessed in Kelvin degrees) and a mean (SD) value of 14.4 (7.58) °C. Among pollutants, the most variable was the 24h-maximum concentration of nitrogen monoxide, with a coefficient of variation of 1.32, much higher than the one for the 24h-maximum concentration of sulphur dioxide, carbon monoxide, nitrogen dioxide, the 24 h mean PM10 concentration or the maximum 8h-moving average of ozone concentration, with coefficients of 0.80, 0.67, 0.54, 0.52 and 0.45 respectively. The 24h-maximum concentration of O3 showed the lowest coefficient of variation, with value 0.39. The mean concentrations for the maximum 24 h levels of SO2, CO, NO2 and O3 and the 24 h mean levels of PM10 were 3.52, 429.6, 50.84, 80.29 and 24.44 mcg/m3, respectively. The concentrations of CO and SO2 did not exceed limits for poor quality on any day between 2010 and 2019, while NO2, O3 and PM10 concentrations were responsible of poor air quality for a total of 326 days along the 10 years (8.93%). Specifically, the PM10 limit was exceeded in 158 days (43 days of them together with NO2 limits and 8 days with O3 limits), the O3 limit was exceeded in a total of 143 days (3.92%), and the NO2 was exceeded in a total of 76 days (2.08%).

In preliminary single-pollutant models, we did not observe statistically significant associations between NO₂, PM₁₀ or SO₂ and respiratory emergency admissions across the lag period considered, so these pollutants were not included as exposure variables in the final O₃–humidity models.

Daily and monthly hospital emergency room admissions for breathing issues per month are shown in Fig. 1 and exhibit a convex curve, with the highest amounts registered at the beginning and the end of the year, or in other words, at the months with the lowest temperatures and solar irradiation values and the highest values of relative humidity and atmospheric pressure, and also the months with more days with poor air quality due to excessive concentrations of nitrogen dioxide. On the other side, poor quality due to excessive ozone concentrations occurred mostly at summer months, with higher temperatures and solar irradiation. The distribution of daily weather measures of temperatures and solar irradiation for the study period is shown in Fig. 2. The ones of NO2 and O3 pollutants is shown in Fig. 3.

Fig. 2. Distribution of daily temperature, relative humidity and solar irradiation measures and monthly cumulated precipitation by month, 2010–2019.

The dashed lines identify the overall mean of each measurement in the colour detailed in the corresponding legend.

Fig. 3. Distribution of daily pollution measures of O3, PM10 and NO2 by month, 2010–2019.

The dashed lines identify the overall mean of each measurement in the colour detailed in the corresponding legend. The red lines identify poor air quality measures as those above red lines. The O3 values show the daily maximum 8 h moving average (ma).

A higher risk of breathing issues was significantly associated with low temperatures, high relative humidities, low solar irradiation and high ozone concentrations, as shown in Fig. 4A, B, C, D respectively for 21 days lag. Days reaching extremely cold temperatures showed a significantly higher risk of hospital emergency room admissions from breathing issues, and so 24h-mean temperatures below the 1th percentile and 24h-maximum temperatures below 10^th percentile showed a significant upward trend, although only for the 21 days lag. In contrast, the pattern from its 90th percentile onwards showed a significant downward trend although exhibiting an upward trend for the extremely hot days. More breathing problems were also associated with high relative humidity and extremely high atmospheric pressures but only for 21 days lag In comparison, a monotone significant association was observed with solar irradiation, and days with extremely low solar irradiation showed significantly higher risks while days with extremely high solar irradiation showed significantly lower risks. Higher concentrations of ozone showed a significant upward trend, although at extremely high ozone concentrations there was a trend inflection. Lower concentrations of ozone showed also a significant upward trend with the same change of trend at extremely low ozone concentrations. Higher concentrations of NO2 or other pollutants did not show any significant higher risk of breathing issues. The percent relative effects for effect delays of 3, 7 or even 21 days are reported in Table 1.

Fig. 4.

Estimated overall cumulative relationship between hospital admissions for breathing issues and exposure to temperature (4a), relative humidity (4b), solar irradiation (4c) and O3 concentrations (4 d) up to lag 21. The O3 values show the daily maximum 8 h moving average (ma).

Table 1.

Percent relative effect associated to weather and pollution exposures.

Percent refered to admissions at the median value of the exposure variable
		10% more extreme values		2.5% more extreme values		1% more extreme values
		Low (P10 [95% CI])	High(P90 [95% CI])	Low (P2.5 [95% CI])	High(P97.5 [95% CI])	Low (P1 [95% CI])	High (P99 [95% CI])
With lag=21
	Temperature (mean)	-7.2 [-24.2, 13.6]	-26.5 [-40.1, -9.7] *	15.1 [-6.9, 42.3]	-21.4 [-36.9, -2.0] *	38.8 [7.0, 80.0] *	-6.4 [-32.8, 27.7]	P50r
	Temperature (max)	30.3 [7.4, 58.2] *	-34.7 [-45.9, -21.2] *	68.3 [33.5, 112.3] *	-33.1 [-45.0, -18.5] *	84.2 [38.0, 145.2] *	-19.1 [-38.4, 6.3]	P50r
	Relative Humidity (mean)	-32.3 [-44.3, -17.6] *	-5.6 [-22.2, 14.7]	-39.5 [-54.8, -19.1] *	39.0 [10.1, 75.6] *	-25.0 [-45.3, 2.9]	60.6 [18.8, 117.2] *	P50r
	Relative Humidity (min)	-9.7 [-19.1, 0.8]	22.2 [-0.8, 50.5]	-1.6 [-21.5, 23.4]	81.2 [38.0, 138.0] *	17.1 [-13.7, 59.0]	99.6 [36.8, 191.1] *	P20
	Pressure (mean)	18.3 [1.1, 38.5] *	-8.4 [-22.3, 8.1]	12.8 [-14.9, 49.4]	8.6 [-9.5, 30.3]	-15.2 [-38.6, 17.1]	38.4 [10.4, 73.4] *	P35
	Pressure (min)	8.2 [-6.5, 25.2]	-17.7 [-28.8, -4.9] *	-3.4 [-24.3, 23.1]	0.4 [-17.2, 21.6]	-26.2 [-44.9, -1.1]*	45.3 [15.3, 83.3] *	P50r
	Pressure (max)	25.1 [8.0, 45.1] *	3.3 [-12.1, 21.3]	38.0 [5.6, 80.4] *	12.4 [-5.8, 34.0]	8.3 [-20.8, 48.1]	38.5 [12.1, 71.0] *	P32
	Solar irradiation (24h)	33.9 [3.0, 74.1] *	-43.9 [-60.2, -21.1] *	48.1 [8.6, 102.1] *	-58.5 [-72.5, -39.6] *	47.2 [2.1, 112.1] *	-63.5 [-77.5, -40.8] *	P50r
	O3 (max 8h ma) vs. P25	24.5 [1.3, 53.1] *	63.3 [30.1, 105.0] *	33.2 [4.7, 69.4] *	18.3 [-9.0, 53.6]	29.2 [-8.4, 82.3]	-11.0 [-37.9, 27.6]	P25
	O3 (max 24h) vs. P25	29.5 [6.6, 57.2] *	61.3 [27.8, 103.6] *	46.0 [15.1, 85.1] *	10.0 [-14.8, 42.0]	38.1 [-1.7, 94.0]	-24.5 [-48.8, 11.2]	P25
With lag=7
	Temperature (mean)	5.7 [-2.8, 15.0]	-13.7 [-29.0, 4.9]	11.2 [-0.7, 24.4]	-11.6 [-28.3, 9.1]	11.3 [-4.2, 29.4]	-0.4 [-23.0, 28.8]	P27
	Temperature (max)	4.0 [-8.4, 18.0]	-23.9 [-33.5, -13.0] *	11.2 [-4.9, 29.9]	-25.2 [-35.7, -13.0] *	10.4 [-8.8, 33.7]	-17.5 [-31.6, -0.4] *	P50
	Relative Humidity (mean)	-12.1 [-21.7, -1.2] *	-5.9 [-16.5, 6.0]	-14.5 [-27.9, 1.5]	9.1 [-5.0, 25.3]	-14.3 [-29.3, 3.8]	19.4 [-0.9, 43.8]	P50
	Relative Humidity (min)	-6.5 [-12.5, -0.05] *	6.0 [-7.0, 20.8]	-10.3 [-21.8, 2.9]	14.7 [-2.3, 34.5]	-7.3 [-21.8, 2.9]	18.6 [-5.1, 48.2]	P20
	Pressure (mean)	-4.7 [-12.8, 4.2]	-10.8 [-18.2, -2.7] *	-8.1 [-21.2, 7.0]	-5.7 [-16.6, 6.7]	-6.0 [-20.8, 11.6]	5.3 [-8.8, 21.6]	P50
	Pressure (min)	-5.6 [-13.4, 2.9]	-12.3 [-19.6, -4.3] *	-9.2 [-21.9, 5.5]	-6.4 [-17.4, 6.0]	-8.4 [-22.6, 8.4]	10.9 [-4.6, 28.9]	P50
	Pressure (max)	-1.7 [-9.7, 7.0]	-8.1 [-15.5, -0.0] *	-0.7 [-14.4, 15.1]	-7.2 [-17.8, 4.8]	-1.6 [-17.1, 16.7]	2.4 [-10.9, 17.7]	P50
	Solar irradiation (24h)	12.3 [-4.2, 31.6]	0.7 [-18.4, 24.1]	14.4 [-3.1, 35.1]	-15.4 [-33.5, 7.7]	13.6 [-6.8, 38.3]	-24.5 [-43.5, 0.8]	P43
	O3 (max 8h ma) vs. P27	14.9 [0.9, 30.7] *	21.5 [4.4, 41.3] *	10.5 [-5.6, 29.4]	7.9 [-9.5, 28.6]	3.9 [-17.7, 31.3]	-1.9 [-22.7, 24.5]	P50
	O3 (max 24h) vs. P28	16.6 [3.1, 31.8] *	21.5 [4.8, 40.8] *	11.9 [-4.4, 30.8]	0.0 [-15.9, 18.9]	2.8 [-18.3, 29.2]	-16.7 [-36.2, 8.7]	P50
With lag=3
	Temperature (mean)	2.2 [-3.1, 7.8]	-5.6 [-20.5, 12.2]	1.9 [-6.3, 10.8]	-7.2 [-22.9, 11.6]	-1.0 [-12.0, 11.5]	-1.0 [-20.4, 23.2]	P22
	Temperature (max)	-2.5 [-12.1, 8.2]	-18.6 [-27.5, -8.5] *	0.5 [-11.5, 14.1]	-21.4 [-31.3, -10.0] *	0.5 [-14.1, 17.5]	-17.2 [-29.5, -2.9] *	P50
	Relative Humidity (mean)	-7.1 [-15.0, 1.5]	1.5 [-7.4, 11.3]	-5.6 [-17.0, 7.3]	5.7 [-5.4, 18.0]	-5.7 [-18.9, 9.5]	9.8 [-4.9, 26.8]	P50
	Relative Humidity (max)	-13.5 [-20.8, -5.5] *	1.2 [-8.3, 12.8]	-16.4 [-17.6, -3.4] *	-6.9 [-18.5, 6.3]	-10.5 [-24.3, 5.7]	-12.9 [-29.0, 6.9]	P50
	Pressure (mean)	-4.6 [-10.7, 1.9]	-7.8 [-13.8, -1.3] *	-2.0 [-12.7, 9.9]	-9.0 [-17.8, 0.8]	0.5 [-11.7, 14.5]	-1.8 [-13.0, 10.9]	P50
	Pressure (min)	-4.9 [-10.8, 1.4]	-8.4 [-14.5, -1.9] *	-3.6 [-13.9, 7.9]	-9.2 [-18.1, 0.6]	-3.1 [-14.7, 10.2]	0.8 [-11.1, 14.4]	P50
	Pressure (max)	1.7 [-3.4, 7.1]	-2.4 [-9.8, 5.6]	5.9 [-5.8, 19.2]	-5.7 [-14.2, 3.7]	5.8 [-7.9, 21.5]	0.7 [-10.6, 13.4]	P20
	Solar irradiation (24h)	9.1 [-3.5, 23.4]	-4.1 [-17.1, 11.0]	15.7 [1.1, 32.5] *	-9.9 [-23.5, 6.1]	17.5 [0.2, 37.8] *	-13.6 [-29.2, 5.5]	P50
	O3 (max 8h ma) vs. P30	11.8 [0.6, 24.3] *	13.9 [1.3, 28.0] *	0.5 [-11.1, 13.6]	7.0 [-7.0, 23.0]	-8.1 [-22.6, 9.0]	0.7 [-16.4, 21.3]	P50
	O3 (max 24h) vs, P30	13.2 [2.6, 24.8] *	13.8 [1.5, 27.5] *	2.5 [-9.3, 15.6]	1.8 [-11.5, 17.0]	-7.2 [-21.5, 9.8]	-8.9 [-26.1, 12.1]	P50

Weather and pollution variables not included showed no significant higher risk association with response.

*Significant association.

The increase in hospital admissions for breathing issues associated with the exposure to daily high concentrations of O3 (measured by the daily maximum 8 h moving average) was still significant after having adjusted by daily mean temperatures, daily mean relative humidities, daily solar irradiation or even by other pollutants exceeding healthy limits such as PM10 or NO2, as shown in Fig. 5A, B, C, D respectively. Furthermore, when adjusted by solar irradiation and relative humidity, only higher values of ozone were associated with a higher risk of hospital admissions for breathing issues. Extremely high concentrations of ozone did not show higher levels of risk.

Fig. 5.

Estimated overall cumulative relationship between hospital admissions for breathing issues and exposure to O3 concentrations up to lag 21 and adjusted by 24 h mean temperature (5a), 24 h mean relative humidity (5b), 24 h solar irradiation (5c) and pollutants exceeding air quality limits (PM10 and NO2). The O3 values show the daily maximum 8 h moving average (ma).

Discussion

Elevated concentrations of tropospheric ozone and high relative humidity levels were found to be significantly associated with increased emergency department (ED) visits and hospital admissions for acute respiratory conditions in Lleida. These associations were evident even after adjusting for confounding meteorological variables such as temperature and solar radiation, suggesting an independent effect of ozone on respiratory morbidity.

Tropospheric ozone (O₃) is not directly emitted into the atmosphere but is formed by photochemical reactions involving nitrogen oxides (NOₓ) and volatile organic compounds (VOCs) under the influence of solar radiation. This makes regions like Lleida particularly susceptible, due to their topography and climatic characteristics—especially during summer months (May to September), when temperature and sunlight are at their peak, and ozone levels frequently exceed both WHO recommendations and EU legal thresholds.

To our knowledge, only a limited number of time-series studies have jointly modelled the independent and delayed effects of both tropospheric O₃ and relative humidity on respiratory morbidity, and almost all of them have been conducted in large coastal cities. By focusing on an inland Mediterranean basin with frequent thermal inversions and marked seasonal contrasts, our study contributes novel evidence on how ozone and humidity exert both short- and longer-term impacts on respiratory emergency admissions.

Our results align with previous studies conducted in similar Mediterranean inland areas. For instance, Tenías et al²⁰. found a 6% increase in COPD-related ED visits in Valencia for every 10 µg/m³ rise in ozone levels. Similarly, Cirera et al²¹. observed significant associations between ozone concentrations and asthma or COPD exacerbations in Cartagena, a city with comparable climate and air quality patterns. With ongoing climate change, the frequency of extreme weather conditions and high ozone events is expected to increase, particularly in Mediterranean inland areas. This underlines the urgency of adaptive health system planning

Regarding humidity, our findings suggest a complex interrelationship with ozone. While low relative humidity in dry summer months can enhance photochemical ozone formation (by reducing the attenuation of UV radiation), very high humidity levels—as observed in autumn and winter in Lleida—may also affect pollutant dispersion and, given its higher concentration, enhance ozone’s respiratory impact. Previous literature has highlighted that high humidity may impair pulmonary function in individuals with pre-existing respiratory conditions and increase susceptibility to infections^22,23. The dual role of humidity—enhancing ozone formation in dry conditions and limiting pollutant dispersion in saturated air—may explain the non-linear patterns observed.

From a mechanistic perspective, ozone is a potent oxidant capable of inducing epithelial injury, increasing airway permeability and triggering inflammatory responses in the respiratory tract. High relative humidity may further influence respiratory defence mechanisms by altering mucociliary clearance, promoting mucus accumulation and modifying the deposition of inhaled particles and soluble gases in the airways. In individuals with pre-existing respiratory diseases, the combination of epithelial damage caused by ozone and humidity-related impairment of mucociliary function could facilitate the penetration of pollutants and pathogens, increase susceptibility to infections and precipitate acute exacerbations. These pathophysiological pathways provide a plausible biological explanation for the joint impact of ozone and high humidity observed in our study.

A global meta-analysis by Liu et al²⁴ confirmed a consistent relationship between short-term ozone exposure and hospital admissions for respiratory diseases, although noting geographic variability influenced by climate, population characteristics, and pollutant mixtures. Likewise, Linares et al²⁵ emphasized the modifying role of humidity and temperature on the respiratory effects of air pollution, highlighting that ozone-related admissions are higher in conditions of heat and stagnant air.

The exposure–response curve for ozone showed a non-linear pattern, with an increase in risk at high concentrations but also a suggestion of elevated risk at very low O₃ levels. Several mechanisms may contribute to this apparent “J-shaped” relationship. Days with very low ozone concentrations in our setting often occur under different atmospheric regimes, for example during winter periods with intense NO scavenging of ozone and higher levels of primary traffic-related pollutants, or during stagnant conditions when people spend more time indoors and are more exposed to indoor pollutants and respiratory infections. In such circumstances, low ambient ozone may act as a marker for a different pollutant mixture or for increased susceptibility related to co-circulating viral infections, rather than indicating a protective effect of ozone itself. In addition, the number of days at the extremes of the ozone distribution was relatively small, leading to wider confidence intervals and more unstable estimates. For these reasons, the inflection at very low concentrations should be interpreted cautiously and further investigated in larger multi-city datasets.

Our study adds to this evidence by demonstrating that in Lleida—a city with a continentalized Mediterranean climate and frequent atmospheric inversions—both ozone and humidity are important contributors to respiratory morbidity. The robustness of the association even after adjusting for solar radiation and temperature supports the independent role of ozone. Studies from other regions with comparable characteristics—such as the Ebro Valley²⁶ or inland areas of Italy²⁷—have similarly reported associations between elevated ozone concentrations and an increased incidence of respiratory conditions in both paediatric and adult populations.

Strenghts and Limitations

First, the ecological design of this study prevents causal inference at the individual level and is susceptible to ecological fallacy. Second, we lacked access to individual-level data on relevant covariates such as smoking status, occupational exposures, medication use, or underlying respiratory conditions, which could act as unmeasured confounders. Third, socioeconomic and environmental factors such as housing quality, indoor air pollution, and differential access to healthcare were not included, despite their potential influence on respiratory morbidity. Fourth, diagnostic classification relied on hospital discharge codes (ICD-10), which may vary in accuracy across settings and over time. Fifth, our analysis relied on daily aggregated counts of respiratory hospital admissions and did not include individual-level information on key determinants such as housing conditions, recent infections, smoking status or which vulnerable groups were most affected. As a result, we could not quantify the proportional contribution of these factors to respiratory morbidity during the lag period, and residual confounding by these variables cannot be ruled out. Future studies incorporating individual-level information are needed to address these questions. Finally, seasonal patterns of respiratory infections—particularly influenza and other viral epidemics—could partially account for some of the observed variations in hospital admissions, despite adjustments for seasonality in the statistical models.

Despite the ecological nature of our study and the lack of individual-level information, several aspects support the plausibility of the observed associations. First, our findings are consistent in direction and magnitude with previous time-series and DLNM studies that have reported increased respiratory morbidity associated with short-term ozone exposure in Mediterranean and other climatic settings. Second, the association between ozone and respiratory emergency admissions remained robust after adjustment for temperature, solar radiation and, in sensitivity analyses, for NO₂ and PM₁₀, which reduces the likelihood that our results are entirely driven by confounding by these factors. Third, the lag pattern and non-linear exposure–response relationship are compatible with the known pathophysiological effects of ozone and high humidity on the respiratory tract, including oxidative damage, airway inflammation and impaired mucociliary clearance. While residual confounding by unmeasured factors cannot be ruled out, these elements strengthen the argument that the associations we describe reflect a real underlying effect rather than purely artefactual correlations.

Despite these limitations, our findings support the growing body of evidence on the adverse respiratory effects of air pollution—particularly tropospheric ozone—and the need to consider meteorological parameters like humidity in public health planning. These results have direct implications for local policy, particularly in designing early-warning systems, enhancing environmental surveillance, and implementing air quality control measures. Timely public health interventions could reduce ED burden and improve respiratory health outcomes, especially among vulnerable populations such as children, the elderly, and those with chronic lung diseases.

These findings are relevant not only for hospital-based care, but also for primary and community health professionals, who can anticipate increased demand or patient deterioration during periods of high ozone and humidity. Integrating environmental risk forecasting into chronic respiratory disease management may improve prevention and reduce acute episodes. From a public health perspective, our findings suggest that local early-warning systems should pay particular attention to days when ozone levels exceed high percentiles of the local distribution, for example above the 90th or 97.5th percentile of the O₃ maximum 8-hour moving average, especially when combined with high relative humidity and stagnant atmospheric conditions. On such days, targeted communication to patients with chronic respiratory diseases, reinforcement of preventive medication and the organisation of healthcare resources could help mitigate the impact on emergency services. In practical terms, integrating real-time forecasts of ozone and humidity into risk prediction models, and defining alert thresholds based on local high-percentile values, could support climate-sensitive health planning, particularly in inland regions prone to thermal inversions and extreme ozone episodes. In settings such as Catalonia, where environmental warning systems already exist, this integration could be implemented within current alert frameworks with relatively minor adaptations.

Conclusions

This study provides robust epidemiological evidence of a significant association between elevated tropospheric ozone concentrations and high relative humidity with increased emergency hospital admissions for acute respiratory conditions in a Mediterranean inland city. These findings confirm and extend previous literature by highlighting the independent contribution of both environmental pollutants and meteorological factors—particularly humidity—to respiratory morbidity.

The results support the need for integrated air quality and public health policies, including the development of real-time environmental alert systems and the incorporation of meteorological parameters into risk prediction models. Moreover, they underscore the importance of climate-sensitive health planning, especially in vulnerable regions affected by extreme weather events and increasing ozone episodes.

Strengthening early warning systems, environmental surveillance, and community-level preparedness could reduce the burden on emergency departments and improve respiratory health outcomes, particularly among high-risk groups such as children, older adults, and patients with chronic respiratory diseases.

Author contributions

MM and CL wrote the main manuscript text and MM prepared the figures. EJ and JO collected all the data and created the database. OY reviewed all the project. All authors reviewed the manuscript.

Data Availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at IRBLLEIDA.

Competing interests

The authors declare no competing interests.

CONSENT FOR PUBLICATION

This manuscript does not contain any individual person’s data in any form (including individual details, images, or videos), and therefore consent for publication is not applicable.