Coverage
What does the data cover?
Only 657 of Mexico's 2,457 municipalities host a hospital that reports emergency department visits, and reporting improves over the sample rather than decaying, from 411 municipalities in 2008 to 624 by 2021. Counted by where patients actually live, though, the records reach 2,442 municipalities, because residents of the other 1,785 travel to a hospital somewhere else.
Descriptive only
Nothing on this page is an estimate. The paper's effects are national, and it produces no municipality-level result, so no map here can be read as a local effect of temperature. These layers describe which places appear in the data and how much they report.
Which municipalities appear in the data
Municipalities reporting, by year
How these two layers were built
Visits are counted by the patient's own municipality of residence, so numerator and denominator describe the same people. Counting instead by where treatment happened would divide visits by the host municipality's population, which makes a small town with a regional hospital read as extraordinary: on that basis the median is 12,339 per 100,000 and the top reaches 327,786, describing catchment areas rather than health. Both versions are in the downloadable file.
The emergency department count is inferred, and checked
The archive identifies each hospital by its CLUES code, whose first two letters give the state but not the municipality, and it contains no crosswalk. Each facility is therefore placed in the municipality where most of its patients live, which for a typical hospital is 91 percent of them. Two things the inference never saw were used to test it: the state it implies agrees with the state encoded in the CLUES code for 99.1 percent of facilities, and 94.2 percent of the implied locations are municipalities that the visit records independently show to host a reporting hospital. The layer counts the 1,052 departments that reported at any point between 2012 and 2021 across 687 municipalities, not a snapshot of any single year.
How these layers were built
Both layers aggregate the deposited municipality-day panel: visits are summed by municipality and year, annualised to account for partly reported years, and divided by municipal population from the same archive.
Geometry is the 2012 INEGI municipality boundary file, projected to EPSG:6372 and simplified. Projection and simplification happen in the export script, not in the browser.
The residence layer covers 2012 and 2014 to 2021. Residence is absent from the 2008 to 2011 files, and the 2013 file records the state code in place of the municipality, so it is excluded. Between 5 and 10 percent of visits carry no usable residence and are dropped, 25 percent in 2012, which makes residence rates a slight undercount.
The rate layer uses quantile breaks, so each shade holds a comparable number of municipalities rather than an equal slice of a very skewed range. The coverage layer uses year bands instead, because most reporting municipalities report the whole period and quantiles would collapse into a single class.