Effects of Extreme Temperatures From Climate Change on the Medicare Population: Preliminary Results

Presentation by Jared Jageler, David Adler, Noelia Duchovny, and Evan Herrnstadt, analysts in CBO’s Microeconomic Studies and Health Analysis Divisions, at the Association of Environmental and Resource Economists Summer Conference.

1. Presentation at the Association of Environmental and Resource Economists
Summer Conference
May 31, 2024
Jared Jageler, David Adler, Noelia Duchovny, and Evan Herrnstadt
Microeconomic Studies Division and Health Analysis Division
Effects of Extreme Temperatures
From Climate Change on the Medicare
Population: Preliminary Results
For information about the conference, see www.aere.org/summer.

2. 1 The information in this presentation is preliminary and is being circulated to stimulate discussion and critical comment.
Climate change is expected to increase average temperatures and affect the
frequency of extremely hot and cold days.
The net effects of those changing temperatures on health and federal health care
spending are uncertain:
▪Extreme heat and cold are associated with a range of negative health effects.
▪Spending and utilization could increase or decrease depending on how health
effects and behavior respond to extreme temperatures.
▪Mortality could increase (because of more hot days) or decrease (because of
fewer cold days).
Those effects may be pronounced among elderly populations, with implications
for the federal budget related to Medicare enrollees and spending.
Introduction

3. 2 Adaptation refers to how areas with different climates respond to increasing temperatures differently, such as by installing more air conditioning.
The Congressional Budget Office uses a complete nationwide data set of Medicare
beneficiaries to examine health care utilization and spending as well as mortality.
In particular, CBO investigates these questions:
▪What is the effect of extreme temperatures on emergency department (ED) visits,
ED spending, and mortality among Medicare beneficiaries?
▪How will those outcomes be affected by projected temperatures resulting from
climate change, after incorporating the effects of adaptation?
This research is preliminary, and the results are subject to change.
Research Questions

4. 3
▪Nationwide studies on the effects of climate change on mortality
–Heutel, Miller, and Molitor (2021)
–Carleton and others (2022)
▪Other recent research on health care utilization (ED visits) and mortality in
California hospitals:
–White (2017)
–Gould and others (2024)
▪A study estimating the effects of exposure to particulate pollution on similar
outcomes in the Medicare population
–Deryugina and others (2019)
Related Research

5. 4 Overall, CBO finds a daily weighted average of 11 deaths, 156 ED visits, and $388,000 in spending per 100,000 Medicare enrollees. ED = emergency department.
Overview: Econometric Estimates
Very hot days cause more
ED visits and higher ED
spending. By contrast, very
cold days cause a reduction
in ED visits and a smaller
reduction in ED spending.
Mortality is higher on both
very hot and very cold days.
The current estimates show
the effect of weather only on
outcomes for the same day
and the following two days,
which could exclude some
delayed and compensatory
effects.
CBO plans to explore longer
windows in future versions
of this analysis, which could
change the results.

6. 5 Daily average temperature is the simple mean of daily minimum and maximum temperatures.
Medicare Outcomes
▪Daily counts of ED visits and spending by zip code (2000 to 2019) for fee-for-
service (FFS) enrollees
–Also disaggregated to inpatient (IP) and outpatient (OP) visits and spending;
IP involves an overnight hospital stay, but OP does not
▪Daily deaths by zip code per 100,000 enrollees (2000 to 2019)
–Includes FFS and Medicare Advantage
Weather
▪Daily average temperature and precipitation from the National Oceanic and
Atmospheric Administration’s nClimGrid-Daily (2000 to 2019)
▪The National Aeronautics and Space Administration’s Earth Exchange Global Daily
Downscaled Projections (through 2099)
–25-kilometer x 25-kilometer grid
–Daily projections of average temperature
▪Both are matched to zip code tabulation areas (ZCTAs) and weighted by inverse
distance to ZCTA centroid
Medicare and Climate Data

7. 6 Medicare data are aggregated to the zip code and temperature data to the ZCTA. Those levels are nearly identical, but CBO performs a crosswalk between the two to stay at the ZCTA level.
�
??????��??????=෍
??????
�
??????????????????????????????
????????????��??????+��
??????�+�
????????????+�
�??????+??????
??????��??????
▪y
isty​ is the three-day sum of the health outcome (ED visits or ED spending per 100,000 Medicare FFS
enrollees, or mortality per 100,000 Medicare enrollees) in ZCTA ??????, in state s, on date ??????, in year �.
▪�
?????? are coefficients of interest.
▪????????????????????????
????????????��?????? are the daily average 5 degrees Fahrenheit temperature bins from 0 to 100 (60 to 65
degrees Fahrenheit is the omitted category), where ?????? indexes the bins.
▪�
??????� is a vector of controls for daily precipitation and fully interacted two- and seven-day temperature
lags plus leads.
▪�
???????????? and �
�?????? are ZCTA-day-of-year and state-year fixed effects. Each observation is weighted by
ZCTA Medicare enrollment. Standard errors are clustered by ZCTA.
Estimating Homogeneous Effects of Temperature Bins

8. 7 ED = emergency department; IP = inpatient; OP = outpatient; ZCTA = zip code tabulation area.
Change in Three-Day Emergency Department Visits per
100,000 Fee-for-Service Medicare Enrollees
All ED visits increase with
higher-temperature days.
ED visits decrease with
cooler temperatures in the
short term. OP visits drive
that effect, because IP visits
are more complex and
therefore more unavoidable.

9. 8 ED = emergency department; IP = inpatient; OP = outpatient; ZCTA = zip code tabulation area.
Change in Three-Day Emergency Department Spending per
100,000 Fee-for-Service Medicare Enrollees
ED spending generally increases
with higher temperatures. The
increase is greater for IP visits than
for OP visits, which is consistent
with the greater complexity and
cost of IP visits (which therefore
result in a hospital stay).
On cold days, an increase in IP
spending drives an overall increase
in ED spending. OP spending
decreases slightly on cold days,
suggesting that the decrease in ED
visits may have been concentrated
in lower-cost visits.

10. 9 CDDs = cooling degree days; ED = emergency department.
Using Differences in Local Average Climate to
Account for Adaptation
Extreme temperatures do not affect all
areas equally because of different
levels of adaptation: For example, a
daily temperature of 90 degrees
Fahrenheit is likely to be more harmful
in Minneapolis than in Dallas.
CBO captures that difference by
allowing the effect of each temperature
bin to differ on the basis of a ZCTA’s
average climate.
CBO computes the average annual
cooling degree days (CDDs) for each
ZCTA and estimates a separate linear
spline in CDDs for each temperature
bin. Annual CDDs capture how often
and by how much the daily
temperature surpassed 65 degrees
Fahrenheit.

11. 10 ED = emergency department.
Change in Three-Day Inpatient and Outpatient Emergency
Department Visits for Three Illustrative Cities
Despite the overall pattern of
fewer ED visits on cold days,
places with warmer climates
experience more ED visits
when the temperature is
below 20 degrees Fahrenheit.
Places with mild and cold
climates experience more ED
visits on hot days.

12. 11 ED = emergency department.
Change in Three-Day Inpatient and Outpatient Emergency
Department Spending for Three Illustrative Cities
Spending increases
substantially at the
extremes, with significant
heterogeneity by climate.

13. 12
CBO takes the unweighted average of NEX-GDDP daily downscaled projections from models reporting a full year of projections. Dallas’s CDDs in 2075 are estimated to be 3300,
exceeding the limits of this figure. CDDs = cooling degree days; ED = emergency department; SSP = shared socioeconomic pathway.
Projecting the Effects of Climate Change
CBO computes the projected change in
the future temperature distribution
relative to the 2015–2019 average and
applies that change to the agency’s
historical weather data.
To account for adaptation, CBO allows
a ZCTA’s temperature response to vary
in the future on the basis of its projected
CDDs using spline estimates.
For example, Minneapolis is gradually
allowed to have a temperature
response by 2075 that approaches the
response of Washington, DC, today.
CBO’s central climate scenario is
SSP2-4.5, but the agency also presents
cases with low (SSP1-2.6) and high
(SSP3-7.0) emissions and warming.

14. 13 The figure shows the change in IP and OP visits relative to the 2015–2019 average. ED = emergency department; SSP = shared socioeconomic pathway.
Projected Change in Annual Emergency Department Visits per
100,000 Medicare Enrollees
Average annual ED visits per
100,000 enrollees from 2030 to
2075 are projected to increase
by an average of 250 under
SSP2-4.5. Under alternative
scenarios, those visits are
projected to increase by 175
(SSP1-2.6) to 285 (SSP3-7.0).
Without adaptation, average
annual ED visits are projected
to increase by 415 per 100,000
enrollees.

15. 14 The figure shows the change in IP and OP spending relative to the 2015–2019 average. SSP = shared socioeconomic pathway.
Projected Change in Annual Emergency Department Spending per
100,000 Medicare Enrollees
Average annual ED spending per
100,000 enrollees from 2030 to
2075 is projected to decrease by
an average of $750,000 under
SSP2-4.5. The projected decrease
in ED spending is driven by a
decrease of $870,000 in OP
spending that is partially offset by
an increase of $120,000 in IP
spending.
Under alternative scenarios,
average annual ED spending per
100,000 enrollees is projected to
decrease by $510,000 (SSP1-2.6)
to $810,000 (SSP3-7.0). Without
adaptation, average annual ED
spending (IP and OP) is projected
to increase by $48,000 per
100,000 enrollees.

16. 15 Total cumulative changes are based on Medicare enrollee projections and on the sum of five-year incremental estimates from 2030 to 2075.
▪Average annual ED visits (IP and OP) are projected to increase by 235,000 per
year from 2030 to 2075.
–About 1 percent of total 2019 ED visits
▪Cumulative changes in ED spending (IP and OP) through 2075 are projected to
decrease by roughly $7 billion.
–An increase of about $1.1 billion in OP spending and a decrease of about
$8.1 billion in IP spending
–An average annual decrease of $700 million, or about 1 percent of total
2019 ED spending
▪Cumulative mortality through 2075 is projected to decrease by roughly 320,000.
–An average annual decrease of about 32,000, or about 2 percent of total
2019 Medicare mortality
Preliminary Projected Outcome Summary for SSP2 -4.5,
2030 to 2075

17. 16
▪Extend the outcome observation window to a 28-day sum
▪Examine differences in mortality effects by subgroups:
–Age
–Race
–Dual enrollees (Medicare and Medicaid)
–FFS vs. Medicare Advantage
▪Perform sensitivity analyses of adaptation:
–Medicare enrollment over time
–Alternate spline specifications
–Delayed adaptation in projections
Next Steps

18. 17
Tamma Carleton and others, “Valuing the Global Mortality Consequences of Climate Change
Accounting for Adaptation Costs and Benefits,” Quarterly Journal of Economics, vol. 137, no. 4
(November 2022), pp. 2037–2105, https://doi.org/10.1093/qje/qjac020.
Tatyana Deryugina and others, “The Mortality and Medical Costs of Air Pollution: Evidence From
Changes in Wind Direction,” American Economic Review, vol. 109, no. 12 (December 2019),
pp. 4178–4219, https://doi.org/10.1257/aer.20180279.
Carlos F. Gould and others, Temperature Extremes Impact Mortality and Morbidity Differently, NBER
Working Paper 32195 (National Bureau of Economic Research, March 2024), pp. 4178–4219,
https://doi.org/10.3386/w32195.
GarthHeutel, Nolan H. Miller,and David Molitor, “Adaptation and the Mortality Effects of Temperature
Across U.S. Climate Regions,”Review of Economics and Statistics,vol. 103, no. 4 (October 2021),
pp. 740–753, https://doi.org/10.1162/rest_a_00936.
Corey White, “The Dynamic Relationship Between Temperature and Morbidity,”Journal of the
Association of Environmental and Resource Economists, vol. 4, no. 4 (December 2017),
https://doi.org/10.1086/692098.
References