UBS Evidence Lab Uses Hospital Data to Profile Regional Recoveries

At the beginning of the pandemic, geolocation data played a key role in measuring the outbreak of the Covid-19 disease, and provided something of a warning system for firms as to where the next series of lockdowns were most likely to unfold.

While geolocation data is still being used to monitor for second- and third-wave outbreaks, it is also being used to highlight a return of business for regional hospitals, says Jeremy Brunelli, global head of frameworks at UBS Evidence Lab, the independent research unit separate from UBS Research.

The team is monitoring hospital admittance by measuring the foot traffic entering and leaving hospitals, looking at factors such as inpatient and outpatient volumes, emergency room numbers, total surgeries, and elective procedures. It is partnering with an undisclosed data vendor that cleanses and anonymizes the hospital traffic dataset.

“This is an exclusive dataset that we have, and we get about two reads every month. We have an initial read, which is not the full sample, but it gives us an early indication, and then we get a final read at the end of the month. This is a dataset where we can see a dramatic decline in the volumes [of elective procedures and surgeries] driven by Covid-19,” he says.

So, for example, in Washington State, which was the first epicenter of the coronavirus in the US, hospital admittance data showed a significant drop in the number of non-emergency procedures back at the beginning of the pandemic, but those numbers are now improving. In April, there was a 70% year-on-year decline in non-emergency medical procedures, May saw a 50% decline compared to the previous May, June saw a still robust 22% decline, but July saw only a 7% decline, according to Brunelli.

The Evidence Lab uses the Bayesian inference approach to combine multiple datasets to see if they support or disprove a probable investment hypothesis. This is where alternative datasets on health care can be valuable, says Brunelli, as they can more accurately reflect world events than traditional market data.

“We would create this big mosaic of datasets and try to present it in a very digestible way for investors, so that they can look across this mosaic and start to increase or decrease their confidence around some hypothesis they have,” he says.

To supplement the hospital admittance data, UBS also conducts a proprietary survey, called a pulse check. The survey puts questions to 40 different C-level hospital executives in the US to learn about the current conditions, in terms of surgery procedures and capital expenditure plans, including funding spent on technologies and beds. Brunelli says that the two data products—the foot traffic and the survey—provide complimentary views on the events unfolding in hospitals

“There are still a significant amount of delays, but they’ve also improved as things have reopened and recovered, especially with orthopedic surgeries—we’ve seen 43% of hospitals are reporting delays versus 58% in May, and plastic surgeries and endoscopies have also seen a decrease in delays,” he adds.

Yet, these various healthcare insights are not used in isolation. Brunelli says they are most effective when used to compliment other foot traffic or mobility datasets that look at activity around retail shopping, public transit usage, and car traffic congestion, to help understand the pace of recovery in different locations.

For example, mobile device application publishers use Google’s software development kit and tie it to Google Maps, and embed it into their own technology frameworks. The location data of the person using the app is then pulled, aggregated, anonymized, and then sold to users like UBS, and they can more accurately see spikes in traffic or activity around retail centers. The Evidence Lab also takes in satellite data from location, navigation, and map technology provider TomTom to monitor auto congestion, and combines that with metrics on app usage from Sensor Tower, which provides data around mobile app usage.

“We use hundreds of vendors and also have, what we call, an internal harvest team,” Brunelli says, “where we web-mine thousands of different types of data from the internet, which we curate and create metrics with.”