The study was conducted in the city of Palma de Mallorca in the Balearic Islands, in the Mediterranean Sea, where the climate is typically Mediterranean with temperate temperatures in winter and characteristically hot and humid summers, although precipitations are scarce and irregular. The healthcare area under study was the Ponent area, with a population of 346,834 inhabitants. Because the local economy is primarily based on tourism, important seasonal changes in population numbers occur and the total population may increase by up to 74% during the summer months. In accordance with the Spanish National Health System regulations, the current healthcare model is based on a universal public healthcare system and the study was carried out in that context.

Data from daily individual emergency admissions to Hospital Universitario de Son Espases (previously named Hospital Universitario Son Dureta until 2011) from 1 January 2005 to 31 December 2021 as well as data from daily individual admissions to critical care units (CCUs) and emergency department admissions from 1 January 2016 to 31 December 2021 were collected from the hospital discharge minimum dataset (Conjunto Mínimo Básico de Datos de las Altas Hospitalarias). The total number of daily emergency admissions (total admissions to hospital), admissions to CCUs, and emergency department admissions (admissions to emergency department) were also collected from this dataset. Information on patient sex, age, and postal code (PC) for each admission was obtained to derive the total number of daily emergency admissions by sex (man or woman), age (stratified into four groups: < 15 years, 15–65 years, 66–80 years, and > 80 years), and PC. PC is a unit of administrative division of the population, and it was used as a reference to analyze emergency admissions by population type (urban vs. rural) and average gross income (AGI). Admission PCs were classified as “urban” or “rural” based on the definition used in the 2020 report on Spanish Urban Areas of the Ministry of Transport, Mobility, and Urban Agenda (Ministerio de Transportes 2021). Data on AGI by PC were obtained from the Spanish Tax Agency’s official web site, which provides personal income tax statistics by PC for the larger municipalities (AgenciaTributaria). Palma de Mallorca’s PCs were divided into three subgroups (low income, high income, and middle income), with 25% of PCs having a low AGI, 25% a high AGI, and the remaining 50% a middle AGI.

Daily individual admissions stratified by primary and secondary diagnosis of diseases for which the impact of heat waves had previously been described in other studies were also collected based on the ICD-9 coding system (until 2015) and the ICD-10 (from 2016 to 2021). Admissions for specific diagnoses included the following six nosologic groups: heat stroke, cardiovascular diseases (acute myocardial infarct, angina, heart failure (HF) exacerbation, arrhythmias, pulmonary thromboembolism (PTE), deep venous thrombosis (DVT), syncope, hypotension, hypertension crisis, hypertension), pulmonary diseases (chronic obstructive pulmonary diseases [COPD] exacerbation, asthma exacerbation, acute respiratory distress syndrome [ARDS]), acute respiratory failure), neurological diseases (ischemic stroke, hemorrhagic stroke, seizure, Parkinson disease, multiple sclerosis), reno-urological diseases (acute kidney injury [AKI], chronic kidney disease (CKD) exacerbation, upper urinary tract stones), and infectious diseases (gastrointestinal infections, pneumonia). Data were processed to obtain the total daily number of admissions by primary diagnosis. However, we decided to include in our analysis diseases that can be underrepresented at admission, such as “exacerbation of COPD” for which the primary diagnosis may be recorded as “pneumonia” while the code indicating “exacerbation of COPD” may be recorded as secondary diagnosis, irrespective of whether they were recorded as primary or secondary diagnosis. These were HF exacerbation, PTE, hypertension, hypotension, hypertension crisis, COPD exacerbation, asthma exacerbation, ARDS, CKD exacerbation, Parkinson disease, and multiple sclerosis.

Maximum, minimum, and mean daily temperatures in Palma de Mallorca (Palma-Puerto weather station) were obtained from the OpenData system of the Spanish Meteorological Agency (Agencia Estatal de Meteorología—AEMET) from 1 January 2005 to 31 December 2021. Daily data on suspended particles (PM10, inhalable particulate matter with diameters between 10 µ and 2,5 µm) and tropospheric ozone (O3) concentrations in Palma de Mallorca (Palma-Foners station) from 1 January 2005 to 31 January 2021 were obtained from the Climate Change and Atmosphere Service (Servicio de Cambio Climático y Atmósfera) of the Regional Ministry of Energy Transition, Productive Sectors, and Democratic Memory (Consejería de Transición Energética, Sectores Productivos y Memoria Democrática). Atmospheric pollution indicators (PM10 and O3) have been considered as confounding factors in previous studies and were therefore adjusted in the statistical analysis.

Currently there is no agreed definition of heat wave (HW) in the scientific community, but there are a series of recommendations to develop a definition (Cicci et al. 2022; Xu et al. 2016; Honda et al. 2014). On the one hand, no significant differences have been found between the use of mean temperature (Tmean) or maximum temperature (Tmax) as a reference unit, although it is advisable to apply a temperature threshold for which an impact on the health of the study population has been evidenced in previous analyses. On the other hand, there is a certain degree of consensus in similar studies to apply a minimum period of 2 consecutive days of temperatures above the threshold to define a period as a HW.

In line with these recommendations, we defined a HW as a level of heat starting at 1 °C below the temperature threshold for the Balearic Islands at which there is a statistically significant increase in mortality. This threshold was determined in the study by Díaz Jiménez, et al. on trigger threshold temperatures for mortality attributable to heat in Spain during the 2000–2009 period (Díaz Jiménez J 2015). This study is currently considered as the consensus study for further research on the heat-related health impact in Spain. We decided to lower the temperature threshold by 1 °C to increase the number of HW events and thus obtain a more homogeneous distribution of these events during the study period. Therefore, in our study, a HW was defined as a period ≥ 2 consecutive days with Tmax ≥ 35 °C.

Patients were considered to have been admitted for exposure to a HW when they were admitted to Hospital Universitario de Son Espases during one of the periods defined as HWs (based on the definition described above) or until 7 days (inclusive) following a HW episode. This was to ensure a time window from the onset of pathophysiological effects secondary to high temperatures to patient admission, with a definitive primary diagnosis. This phenomenon is called “lag effect” (Li et al. 2015; Honda et al. 2014).

The data collected from the daily admissions to Hospital Universitario de Son Espases (previously called Hospital Son Dureta) during the period 2005–2021, both inclusive, were analyzed. A quasi-Poisson distribution was used with a generalized linear model (GLM) to estimate relative risks (together with their 95%CIs) associated with HWs compared to periods of no-HWs for each study variable separately. The analysis was performed using the programming language Rv4.2.2 and was as follows for each study variable:

Quasi-Poisson (X)—> glm(αf ~ HWf + fechaf + o3f + pm10f, data = Y, family = quasipoisson(link = “log”)).

where X is the study variable, αf is the number of admissions with the study variable on day f, and HWf is an independent binary variable identifying if day f is within a HW or a no-HW period (1 and 0, respectively). The factor fechaf is an independent ordinal variable identifying the day, month, and year of the day f to control for changes in trends in admissions over time. Additionally, o3f and pm10f are independent quantitative variables representing the level of pollution on day f based on the concentration of O3 and the concentration of suspended particles in the air in micrograms per cubic meter (μg/m3). Thus, the analysis was adjusted to these parameters that were assumed to act as confounding factors. Lastly, data = Y provides information on the source from which the data had to be extracted, and the function family = quasipoisson identifies the distribution to which the analysis was adjusted. All statistical tests were two-sided and values of p < 0.05 were considered statistically significant.

Study variables were the items already described above: total emergency admissions, total CCU admissions, emergency department admissions, sociodemographic variables, and admissions with specific diagnoses. As mentioned, the sociodemographic variables recorded (sex, age, and PC) were divided into 2, 4, and 5 study variables respectively (man and woman for sex; < 15 years, 15–65 years, 66–80 years, and > 80 years for age; low income, middle income, high income, urban, and rural for PC).