We used data on all 85,854,176 deaths in the USA from 1980 to 2016 from the National Center for Health Statistics (NCHS). Age, sex, state of residence, month of death, and underlying cause of death were available for each record. The underlying cause of death was coded according to the international classification of diseases (ICD) system (9
th revision of ICD from 1980 to 1998 and 10
th revision of ICD thereafter). Yearly population counts were available from NCHS for 1990 to 2016 and from the US Census Bureau prior to 1990 (Ingram et al., 2003 (
link)). We calculated monthly population counts through linear interpolation, assigning each yearly count to July.
We also subdivided the national data geographically into nine climate regions used by the National Oceanic and Atmospheric Administration (
Figure 18 and
Table 2) (Karl and Koss, 1984 ). On average, the Southeast and South are the hottest climate regions with average annual temperatures of 18.4°C and 18°C respectively; the South also possesses the highest average maximum monthly temperature (27.9°C in July). The lowest variation in temperature throughout the year is that of the Southeast (an average range of 17.5°C). The three coldest climate regions are West North Central, East North Central and the Northwest (7.6°C, 8.0°C, 8.2°C respectively). Mirroring the characteristics of the hottest climate regions, the largest variation in temperature throughout the year is that of the coldest region, West North Central (an average range of 30.5°C), which also has the lowest average minimum monthly temperature (−6.5°C in January). The other climate regions, Northeast, Southwest, and Central, possess similar average temperatures (10°C to 14°C) and variation within the year of (23°C to 26°C), with the Northeast being the most populous region in the United States (with 19.8% total population in 2016).
Data were divided by sex and age in the following 10 age groups: 0–4, 5–14, 15-24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, 85+ years. We calculated monthly death rates for each age and sex group, both nationally and for sub-national climate regions. Death rate calculations accounted for varying length of months, by multiplying each month’s death count by a factor that would make it equivalent to a 31 day month.
For analysis of seasonality by cause of death, we mapped each ICD-9 and ICD-10 codes to four main disease categories (
Table 1) and to a number of subcategories which are presented in the Supplementary Note. Cardiorespiratory diseases and cancers accounted for 56.4% and 21.2% of all deaths in the USA, respectively, in 1980, and 40.3% and 22.4%, respectively, in 2016. Deaths from cardiorespiratory diseases have been associated with cold and warm temperatures (Basu, 2009 (
link); Basu and Samet, 2002 (
link); Bennett et al., 2014 (
link); Braga et al., 2002 (
link); Gasparrini et al., 2015 (
link)). Injuries, which accounted for 8% of all deaths in the USA in 1980 and 7.3% in 2016, may have seasonality that is distinct from so-called natural causes. We did not further divide other causes because the number of deaths could become too small to allow stable estimates when divided by age group, sex and climate region.
We obtained data on temperature from ERA-Interim, which combines predictions from a physical model with ground-based and satellite measurements (Dee et al., 2011 (
link)). We used gridded four-times-daily estimates at a resolution of 80 km to generate monthly population-weighted temperature by climate region throughout the analysis period.