Online Resource 6. Mortality Displacement Impacts in Health Impacts Assessment of Projected

Online Resource 6. Mortality Displacement Impacts in Health Impacts Assessment of Projected

Online Resource 6. Mortality Displacement Impacts in Health Impacts Assessment of Projected Extreme Temperatures

Authors: David Mills,1 Joel Schwartz,2Mihye Lee,2 Marcus Sarofim,3 Russell Jones,1Megan Lawson,4 Michael Duckworth1, and Leland Deck5

Affiliations: 1Stratus Consulting Inc., 1881 Ninth Street, Suite 201, Boulder, CO 80302, USA; 2Harvard University, Harvard School of Public Health, Department of Environmental Health, Department of Epidemiology, 677 Huntington Avenue, Boston, MA 02115, USA; 3Climate Change Division (6207-J), U.S. Environmental Protection Agency, 1200 Pennsylvania Avenue NW, Washington, DC 20460, USA; 4Headwaters Economics, P.O. Box 7059, Bozeman, MT 59771, USA; 5Clean Air Markets Division, U.S. Environmental Protection Agency, 1310 L Street NW, Washington, DC 20005, USA

Corresponding author: David Mills, 1Stratus Consulting Inc., 1881 Ninth Street, Suite 201, Boulder, CO 80302, USA

Telephone: 303-381-8248

Fax 303-381-8000

Email:

Overview of mortality displacement and extreme temperature events

A question of interest when evaluating mortality estimates from current extreme temperature events or projected impacts under potential future climates is whether an the mortality impact is realized primarily by individuals who are already in poor health and believed to be near death or by individuals who are seen as generally healthy.

The share of excess deaths likely to have been realized by individuals considered to be near death before an extreme temperature event can be estimated from mortality patterns during and after the event. This measures what is formally recognized as the mortality displacement or “harvesting” of the event. Mortality displacement occurs when an event causes increased mortality. Figure 1 presents a hypothetical example of an extreme temperature event that results in mortality displacement to help visualize this concept.

Figure 1. Mortality displacement in a hypothetical extreme temperature event.

The mortality displacement of this hypothetical event is calculated by comparing estimates of the excess deaths attributable to the EHE (shaded area A) to observed reductions in mortality after the EHE (shaded area B). The mortality displacement calculation of B/A estimates the percentage of the EHE-attributable deaths attributable to individuals assumed to be in a near-death state as the event developed.

This calculation assumes that a population has a relatively constant number of people living in the near-death and healthy states and that these populations are kept in balance by additions to the population (births, relocation), the movement of persons into the near-death category, and the death of persons in the near-death category (movement from being near death to healthy is not an option in this framework). If extreme heat events also influence recruitment into the near death category, the situation could be more complicated, and net mortality displacement reduced.

In this view, as the A/B ratio increases, the role of mortality displacement in an EHE decreases reflecting that more of the event-attributable deaths were of healthy compared to near-death individuals. Visually, an extreme temperature event with minimal mortality displacement would show a spike in mortality during the event and then return quickly to long-term average levels following the event. Such a pattern would reflect that pool of near death individuals was not disproportionately depleted by the event.

Table 1provides estimates of mortality displacement calculated for a series of specific extreme heat events experienced in the United States.

Table 1. Mortality displacement estimates for select EHEs in the United States
City
(year of EHE) / Mean daily summer mortality
(deaths per 100,000) / Sum of mortality above mean associated with the EHE
(A) / Sum of mortality below mean following the EHE
(B) / Percent of EHE deaths attributable to mortality displacement
(B/A)
St. Louis (1980) / 2.82 / 17.13 / 2.67 / 16%
St. Louis (1966) / 2.90 / 33.37 / 7.34 / 22%
New York (1984) / 2.67 / 7.75 / 3.00 / 39%
New York (1975) / 3.02 / 8.65 / 3.45 / 40%
New York (1966) / 3.50 / 19.76 / 8.56 / 43%
Birmingham (1980) / 2.44 / 14.16 / 10.19 / 72%
Source: Table 1 in Kalkstein, 1997.

The results in Table 1 vary substantially for the percentage of deaths attributable to mortality displacement from the EHE, from 16% to 72%. In most of the EHEs studied, however, most of the excess mortality appears to have been experienced by healthy individuals because the mortality displacement estimates are below 50%. Moreover, most of these events occurred several decades earlier and changes in air conditioning prevalence, health practices, etc. make it hard to extrapolate. Results from other notable,and more recent,extreme heat events include an estimate that less than 10% of the excess mortality from the nine largest French cities during the 2003 European heatwave reflected harvesting, although the results varied by city (Le Tetre et al., 2006). Considering the range of these values, characteristics of these events reflecting their deviation from normal conditions and additional research into the nature of the mortality attributable to extreme temperature events (e.g., Kinney et al., 2012), it appears that the percentage of the deaths attributable to harvesting declines with the severity of the event. In other words, as the event increasingly represents a departure from normal conditions harvesting accounts for less of the observed impact as mortality in the pool of those near death is increasingly offset by mortality from those not otherwise expected to die.

Mortality displacement in time-series studies

For extreme temperatures, the emphasis of much of the investigation surrounding mortality displacement has been on specific events. In studies that evaluate the mortality impact of extreme temperatures on populations over multiple years, the potential impact of mortality displacement is evaluated based on a consideration of the results from model specifications that consider the sign and significance of the impact coefficient for the exposure over different time periods (i.e., lags). For example, to what extent is a day’s mortality total a function of the minimum temperature on the same day (lag 0), three days before (lag 3), or two weeks prior (lag 14).

When modeling results return values that are consistently statistically insignificant for increasing lag periods, a conclusion that harvesting is not still occurring can be justified. For example, because the Medina-Ramon and Schwartz (2006) study only considered the impact of same day and previous day extreme temperatures (i.e., lag 0 and lag 1) it could not draw conclusions about the presence of harvesting over longer exposures. As noted specifically in the study:

“However our study was not designed to appropriately assess the presence of harvesting by examining longer lags, and thus we cannot rule out the occurrence of a harvesting effect for extreme temperatures beyond lag 1.” (Medina-Ramon and Schwartz, 2007, p. 831-832).

Another consideration is that as longer lags are examined it becomes increasingly difficult to separate the effects of longer lags of temperature and the effects of season, making it harder to examine mortality displacement for lags much more than two weeks.

Potential impact of mortality displacement in the current results

Reviewing the studies referenced above, Medina-Ramon and Schwartz (2007) consider the impact on their results if 10% - 40% of the mortality attributed to extremely hot days reflected mortality displacement. Incorporating a similar assumption, our study’s general conclusion, which is climate change is likely to increase the risk and incidence of mortality attributable to extreme temperatures accounting for impacts to extremely hot and extremely cold days, would still hold. Future studies may be able to more fully address the question of mortality displacement through modeling design but the ultimate impact on projected health impacts will remain an area where precision remains elusive but general conclusions can be evaluated by considering alternative assumptions as we have done here with respect to evaluating the impact of alternative threshold temperatures.

More promisingly, quantitative approaches to estimating the mortality impact of temperatures that estimate the impact over the full range of observed temperatures in the historical record hold the promise of reducing the extent to which mortality displacement is a concern. Specifically, using approaches such as estimating the impact of temperature in a location over a season using spline functions would allow for developing quantitative impact estimates where changes on all future days are accounted for. In this scenario mortality displacement concerns would be addressed as the seasonal nature of the estimation would mean that mortality displacement that occurred at any point following an event in the historical record would have a corresponding impact on following days that would be captured in the estimated temperature impact.

References

Kalkstein, LS (1997). Climate and human mortality: Relationship and mitigating measures. Advances in Bioclimatology 5:161-177.

Kinney PL, Pascal M, Vautard, R, Laaid, K (2012) Winter mortality in a changing climate: will it go down. Bull EpidémiolHebd, 12-13(March):148–151.

Le Tetre, A, Lefranc A, EilsteinDeclercq C, Medina S, Blanchard M, Chardon B, Fabre P, Filleul L, Jusot JF, Pascal L, Prouvost H, Cassadou S, Ledrans M (2006) Impact of the 2003 heatwave on all-cause mortality in 9 French cities. Epidemiology 17: 75-79.

Medina-Ramon M, Schwartz J (2007) Temperature, temperature extremes, and mortality: A study of acclimatization and effect modification in 50 US cities. Occup Environ Med 64:827833

1