What healthcare can learn from Big Weather
When it comes to analytics and big data in healthcare, many organizations struggle to understand what data they have at their disposal, never mind analyzing it and reaping the benefits.
It's easy to get lost trying to make sense of the ever increasing volumes of data being collected inside a healthcare organization, but by recognizing the interconnectivity of data, it's possible to reap direct rewards from understanding the correlation to data outside the organization.
Take the weather for example. Weather affects everyone and predicting it has always been both a challenge and fascination. Meteorologists use sensors to track atmospheric behavior that are combined with models to produce event notifications of extreme weather.
A typical alert message from the National Weather Service looks like this:
Severe Weather Statement
Alert: ...A SEVERE THUNDERSTORM WARNING REMAINS IN EFFECT UNTIL 915 PM CDT
FOR WESTERN GENEVA AND SOUTHERN COFFEE COUNTIES...
AT 850 PM CDT...DOPPLER RADAR INDICATED A SEVERE THUNDERSTORM CAPABLE
OF PRODUCING DAMAGING WINDS IN EXCESS OF 60 MPH. THIS STORM WAS
LOCATED 8 MILES SOUTHWEST OF ENTERPRISE...MOVING SOUTHEAST AT 30 MPH.
LOCATIONS IMPACTED INCLUDE...
BATTENS CROSSROADS...CENTRAL CITY...SELLERSVILLE...ENTERPRISE
MUNICIPAL A/P...MARL...EUNOLA...TURNER CROSSROADS AND WEEKS
In order for a healthcare provider to automatically derive a meaningful impact from this message, it needs to be parsed and mapped into a location grid. Then, the healthcare provider needs to assess if they have any workers or patients in those locations and start implementing the necessary arrangements if so.
Processing these messages is computationally costly for businesses and knowing which events to 'listen' for is an unnecessary burden for already overtaxed legacy computer systems. Thankfully there are alternatives available that healthcare providers should consider.
Weather forecast models in the U.S. typically have a horizontal resolution of 12 kilometers, meaning they are based on data gathered from grid points spaced 12 kilometers apart. This level of poor resolution means a hospital may have to alert their services in an entire county when in reality there may only be a localized flash flood. Luckily this is changing with the help of select technology vendors.
IBM recently created a Big Data system that enables highly-targeted and locale-specific weather forecasting. The system, dubbed 'Deep Thunder', provides high–resolution forecasts for a region using a diversity of public data from the National Oceanic and Atmospheric Administration, National Aeronautics and Space Administration, U.S. Geological Survey, as well as private data from companies like Earth Networks. With accuracy and precision, IBM's Deep Thunder can deliver hyper-localized weather predictions up to three days in advance, with calculations as fine as one kilometer and as granular as every 10 minutes. In addition, Deep Thunder has demonstrated the ability to predict snowfall amounts accurately to the nearest inch, which can indicate the level of disruption expected to local infrastructure and transportation.
These capabilities prove even more significant for homecare providers, as they can now receive warning that bad weather is coming three days in advance – giving them plenty of time to prepare both patients and caregivers. Ambulance routing services are already wired into real-time traffic information and knowing where at risk patients are in the path of a weather event will provide for even greater optimization of scarce resources.
Internet of Things
A new Kickstarter project called StormTag bundles a collection of sensors – Temp, Barometric Pressure, Humidity and UV that can be sent to your phone which then reports the information using apps such as WeatherSignal from the team at Open Signal. This level of data collection will bring down the cost of weather event prediction even further in the future.
In addition to the StormTag project, the Internet of Things (IoT) has expanded the connected fabric of medical devices and patients to a new level where providers can track the location of their assets and patients in real-time. Hyper location tracking of medical devices such as wheelchairs, infusion pumps and defibrillators that might need to be mobilized in the event of an emergency can be routed to targets without having to first discover where they may have been put in storage.
As hospitals struggle to lower operating costs and remain competitive, the IoT in healthcare offers a way to tighten budgets and improve a patient's journey through a medical facility. For example, the explosion of crowdsourcing apps like 'Waze' that use the location and movement of cell phones to track traffic patterns in real-time, are opening the doors to new low cost methods of collecting data on a massive scale.
Consider the following scenario:
Hand hygiene remains a top issue in US hospitals as a major source of disease with about 100,000 people dying from hospital- or healthcare-related infections in North America every year. Through the use of the IoT and RFID-enabled patient and visitor badges that can be tracked through the hospital, providers can remind its wearer when to wash their hands and highlight the nearest sanitizer station.
This is just one common, yet easy way the IoT and the data collected from such connected devices can allow healthcare organizations to improve care and cut down on costs.
Listening to Social Feeds
Twitter and Facebook data 'fire hoses' can be used to filter crowd-sourced messages about weather emergency related events or natural disasters. In 2010, the Red Cross was able to target survivors of the Haiti earthquake based on cellular phone positions after they texted, called and tweeted for help.
Insurance companies are already listening on social data feeds for signals that may represent an event related to their policyholders. In addition to checking claims against historical data, insurers can use predictive models of the weather to alert policyholders before the event occurs.
Call to action
Healthcare service providers can make use of the 'small signals' from 'big data' by registering with businesses offering hyper-local data services whether they come from super computer projects like Deep Thunder or a simple crowd-sourcing of user observations made from cell phones.
Most of these services offer web service integration models for feeding alerts directly into existing monitoring dashboards. Doing so will allow healthcare providers to better prepare for disasters, while also benefiting from improved efficiencies and costs savings on a day-to-day basis.
Stuart Sim is a director in West Monroe Partners' Advanced Analytics practice, based in New York City.