On a sweltering afternoon in Phoenix, a man collapses on a pavement that has reached 70°C. The paramedics arrive. The hospital records his death as cardiac arrest. Heat never appears on the death certificate — and so, officially, the heat did not kill him.

Multiply that moment by thousands, and you begin to understand what an NPR investigation has uncovered: America's heat death toll is dramatically, dangerously undercounted. The reasons for the gap are surprisingly mundane. Doctors are trained to record the immediate cause of death — a failing heart, struggling kidneys, a stroke — rather than the environmental trigger that pushed the body past its limit. Heat is what epidemiologists call a "contributing factor," and contributing factors have a way of disappearing from official paperwork.

Add to that a patchwork of state reporting standards, and the national picture becomes almost meaningless. Some counties count carefully; others barely count at all. The human cost of that paperwork problem is anything but abstract. Researchers who have compared death records with local temperature data — a method called excess-mortality analysis — consistently find that the real toll is several times higher than official figures suggest.