1. Epidemiology and Global Health
Download icon

National and regional seasonal dynamics of all-cause and cause-specific mortality in the USA from 1980 to 2016

  1. Robbie M Parks
  2. James E Bennett
  3. Kyle J Foreman
  4. Ralf Toumi
  5. Majid Ezzati  Is a corresponding author
  1. Imperial College London, United Kingdom
  2. University of Washington, United States
Research Article
Cite this article as: eLife 2018;7:e35500 doi: 10.7554/eLife.35500
20 figures, 2 tables, 4 data sets and 1 additional file

Figures

Wavelet power spectra for national time series of all-cause death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.004
Wavelet power spectra for national time series of all-cause death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.005
Wavelet power spectra for national time series of cancer death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.006
Wavelet power spectra for national time series of cancer death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.007
Figure 5 with 3 supplements
Wavelet power spectra for national time series of cardiorespiratory disease death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.008
Figure 5—figure supplement 1
Wavelet power spectra for national time series of cardiovascular disease death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.009
Figure 5—figure supplement 2
Wavelet power spectra for national time series of chronic respiratory disease death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.010
Figure 5—figure supplement 3
Wavelet power spectra for national time series of respiratory infection death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.011
Figure 6 with 3 supplements
Wavelet power spectra for national time series of cardiorespiratory disease death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.012
Figure 6—figure supplement 1
Wavelet power spectra for national time series of cardiovascular disease death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.013
Figure 6—figure supplement 2
Wavelet power spectra for national time series of chronic respiratory disease death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.014
Figure 6—figure supplement 3
Wavelet power spectra for national time series of respiratory infection death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.015
Figure 7 with 2 supplements
Wavelet power spectra for national time series of injury death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.016
Figure 7—figure supplement 1
Wavelet power spectra for national time series of intentional injury death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.017
Figure 7—figure supplement 2
Wavelet power spectra for national time series of unintentional injury death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.018
Figure 8 with 2 supplements
Wavelet power spectra for national time series of injury death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.019
Figure 8—figure supplement 1
Wavelet power spectra for national time series of intentional injury death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.020
Figure 8—figure supplement 2
Wavelet power spectra for national time series of unintentional injury death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.021
Figure 9 with 5 supplements
Wavelet power spectra for national time series of death rates from causes other than cancers, cardiorespiratory diseases and injuries for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.022
Figure 9—figure supplement 1
Wavelet power spectra for national time series of substance use disorder death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.023
Figure 9—figure supplement 2
Wavelet power spectra for national time series of perinatal condition death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.024
Figure 9—figure supplement 3
Wavelet power spectra for national time series of endocrine disorder death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.025
Figure 9—figure supplement 4
Wavelet power spectra for national time series of genitourinary disease death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.026
Figure 9—figure supplement 5
Wavelet power spectra for national time series of neuropsychiatric disorder death rates for 1980–2016, by age group for males.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.027
Figure 10 with 6 supplements
Wavelet power spectra for national time series of death rates from causes other than cancers, cardiorespiratory diseases and injuries for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.028
Figure 10—figure supplement 1
Wavelet power spectra for national time series of substance use disorder death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.029
Figure 10—figure supplement 2
Wavelet power spectra for national time series of maternal condition death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.030
Figure 10—figure supplement 3
Wavelet power spectra for national time series of perinatal condition death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.031
Figure 10—figure supplement 4
Wavelet power spectra for national time series of endocrine disorder death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.032
Figure 10—figure supplement 5
Wavelet power spectra for national time series of genitourinary disease death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.033
Figure 10—figure supplement 6
Wavelet power spectra for national time series of neuropsychiatric disorder death rates for 1980–2016, by age group for females.

Wavelet power values increase from blue to red. The shaded regions at the left and right edge of each box indicate the cone of influence, where spectral analysis is less robust. P-values for the presence of 12 month seasonality are to the right of each figure at the 12 month line.

https://doi.org/10.7554/eLife.35500.034
Figure 11 with 3 supplements
Mean timing of maximum and minimum all-cause and cause-specific mortality at the national level, by sex and age group for 1980–2016.

Red arrows indicate the month of maximum mortality, and green arrows that of minimum mortality. The size of the arrow is inversely proportional to its respective variance.

https://doi.org/10.7554/eLife.35500.035
Figure 11—figure supplement 1
Mean timing of maximum and minimum mortality for specific cardiorespiratory diseases at the national level, by sex and age group for 1980–2016.

Red arrows indicate the month of maximum mortality, and green arrows that of minimum mortality. The size of the arrow is inversely proportional to its respective variance.

https://doi.org/10.7554/eLife.35500.036
Figure 11—figure supplement 2
Mean timing of maximum and minimum mortality for specific injuries at the national level, by sex and age group for 1980–2016.

Red arrows indicate the month of maximum mortality, and green arrows that of minimum mortality. The size of the arrow is inversely proportional to its respective variance.

https://doi.org/10.7554/eLife.35500.037
Figure 11—figure supplement 3
Mean timing of maximum and minimum mortality for the cluster of causes other than cancers, cardiorespiratory diseases and injuries at the national level, by sex and age group for 1980–2016.

Red arrows indicate the month of maximum mortality, and green arrows that of minimum mortality. The size of the arrow is inversely proportional to its respective variance.

https://doi.org/10.7554/eLife.35500.038
Figure 12 with 3 supplements
National percent difference in death rates between the maximum and minimum mortality months for all-cause and cause-specific mortality in 2016 versus 1980, by sex and age group.
https://doi.org/10.7554/eLife.35500.039
Figure 12—figure supplement 1
National percent difference in death rates between the maximum and minimum mortality months for specific injuries in 2016 versus 1980, by sex and age group.
https://doi.org/10.7554/eLife.35500.040
Figure 12—figure supplement 2
National percent difference in death rates between the maximum and minimum mortality months for specific cardiorespiratory diseases in 2016 versus 1980, by sex and age group.
https://doi.org/10.7554/eLife.35500.041
Figure 12—figure supplement 3
National percent difference in death rates between the maximum and minimum mortality months for the cluster of causes other than cancers, cardiorespiratory diseases and injuries in 2016 versus 1980, by sex and age group.
https://doi.org/10.7554/eLife.35500.042
Mean timing of maximum all-cause mortality for 1980–2016, by climate region and age group for males.

Average temperatures (in degrees Celsius) are included in white for the corresponding month of maximum and minimum mortality for each climate region.

https://doi.org/10.7554/eLife.35500.043
Mean timing of minimum all-cause mortality for 1980–2016, by climate region and age group for males.

Average temperatures (in degrees Celsius) are included in white for the corresponding month of maximum and minimum mortality for each climate region.

https://doi.org/10.7554/eLife.35500.044
Mean timing of maximum all-cause mortality for 1980–2016, by climate region and age group for females.

Average temperatures (in degrees Celsius) are included in white for the corresponding month of maximum and minimum mortality for each climate region.

https://doi.org/10.7554/eLife.35500.045
Mean timing of minimum all-cause mortality for 1980–2016, by climate region and age group for females.

Average temperatures (in degrees Celsius) are included in white for the corresponding month of maximum and minimum mortality for each climate region.

https://doi.org/10.7554/eLife.35500.046
The relationship between percent difference in all-cause death rates and temperature difference between months with maximum and minimum mortality across climate regions, by sex and age group in 2016.
https://doi.org/10.7554/eLife.35500.047
Author response image 1
Percent difference in death rates between the maximum and minimum mortality months for all-cause mortality in 2016 versus 1980 by sex, age group and region.
https://doi.org/10.7554/eLife.35500.052
Author response image 2
The relationship between annual mean temperature (used in European papers) and temperature range between maximum and minimum mortality months (used in our paper).
https://doi.org/10.7554/eLife.35500.053

Tables

Table 1
Number of deaths, by cause of death and sex from 1980 to 2016.
https://doi.org/10.7554/eLife.35500.003
CauseMaleFemaleTotal
All cause43,558,20342,295,97385,854,176
Cancers10,481,5829,476,53019,958,112
Cardiorespiratory diseases20,168,04921,109,52541,277,574
Cardiovascular diseases16,238,34417,210,55633,448,900
Chronic respiratory diseases2,791,6522,595,9505,387,602
Respiratory infections1,138,0531,303,0192,441,072
Injuries4,034,8761,768,1705,803,046
Unintentional2,489,1421,348,1873,837,329
Intentional1,545,734419,9831,965,717
Other causes8,873,6969,941,74818,815,444
Table 2
Characteristics of climate regions of the USA.
https://doi.org/10.7554/eLife.35500.049
Climate regionConstituent statesPopulation (2016)Mean annual temperature (1980–2016) (°C)
CentralIllinois, Indiana, Kentucky, Missouri, Ohio, Tennessee, West Virginia50,191,32611.6
East North CentralIowa, Michigan, Minnesota, Wisconsin24,418,7388
NortheastConnecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont64,046,74110.6
NorthwestAlaska, Idaho, Oregon, Washington13,811,8108.2
SouthArkansas, Kansas, Louisiana, Mississippi, Oklahoma, Texas45,388,41418
SoutheastAlabama, Florida, Georgia, North Carolina, South Carolina, Virginia59,356,07218.4
SouthwestArizona, Colorado, New Mexico, Utah17,613,98113.6
WestCalifornia, Hawaii, Nevada43,708,57416.6
West North CentralMontana, Nebraska, North Dakota, South Dakota, Wyoming5,168,7537.6

Data availability

The ERA-Interim temperature data are available at https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim. The US Census Populations With Bridged Race Categories data for 1990-2016 are available at https://wonder.cdc.gov/bridged-race-population.html. Pre 1990, the County Intercensal Tables are available at https://www.census.gov/data/tables/time-series/demo/popest/1980s-county.html. The Vital statistics data are available at https://www.cdc.gov/nchs/nvss/dvs_data_release.htm through a request to NAPHSIS (https://www.naphsis.org/).

The following previously published data sets were used
  1. 1
    National Center for Health Statistics
    1. Center for Health Statistics National
    (2017)
    Vital statistics (1980-2016).
  2. 2
    European Centre for Medium-Range Weather Forecasts
    1. Centre for Medium-Range Weather Forecasts European
    (2016)
    ERA-Interim temperature data (1979-2016).
  3. 3
    CDC WONDER
    1. Wonder CDC
    (2016)
    US Census Populations With Bridged Race Categories (1990-2016).
  4. 4
    United States Census Bureau
    1. States Census Bureau United
    (2016)
    County Intercensal Tables (1980-1989).

Additional files

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)