Coronavirus Disparities: Smart Thermometers

Increasingly, I am reflecting on the disparities in the coronavirus coverage, data sources, data analysis, and expected consequences. This is my attempt to begin tracking my observations.

Kinsa Health, the smart thermometer manufacturer, tracks fevers and monitors whether there is an atypical spike in fevers. Their company finds that fevers in their devices spiked in mid-March and has been decreasing since social distancing measures were enacted. This data of 162,000 temperatures nationally has been lauded as a harbinger of good news and evidence that social distancing works.

This dataset raises serious bias questions (aside from the obvious privacy concerns). Where are these thermometers distributed/purchased, and who within those counties has the thermometer? Street Bump, the Boston smart app to track potholes, already highlighted that smart apps privilege more affluent people and more affluent areas. Why should we believe that smart thermometers are any different?

This is map from Kinsa Health on April 2, 2020. According to the legend on the bottom, the red represents the High atypical fevers whereas the dark taupe represents No atypical fevers. Most striking to me is the light taupe, which represents the areas with too few thermometers to make a determination.

There are some enormous gaps in this dataset, especially in rural and predominantly minority areas. Compare the map below with the Census 2010 maps of the Black and Hispanic population.

We can see that western Texas and up through the Dakotas are fairly devoid of smart thermometers, as is the interior of the South. Let’s compare that to the Black and Hispanic population by county:

In the interior of the south, we see majority Black population. And insufficient fever information.

In the western part of Texas, we see majority Hispanic population. And insufficient fever information.

So… there’s directional evidence that the smart thermometer data excludes areas with high minority population.

Even within the counties with sufficient devices to make a determination about fevers, we do not know who has the devices. Considering the thermometers range from $35.99 to $69.99, it’s likely affluent people who own smart thermometers and who can also stay at home. Black and Hispanic/Latinx workers, who are less affluent on average and are less likely to have remote work options, may not be represented in the smart thermometer dataset.

This dataset is valuable to understand the pandemic in real-time. But it does not cover the entire population, especially people most worried about — and most vulnerable — to the virus.