Quantifying Disaster Deaths

I have recently realized the depth of confusion surrounding the issue of how disaster death tolls are calculated. Almost all death tolls numbers from large disasters/ catastrophes are estimates (sometimes smaller events too). We rarely have an exact count. As people are introduced to the idea that we will likely never know the exact death toll from Hurricane Maria, I think it's important to explore this in more depth. 

I’ll discuss death tolls more broadly and then address Hurricane Katrina & the Federal Levee Failure and Hurricane Maria in Puerto Rico. Read this post for a less clinical discussion of Puerto Rico’s death toll.

Media Reporting

One of the first things the media reports when a disaster happens, and that people from away want to know, is how many lives have been lost. During the response, as the situation is still unfolding, there are inconsistent reports as the media adjusts their totals based on information they are gathering, often in high-pressure, urgent situations. It is not unusual to see counts exaggerated or underestimated in the hours and days following these events. Depending on the situation it can take a long time to come up with a total (if logistics even allow) as search and rescue, medical interventions, body recovery, official reporting, and family notification all need to occur before or as totals are released to the public.

Who Counts

Wayne Blanchard describes accounting for the death tolls from disasters and other events in US history; his life’s work. He has created a phenomenal website where he has compiled, to my knowledge, the most complete data set of deaths from events in the United States. His goal is for the website “to be the most complete, and the most authoritative, source in existence on large-loss-of-life events in American history.” Blanchard has identified 5,400 events with death tolls over ten and another 400 events that had less than 10 deaths but were otherwise exceptional. 

A number of agencies are involved in assessing the number of deaths. Blanchard provides an example on his website: 

“Several agencies look at the same event through different lenses, definitions and report slices of information but not the complete picture. For example, if one were to look at the National Weather Service/NOAA figure for 2013 heat fatalities, you would find 92. If one then looks at one of several sources that look at heat associated deaths of young children in vehicles for 2013, you would find details on 44 such deaths, only several of which are noted by the NWS/NOAA or NCDC/NOAA1 within their 92 deaths figure. Then if one consults the Occupational Safety and Health Administration (OSHA) for worker deaths for the year and look for heat-related deaths, one finds more than a dozen, only which several, again, are included in the NWS figure. Finally, when one reviews a State Department of Health, or similar State agency document, one often finds yet more losses of life in this heat event. Thus whereas the NWS/NOAA shows 7 heat-related deaths for Arizona and the NCDC/NOAA shows 9, the Arizona Department of Health Services shows 139.2 Then there is from the Centers for Disease Control and Prevention, the CDC WONDER mortality data website. It shows for AZ for 2013, using ICD-10 X30 code (exposure to excessive natural heat), 151 such deaths. It also shows 372 such deaths nationwide When I draw from all sources I arrive at 422 heat-associated during 2013, as opposed to the 92 noted by the National Weather service and the National Climatic Data Center.”

A huge factor is determining what "counts" as a disaster death. Different groups and agencies have different guidelines for what to consider. Generally, there are three categories. Direct deaths are the obvious ones -- e.g., someone drowns during a flood. Secondary effects are sometimes mixed in with direct deaths but are sometimes kept separate. These are deaths that result from a secondary effect of the primary hazard -- e.g., someone is killed in a landslide that resulted from an earthquake. Finally, there are indirect deaths. This is where the real controversy is found. Does someone who dies while sitting for 24 hours in a car trying to evacuate be counted in the total? Research has found there is a significant increase in suicides following hurricanes, flooding, and earthquakes. It is unlikely that most death totals include these deaths. Where are lines drawn and who gets to draw those lines? Again, there are guidelines out there but they differ dramatically from group to group and of course, logistical issues are situation dependent. 

We should absolutely track and account for all deaths related to the disaster both direct and indirect. Indirect impacts, like job and wage loss, are counted. I can't think of a legitimate reason to not track deaths through the recovery. Why would we not also count people?

In his quest to compile this data he acknowledges one particular challenge as we look back at previous death tolls: “A fatality number gets printed and then is picked up and used by others year after year without searching for actual confirmation the number is factual or defendable. Only sifting through a variety of sources allows one to get a good, or at least better, handle on the mortality of an event.”

Disaster researchers know these death tolls are messy (a lot of disaster data is messy). If you don't get an exact count right away, it can be really difficult, if not impossible, to retroactively count the dead without using statistics to estimate the total. Researchers usually have some type of qualifier in their studies that acknowledge they’re working with imperfect/ incomplete data sets. Two examples:

It's not wrong or unusual to have different groups using different methods to come up with death tolls, but it can directly affect the response.

The Implications

Acknowledging the number of people killed can bring solace to victim's families. This alone should be enough, but there are a number of other logistical and legal reasons having an accurate death toll is important.

Death tolls are one of the primary ways that people “from away” conceptualize the size of a disaster which is directly tied to aid. Although we should be responding based on need rather than impact. The death toll does not perfectly equate with the size of the disaster, sometimes large disasters have small death tolls and small disasters have large death tolls. Still, it is an important indicator. The death toll is one factor that influences donations. Other countries have been accused of inflating numbers to increase the amount of foreign aid and donations they receive. Conversely, some have been accused of underestimating numbers in an effort to hide dire situations. 

Cause of death can influence who is responsible legally and provide federal disaster funeral assistance funding

Cause of death influences who is eligible for burial assistance post-disaster (federal funding). 

Screen Shot 2018-06-04 at 6.11.20 PM.png


Hurricane Katrina & The Federal Levee Failure


The first time I learned about the controversy in how we tally disaster death totals was in the context of Hurricane Katrina and the Federal Levee Failure. In college, as I was artfully finding ways to make every paper I wrote about Katrina, I noticed the wide variety of death tolls published. Many put the death toll around 1200. I had thought it was 1600 because of a sign I always saw in the Lower Ninth Ward. Others put the total around 2000. Some put it as low as 800. Though there was agreement that many people died, there was little agreement in exactly how many people died and who should be included in the total.

As I started tracing back the citations of these death tolls I figured out why. At the time of publication some people were still considered “missing”. Some totals had originally only considered the people who had died in New Olreans but overtime had evolved being cited as the total for the entire United States. Cause of death was not always clearly articulated on death certificates. Finally, there were discrepancies in which causes of death qualify as stemming from the hurricane and flooding. This study is a quick read that can give you a sense of some of these complications. This excellent article from Five Thirty Eight reviews the efforts undertaken in the decade following the flood to more accurately account for the indirect deaths. 

I personally use 1833 because it is the total used by the Times-Picayune. As local journalists, I trust them the most in this case.


Hurricane Maria

The official government total in Puerto Rico following Hurricane Maria stuck at 64 but researchers quickly knew that was incorrect.  Sheri Fink wrote an excellent summary for the New York Times that explains the efforts to arrive at an official count in Puerto Rico. Last week a Harvard study found that upwards of 4,645 people died as a result of Hurricane Maria. At 70 times the offical total, the study made headlines. 1/3 of deaths were attributed to delay and disruptions in care (i.e., indirect deaths).

Did exactly 4,645 people die from Hurricane Maria? No, it's an estimate. But knowing all that we do about how difficult it is to come up with an accurate count, especially in the case of Puerto Rico where it appears that records were not well kept and so many survivors moved to the mainland. For now, it's okay to use this number while more data is collected to refine the findings. 

It’s really easy to manipulate death toll numbers, intentionally or unintentionally. It's also important to point out that the way we have historically recorded death tolls in the US is a really good example of systemic failures of the emergency management for marginalized groups. We have not collected data on the full effects felt by people of color, white women, and other marginalized groups who we know encounter poor access to adequate mental/ healthcare post-disaster.

Disaster data is messy… which is even more evidence for why it is so important to have disaster researchers out in the general public talking about their research and the research of others in our field. There are nuances, qualifiers, and context that must be communicated.



*Forever blog caveat that this is a simplified version of this topic.