Changes of public behavior alter weather-dependent strain on emergency medical services

As the healthcare system advances and expands in its services, the challenges of remaining efficient become more important. Emergency medical services (EMS) are vital cornerstones of communities. In many countries, EMS is available for every individual, regardless of their social or insurance status, via a toll-free telephone number. Understanding the risk factors for busy days in EMSs might be helpful for improving the allocation of resources, which is the key to better care for all patients in the prehospital setting.^[1] An important factor influencing ambulance call volume could be the interplay of public behavior and weather.

The COVID-19 pandemic has led to far-reaching changes in social life. Keeping physical distance from others and wearing a face mask in public places was perceived to be troublesome by many individuals.^[2] People were forced to actively change their professional and private lives due to a generalized fear of disease transmission. This state of persistent discomfort leads to the exacerbation or onset of physical and mental diseases.^[3] Similar social dynamics have been observed during outbreaks of other infectious agents, such as seasonal influenza, measles, Ebola, wars, or economic crises.^[4-8] Children, the elderly, and people living in poverty might be especially susceptible to the resulting negative effects due to their limited coping resources.^[9] In contrast, the new options of working remotely and in home offices might have led to less road traffic and, therefore, fewer accidents.

Aside from implications for individuals, such a crisis has a severe impact on public health in general and emergency medicine in particular.^[10,11] In addition to the consequences for healthcare professionals, patients have often been more reluctant to use medical resources since the beginning of the pandemic.^[12,13] Many people lost their jobs and, therefore, insurance coverage for themselves and their families.^[12] The anxiety of costly bills or of contracting the disease might have kept patients away from primary care providers and emergency departments.^[14] Other systems were overwhelmed by the suddenly changing demands of the population they served, combined with staffing challenges.^[10,11] These circumstances, i.e., radical changes in public behavior due to a social stressor, might also affect EMS call volume.^[15]

The current literature provides many different approaches for predicting ambulance calls and optimizing resource allocation.^[16-19] However, uninfluenceable external circumstances, such as weather, pose challenges to these concepts.^[20-22] Essential attributes of weather are temperature, wind, precipitation, and fog.^[23] Prior data suggest that very high or low temperatures might be associated with cardiovascular emergencies.^[24,25] In this context, apparent temperature might better reflect physiologic stress than crude values.^[24,25] Strong winds, snow, and rain also have the potential to cause direct (i.e., injuries) or indirect harm (e.g., through power outages) to humans.^[21] The associations among the abovementioned weather-related risk factors, a social crisis leading to drastic changes in public behavior, and medical 9-1-1 calls have not yet been investigated.

Rhode Island is a suitable model region for EMS research. First, it is the smallest state within the United States. Its approximately one million inhabitants live in urban and suburban areas around the centrally located capital, Rhode Island.^[26] Second, all EMS data are stored in the state’s EMS database, which is monitored and maintained by the Rhode Island Department of Health, in the standardized format of the National EMS Information System (NEMSIS).^[27] ImageTrend is the largest provider of ambulance data in the state.^[28] Third, the sociodemographic characteristics of Rhode Island’s population are around the national average of the US.^[28] Rhode Island provides prehospital aid for every individual in the state via the toll-free telephone number 9-1-1. March 1^st, 2020 can be considered the beginning of the COVID-19 pandemic in Rhode Island, with the first cases being diagnosed.^[29] Only a few days later, the governor officially declared a state of emergency.^[30] Rhode Island is located in a temperate climate zone.^[31]

Our aim is to investigate the associations of EMS call volume with weather-related risk factors under the influence of the public stressor of the COVID-19 pandemic. The models developed might be helpful tools to anticipate increased ambulance demand.

This study is a retrospective study of all primary 9-1-1 emergency ambulance responses in the state of Rhode Island from January 1^st, 2018, to August 31^st, 2022. Data were downloaded from the state’s EMS database via ImageTrend. Duplicate records, reports without a unique identifier, and missions not dispatched via 9-1-1 were excluded.

We examined the association between the outcome of daily numbers of ambulance calls and the following meteorological risk factors under the social risk factor of the COVID-19 pandemic: precipitation (rain and snow), fog, thunderstorms, and apparent temperature. The respective data were provided by the National Oceanic and Atmospheric Administration (NOAA).^[32] Information on relative air humidity, which was used to calculate apparent temperature, was downloaded from the Automated Surface Observing System (ASOS) network via the homepage of Iowa State University.^[33] All weather data were collected at the Rhode Island International Airport station, which is close to the state’s geographical center. The presence of snow, rain, fog, or thunderstorms during an observation day was considered positive for the entire day.

The US Census Bureau provided statewide population data.^[34] We controlled for the population at risk on an annual basis. March 1^st, 2020, was considered the onset of the COVID-19 pandemic.

We computed the apparent temperature from the daily average wind speed, average crude temperature, and average relative air humidity using the Magnus formula and the Steadman equation for outdoors in the shade.^[35,36] The implementation of a combined formula for the setting of the present study can be found in Supplementary file 1. The apparent temperature over the observation period was broken down into quintiles: values in the first quintile were defined as low, those in the fifth quintile were considered high, and those in between were considered the baseline values.

In the first step, we developed separate Poisson regression models using EMS call volume as the dependent and individual weather-related risk factors (low and high apparent temperatures, fog, rain, snow, and thunderstorms) as independent variables for the periods before and since the COVID-19 pandemic. In the second step, the same constellations of variables were used to calculate the P-values for the interactions between the pandemic and weather-related risk factors.

To account for possible changes in the EMS system (e.g., staffing, structural modifications), we controlled for the year of the call in all regression models. A day served as the unit of analysis. We report our regression results as incident rate ratios (IRRs) with their corresponding 95% confidence intervals (CIs) and their respective P-values. Changes in call volume were then calculated from the median number of calls per day per 100,000 inhabitants, using the IRR and its 95% CI as the measure of dispersion. We considered P-values <0.05 statistically significant. We used Microsoft Excel 16.62 (Microsoft Corporation, USA) for data curation and Stata SE 17.0 (Stata Corporation, USA) for data analysis and graphical depiction.

We did not undertake formal sample size planning because of the lack of previous weather data as a risk factor for EMS calls during a public crisis. Instead, we chose to include all ambulance records throughout the study period in our analysis. In light of the rule that at least ten observations are needed for every covariable investigated in a multivariable regression model, we expected our analyses to reach sufficient power.

Our protocol was approved by the institutional review board of the Rhode Island Department of Health, with an exemption from full review (vote #2022-17). We conducted our study in accordance with the principles of the Declaration of Helsinki. The RECORD statement^[37] for this manuscript is provided in Supplementary file 2.

We identified 695,535 records falling within the study period of 1,704 days. The median rates of EMS calls per day were 38 (IQR 35.2-40.1) and 39 (IQR 35.8-41.5) calls per 100,000 inhabitants before and since the beginning of the COVID-19 pandemic, respectively. Supplementary file 3 provides the detailed daily characteristics over the study period. The descriptive statistics of low, baseline, and high apparent temperatures can be found in Supplementary file 4.

Poisson regression models for the period before the pandemic showed slightly fewer calls on days with low apparent temperatures (decrease in call rate by 1.2 calls per 100,000 inhabitants, 95% CI: 0.9-1.6) and snow (decrease in call rate by 1.2 calls per 100,000 inhabitants, 95% CI: 0.8-1.5). High apparent temperatures (increase in call rate by 3.6 calls per 100,000 inhabitants, 95% CI 3.2-4.0), and thunderstorms (increase in call rate by 2.1 calls per 100,000 inhabitants, 95% CI: 1.5-2.7) were associated with increased call volumes.

The effects of low apparent temperatures (decrease in call rate by 2.2 calls per 100,000 inhabitants, 95% CI: 1.9-2.6) and snow (decrease in call rate by 1.9 calls per 100,000 inhabitants, 95% CI: 1.5-2.2), have been more pronounced since the beginning of the pandemic. High apparent temperatures (increase in call rate by 2.6 calls per 100,000 inhabitants, 95% CI: 2.3-2.9) and thunderstorms were again associated with higher call volumes (increase in call rate by 1.5 calls per 100,000 inhabitants, 95% CI: 1.0-2.0). We found significant interactions between the pandemic and extreme apparent temperatures (low: P=0.001, high: P<0.001) as well as snow (P=0.009).

Table 1 provides information on the interaction between the pandemic and weather-associated risk factors expressed as IRR. Figure 1 displays our findings regarding AT graphically.

Table 1.

Association between emergency medical service calls and weather-dependent risk factors (adjusted for the year of the call)

Weather-dependent risk factor	IRR (95% CI)		P-value for interaction
	Before pandemic	Since pandemic
Low apparent temperature	0.97 (0.96-0.98)	0.94 (0.93-0.95)	0.001
High apparent temperature	1.10 (1.09-1.11)	1.07 (1.06-1.07)	<0.001
Snow	0.97 (0.96-0.98)	0.95 (0.94-0.96)	0.009
Rain	1.00 (0.99-1.00)	0.99 (0.98-1.00)	0.150
Fog	1.01 (1.00-1.02)	1.00 (1.00-1.01)	0.100
Thunderstorms	1.05 (1.04-1.07)	1.04 (1.03-1.05)	0.070

IRR: incidence rate ratio; 95% CI: 95% confidence interval.

Figure 1. Association between apparent temperature (AT) and daily ambulance calls before and since the beginning of the COVID-19 pandemic. The dashed grey line indicates median values.

Our data suggest an association between extreme apparent temperatures and weather phenomena in combination with changes in public behavior due to a pandemic and the daily number of EMS calls. We believe that our findings from a comprehensive statewide model developed by a multidisciplinary team are generalizable to other settings with continental and temperate climate conditions globally.

According to our data, up to 10% of EMS calls may be associated with high apparent temperatures on hot days, which in turn make up 20% of the study period. This number corresponds to a total of 38 calls per hot day in Rhode Island, i.e. approximately 2,800 calls per year. Assuming that a team of two prehospital clinicians would be able to respond to 10 calls over a 12-hour shift, this sums up to almost 600 possibly heat-associated shifts for one person. Considering that the median hourly wage for EMS personnel in the US of around US $ 18,^[38] the expenditures for EMS missions contributable to high apparent temperatures in Rhode Island are US $ 13,000 per year per 100,000 inhabitants, solely for salaries. Aside from the economic perspective, even relatively small increases in EMS call volume might substantially contribute to the decompensation of a healthcare system during a crisis.

Prior research describes increased rates of certain health problems, such as cardiovascular diseases, during extreme weather conditions.^[39,40] The respective models from studies pre-COVID-19 focused either on single types of diseases or on emergency department visits and not on EMS calls in general. The threat imposed by a public stressor might force people not to leave their houses, even in an emergency.^[41] Furthermore, employees may have worked from home rather than commuting to work since March 2020. These circumstances might have prevented them from being exposed to dangerous weather conditions, the hazards of road traffic, risky hobbies, and human beings outside their home environment as potential vectors for infectious diseases.^[42] This theory is supported by the fact that we found fewer EMS missions on days with snow and low temperatures when people might choose to stay at home if they have the option to do so. Research from the police also indicates that fewer emergency calls were placed, especially during the initial phase of the pandemic.^[42,43] Furthermore, more people have died in the field, possibly because they did not receive sufficient first aid.^[42]

Aside from the direct impact of weather, the association between testing positive for an infectious disease and ambulance call volume might be important. Individuals who are aware of a contagious illness might be more reluctant to call an ambulance due to the fear of infecting others or of being admitted in isolation. In a previous study, we were able to show that EMS calls due to potential infectious diseases correlate with subsequent positive COVID-19 testing.^[44] As weather also seems to be a risk factor for EMS calls during the surge of an infectious disease, it might, therefore, also be associated with the number of people diagnosed with the illness.

Our study has several limitations. First, our model includes all 9-1-1 ambulance calls throughout Rhode Island from the most prominent data provider in the state. Calls from other sources and those without a unique identifier might have been missed. However, we estimate that these account for only approximately 5% of all missions. Second, our model is based on the daily characteristics of the respective day. Analysis on a more granular level (e.g., hourly or with time lags of hours to days after the incidents) might reveal different findings. Third, our analyses did not take demographic characteristics such as patient age and occupation into account. Finally, the weather data we used were collected at a single location. However, as the state’s area is small, we believe this to be a minor issue.

Our findings suggest an association between high apparent temperatures and higher EMS call rates. Low apparent temperatures and snow were associated with fewer calls. The COVID-19 pandemic has influenced the effect of weather on ambulance call rate. Weather is often predictable well in advance. One way to optimize an EMS system could be tailoring resource allocation to the individual demands of climate and weather conditions expected after receiving weather reports in the near future. In the long run, global warming could be associated with a constant increase in EMS demand over the next decades.

Funding: None.

Ethical approval: Institutional review board (IRB) approval was obtained for all aspects of this study in accordance with institutional policies (vote #2022-17).

Conflicts of interest: The authors declare that they have no conflicts of interest.

Author contributions: conceptualization: CK, JR, HH, KW; data curation: CK, JR; formal analysis: CK, HH; investigation: all authors; methodology: CK, JR, HH, KW; project administration: CK, JR, KW; resources: CK, NR, IA, EB, JR, HH, FA, KW; software: CK, JR, FA; supervision: CK, JR, HH, KW; validation: CK, JR, HH, FA, KW; visualization: CK; writing - original draft preparation: CK; writing - review & editing: all authors.

All the supplementary files in this paper are available at http://wjem.com.cn.