“One day man will have to fight noise as fiercely as cholera and pest” - Robert Koch (1919)

Summary

TLDR: A bounty was placed trying to assess the annual impact of emergency siren noise pollution on well-being globally^[1], and this is my attempt to answer it: roughly 24,341 DALYs lost. A separate (and much rougher) calculation indicates siren noise pollution may cause at least an annual excess 92 deaths each year.^[2]

Epistemics: This is the result of roughly eight hours of research. There’s a lot of assumptions here, and I’ve tried to be explicit in marking them so you understand the weaknesses in my BOTEC and where it could be strengthened, so please see the conclusion as uncertain.

Housekeeping:

• Assumptions
  • [super major assumption] - means I’m taking my best guess at a number or assumption that I have extremely little confidence in
  • [major assumption] - means I’m making an assumption with extremely little numerical grounding that is mostly intuitive
  • [moderate assumption] - means I have some numbers or evidence to support the general thrust of the assumption, but only at a more general level that means I have to assume to something more specific
  • [mild assumption] - is something I think is pretty reasonable but that I’m flagging because with more time it would be the type of thing I’d fact check just a bit more
• Citation style: (Author, Date Published) for the first instance, thereafter (Author)

Effects of Siren Noise Pollution

Effects of Road Traffic Noise (RTN) Pollution

Excessive road traffic noise (RTN) can have both a negative acute and long term effect on health. Acute exposure has been shown to increase blood pressure, heart rate, and cardiac output (Munzel et al., 2014). Long-term exposure has been correlated with a significant increase to myocardial infarction (MI), a 10 dB increase above a 56dB baseline leading to a 46% increase in nonfatal MIs (Lim et al., 2021), a number that should give us pause considering roughly 40% of the EU population is exposed to RTN exceeding a LDN^[3] of 55dB and 30% exposed to an excess of 55dB at night. And even though Sivakumaran et al. (2022) notes problems of comparing across studies due to different methodologies and terminologies, their meta-analysis indicates there is (weak) evidence for a correlation between acute and long-term RTN exposure and increases in blood pressure, hypertension, and heart rate. Munzel et al.’s meta-analysis further backs this up, finding a significant relationship between increases in RTN and increases in blood pressure, hypertension, and MI^[4]. The evidence is less sure, but there’s also multiple indicators for an association between RTN and stroke (Harding et al., 2013; Munzel).

What do all these studies amount to? Taken together, they paint a convincing picture of some of the negative consequences of RTN, making a strong case for at least a slight negative effect on heart health. That being considered, what exactly does the global burden look like? A starting point to consider is the UK, where day time noise levels > 55dB have been estimated to cause an annual 542 cases of hypertension-related MI, 788 cases of stroke, and 1169 cases of dementia, costing over £1.09 billion (Harding et al.). But this is just the UK, how do we generalize out to the rest of the world?

Thankfully, the World Health Organization (2011) commissioned an in-depth analysis of a subsect of Western European countries (“EUR-A”^[5] countries). Even under conservative assumptions, the report indicated 1 to 1.6 million Disability Adjusted Life Years (DALYs) are lost each year due to environmental noise. The breakdown is as follows: 61,000 DALYs for ischaemic heart disease; 45,000 DALYs for cognitive impairment of children; 903,000 DALYs for sleep disturbance; 22,000 DALYs for tinnitus; and 587,000 DALYs for annoyance. Because we are concerned with RTN and not environmental noise, we need to estimate what proportion of the lost DALYs are due to RTN. Luckily, they included a breakdown for their DALY estimate for annoyance, which indicated roughly 83% of the DALY loss is due to RTN. I’m going to make a [moderate assumption] that this is roughly the same for the other categories, assuming that RTN is responsible for 83% of the loss identified by the report.^[6] So if we take the midpoint of their projected DALY loss (1.3 million), that would be 1.079 million DALY loss yearly due to RTN in EUR-A.

To generalize this, I need to make a [major assumption] that the rates found in this study are similar throughout the rest of the world and that I can then just multiply the rate out to the whole population to get a sense of the impact of RTN. There are of course way too many variables to properly sort out, but I will speak briefly on one major concern, that most of the countries in the World Health Organization (WHO) study are developed and much of the world lives in developing countries.  Though the methodology differs significantly in places, there is a decided demonstration of RTN in developing countries (Moroe, 2022), and an indication that it leads to some of the same effects (Gilani, 2021). This doesn’t settle other problems, like that of being able to generalize to rural areas, as urban areas experience more RTN (WHO) and the WHO sample probably overselects for largely urbanized countries, but it does at least give an indication that generalizing the WHO numbers may not be as crazy an assumption as a first approximation might indicate. But continuing on, these countries has a combined population of 437,773,888 in 2011^[7], or 6.2% of the world population^[8]. If we just generalize out using the WHO report’s numbers, that would indicate a potential yearly loss of 17,386,295 DALYs in the world^[9].

Siren’s Contribution to RTN

Of the studies that seek to quantify RTN, few seem to separate out and measure individually the individual contributors to the noise, preferring to just measure a general dB (decibel) level instead. McAlexander et. al (2015) are an exception, a team that took 329 ten minute measurements from around New York at variable times and marked down instances of events of various categories. They noted 18 instances of sirens ([mild assumption] that sirens is tracking emergency sirens as they mention ambulance sirens as an example elsewhere), so a presence of sirens in 5.5% of the measurements.

From here it gets tricky, because they don’t note the average duration of each type of sound. So I take a [major assumption], that based on myself and two other’s experience (in lieu of finding better on this), sirens are generally audible for 10 to 20 (15) seconds on average across circumstances^[10]. Based on this, we might expect that sirens were present for 4.5 minutes or 0.14% of the sample time^[11]. Making the jump to assume that we can generalize this rate out is another [major assumption] due to this being a single study, and occurring all within New York which is heavily urbanized and has a robust emergency response system. It fails to account for the variance in laws around siren use that can change not only from country to country, but from state to state in the US, and even city to city within a state, amongst other things. If fails to account for the fact that though sirens may have been present 0.14% of the sample time, they were consistently louder than most other factors (McAlexander et. al) meaning that they might overcontribute in their allotted time, leaving a greater impact than 0.14% would indicate. Nevertheless, it serves as the only rough sample I could find, and thus, without better data, is the best we have to go on.

In Sum

So for the finale, assuming 0.14% of RTN is sirens, and that annual global burden from RTN is 17,386,295 DALYs, sirens would account for a loss of roughly 24,341 DALYs annually. I’d probably like to throw a massive confidence interval around this, and [majorly assume] the value could be +/- 15,000 DALYs so perhaps somewhere between 9,341 and 39,341 DALYS^[12].

Alternate: Deaths Due to Siren Noise Pollution

As this was the original goal, I first searched for any connection between siren use and death, and aside from some literature on L&S use causing increased crashes and fatalities, I found little. After finding the breadth of literature on environmental and RT noise, I began my endeavor to estimate the contribution of siren use in hopes to eventually work back to deaths. The briefly mentioned UK study gave me hope of generating an estimate here, and I’ll walk you through what that calculation looked like and why I didn’t end up including it.

The biggest reason for going with the WHO report over Harding et al. is that their study uses increase in blood pressure due to noise to estimate the disease burden, meaning they miss certain other potential negative effects like cognitive impairments for children. The study does come with some decided strengths, such as using data from both rural and urban areas in the UK to inform the estimates for percentages of the population exposed to each dB level of noise. They also take it a step further than WHO does to estimate the economic burden of environmental noise, which is useful. But because they limit their study in this way, it only gives us a smaller slice of the effect of noise pollution.

Nevertheless, Harding et al. estimate that “environmental noise” causes an annual excess of 1,257 cases of hypertension-related acute myocardial infarction (AMI, aka heart attacks), 1,818 cases of stroke, and 2,707 cases of dementia. We can make a [moderate assumption] that the environmental noise that they reference is similar to what was used in the WHO document, meaning roughly 83% is attributable to RTN, or: 1,043 cases of AMI, 1,509 cases of stroke, and 2,247 cases of dementia. Taking our estimated contribution of sirens to RTN of 0.14%, we come up with: 1.46 cases of AMI, 2 cases of stroke, and 3.15 cases of dementia. If we then take the [super major assumption] that similar rates will bear out for the world, we can divide the 2013 world population by the UK’s^[13] and multiply the result by the cases due to siren noise pollution to get a global estimate of: 165 cases of AMI, 225 cases of stroke, and 355 cases of dementia^[14].

From here, we can try to estimate the excess deaths caused by each of the three categories. OECD (2021) notes that the average mortality rate for AMI in 2009 was 11.4%, so in expectation, siren use causes 19 excess deaths due to AMI. According to the Stroke Association (2022), 25% are fatal within the first year, so in expectation, siren use causes 56 excess deaths due to strokes. Kramarow (2019) notes 261,914 deaths attributable to dementia in the US in 2017, which, as a percentage of the 5,500,000 total reported cases of dementia (Alzheimer’s Association Report, 2017) is 4.76%. Taking a [major assumption] that mortality rates for dementia are roughly similar across the globe, we would expect an excess of 17 deaths due to dementia.

In Sum

So in total, we can state [with incredible uncertainty] that siren use causes an annual excess of 92 deaths globally^[15], perhaps with a subjective CI of (20 to 300)^[16].

Other Effects of Siren Use

On Drivers

An ems.gov report clearly states that “there is clearly an increased risk of EMS vehicle crashes or collisions when operating with L&S [lights and sirens]” with estimates from the 1980s indicate that over 12,000 EMS Vehicle crashes occur annually (Kupas, 2017). More recent data indicates that number might be lower, at around 4,500 “accidents” annually (NHTSA, 2014). Only 5% of patients clinically benefit from quicker transports (Murray, 2017), and L&S use is associated with a 50% greater chance of crashing and almost 3x the chance of crashing with a patient on board (Kupas, 2022).

These crashes don’t just mean risk for the EMS work or patient on board: 84% of them involve three or more people. These are often due to “wake effects” where the ambulance L&S doesn’t directly cause the crash but rather increases the incidence of erratic driving from those around it, often causing people to behave unpredictably (Kupas, 32-33). This is perhaps unsurprising, when considering one simulation indicated that having the radio on and being on the phone in the car yielding a mere four seconds of sirens being audible before the ambulance catches up with you (EMS1, 2013). The consequences in the US alone accrue to roughly 2,600 injuries and 33 deaths annually based on data from 1992 to 2011 (NHTSA). All of these deaths can’t be attributed to sirens, but if we apply the 50% greater chance of crashing from before, we get 1,300 injuries and 16.5 deaths. Now, take the [super major assumption] that sirens are responsible for about 10% of this (given that the majority of the danger is potentially posed by the style of driving, and that lights are also a factor) and we arrive at the conclusion that sirens effect on other drivers potentially cause an annual excess of 130 injuries and 1.65 deaths in the US alone.

On EMS Workers

The NIH (2022) has stated that “long or repeated exposure to sounds at or above 85 dBA can cause hearing loss”. Hansen et al. (2017) found high-noise exposure levels > 80 dB(A) during use of sirens ”. The resulting health effects are in evidence, as ambulance paramedics lose hearing at a faster rate (Johnson, 1980), one study finding it to be 150% that expected for age matched individuals (Kupas, 2017). Given the loss of hearing and consistent exposure to high levels of noise, it would be unsurprising if further research on EMS workers health turned up similar negative effects as seen before on heart health and other physiological factors. Though at the moment, it seems the only strong connection is to degradation of hearing, which itself comes with its own negative health effects (WHO, 2023). 

On Lives Saved

On ambulances: Though there are differing data points out there, one widely cited study indicated use of L&S leads to an average reduction of 43.5 seconds in transport times. They note that this degree of time savings “does not warrant the use of lights and siren during ambulance transport, except in rare situations or clinical circumstances” (Hunt et al., 1995). Another study finds a higher 105.8 second reduction in transport time, but similarly notes that “although statistically significant, this time saving is likely to be clinically relevant in only a very few cases” (Brown et al., 2000). A proper accounting would seek to estimate how many lives are saved in the “very few” or “rare” cases, but I think it’s sufficient to assume this number is quite low. There are (many) (interviews) with experts and EMS professionals that echo the same and speak to the overuse of L&S, not just in the case of ambulances mentioned above, but also for fire services as well.

On fire services: While I couldn’t find any straightforward data for the effect of response times for fire services on lives saved, even a report by The UK Fire Brigades Union (2010) (people incentivized to make the case for the importance of response times) only estimates 13 additional deaths per year (in the UK) due to increased response times (generally over the 44 or 106 seconds savings mentioned by ambulance siren use above). Given data from the On Drivers section above, there’s good reason to believe L&S use probably causes more deaths than it saves in this scenario. One slight counterpoint is that in the case of fire services, damage to buildings does also have to be taken into account, where the Fire Brigade report indicates longer response times might lead to an annual excess of £85 million in damages. But again, I think considerations from the On Drivers section are likely to reduce a lot of this gain, in damages to cars due to crashes^[17] and in medical costs from the resulting injuries, potentially leading to even greater monetary damages.

Other Ways I Thought of Answering and Quick Responses

Deaths and Acute Effects of Sirens

I found some studies that indicated sirens and similar noises could have an acute effect on the body, but as far as I could tell, none of these studies tried to give estimates of what that effect would amount to in the environment. This is most likely because many of the negative health effects of RTN are cumulative rather than immediate. A common chain of causality looks something more like: momentary siren noise^[18] leads to momentary increases in blood pressure which overtime results in a cumulative increase in strain on the heart which is associated with an increase in cases of MI. Whereas one might think the acute effects skip the line of causality to go straight from momentary siren noise to MI, something that I just didn’t find any support for in the literature.

1. ^{^}

    More precisely, it was “How many lives are lost globally due to emergency siren noise pollution?”, see below for why I came to DALYs instead of lives lost

2. ^{^}

    While I haven’t done the calculations for net lives/DALY loss, I think any potential positives are counterweighted by the increase to injury and death caused by use of L&S (lights and sirens). See the On Lives Saved and On Drivers section for more.

3. ^{^}

    Day–night level, a measure of noise which summarizes average noise exposure over a 24 hour period

4. ^{^}

    Also of note: one study indicates an increase in strokes for those older than 64.5.

5. ^{^}

    A collection of Western European countries chosen due to data availability: Andorra, Austria, Belgium, Croatia, Cyprus, the Czech Republic, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Luxembourg, Malta, Monaco, the Netherlands, Norway, Portugal, San Marino, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom

6. ^{^}

    Though I should note I would explore this more with time, because the report also indicated that other noises can have a more acute effect on certain categories, rail for example having a more acute effect on cognitive impairment, even though it still probably only makes up only a small fraction of the overall noise. Perhaps an estimate for RTN’s contribution to the cognitive impairment burden should be something more like 80% or 75%. But my assumption is further bolstered by other statements elsewhere, like “sleep disturbance and annoyance, mostly related to road traffic noise, comprise the main burden of environmental noise” (v).

7. ^{^}

    Andorra (70,567), Austria (8,392,000), Belgium (11,040,000), Croatia (4,281,000), Cyprus (1,145,000), the Czech Republic (10,500,000), Denmark (5,571,000), Finland (5,388,000), France (65,350,000), Germany (80,270,000), Greece (11,100,000), Iceland (319,014), Ireland (4,588,252), Israel (7,766,000), Italy (59,380,000), Luxembourg (518,347), Malta (416,268), Monaco (33,945), the Netherlands (16,690,000), Norway (4,953,000), Portugal (10,560,000), San Marino (32,495), Slovenia (2,053,000), Spain (46,740,000), Sweden (9,449,000), Switzerland (7,912,000), and the United Kingdom (63,260,000) = 437,773,888 total

8. ^{^}

    World population (2011): 7,054,000,000

9. ^{^}

    1,079,000 x 16.129 = 17,386,295 DALYs lost

10. ^{^}

     Inside not near window, inside near window, walking on the sidewalk, etc

11. ^{^}

     (0.25 minutes x 18 instances) / (329 instances x 10 minutes) = 0.00137

12. ^{^}

     The choice of 15,000 on either end is an intuitive choice that is meant just to reflect the fact that there are significant reasons it could be higher or lower, two of which I’ll mention here. A reason it could be much higher: Harding’s analysis indicates there's a significant effect from RTN on dementia, something not considered at all in the WHO report. A reason it could be much lower: the estimate for the prevalence of sirens potentially too high, because it is based on data from New York City, an urban location that is more likely to have a higher incidence of sirens than rural areas, or potentially even smaller cities.

13. ^{^}

     7,229,000,000 (world) / 64,130,000 (UK) = 112.72

14. ^{^}

     112.72 x 1.46 = 165 (AMI); 112.72 x 2 = 225 (stroke); 112.72 x 3.15 = 355 (dementia)

15. ^{^}

     19 + 56 + 17 = 92 deaths

16. ^{^}

     There is again reason to think this number is lower than it should be (because the siren estimate may be too large due to it coming from an urban environment) but I think there is stronger evidence here for it being lower than the real amount, due to the fact that Harding et al.’s estimates were restricted to a smaller range of potential negative effects.

17. ^{^}

     Especially when you consider the exorbitant costs to buy and repair fire trucks

18. ^{^}

     As a part of RTN