The response

Does temperature send people to the emergency room?

A day below 10 °C reduces emergency department visits by 9.1% compared with a day between 20 and 25 °C. A day above 30 °C raises them by 3.7%.

The response on the day itself, and accounting for the following 30 days

Show:
-20-15-10-50+5Change in ED visits (%)≤10 °C−16.0%−9.1%10–15 °C−11.8%−5.3%15–20 °C−5.3%−2.7%20–25 °Creference25–30 °C+4.5%+2.5%>30 °C+2.7%+3.7%Daily mean temperature
same daycumulated over 30 days
Point estimates with 95% confidence intervals. The 20–25 °C interval is the omitted reference category, so every effect is measured against a day in that range.
Estimates behind this figure:m1_curve.csv
Full dataCode

How the effect accumulates over 30 days

Confidence interval for:
-20-15-10-50+5Cumulative change in ED visits (%)051015202530Days since the temperature day≤10 °C−16.0%10–15 °C−11.8%15–20 °C−5.3%25–30 °C+4.5%>30 °C+2.7%
Each line is the cumulative effect of one day in the given temperature interval, summed from that day through the following 30. The shaded ribbon is the 95% confidence interval for the selected interval only.
Estimates behind this figure:m1_lagpath.csv
Full dataCode

What is being estimated

The outcome is the rate of emergency department visits per 100,000 people, by municipality and day, for the universe of Mexican public hospitals from 2008 to 2021, and the model is a Poisson distributed lag specification with 30 lags.

Effects are measured against a day between 20 and 25 °C, the omitted reference category. Every specification absorbs municipality-by-year-month and municipality-by-weekday fixed effects, and controls for precipitation and relative humidity with the same 30 lags. Standard errors are clustered at the municipality level.

Percentages are 100 times the estimated coefficient, which is how the paper reports them.