Erik Reinertsen     Blog     Research     Teaching     Tech     Talks

Three ways mobile is transforming healthcare


Mobile technology (i.e. smartphones) is transforming healthcare in three key ways:

  1. Ubiquity
  2. Context
  3. Computation

Note: Mobile technology applied to healthcare is referred to as “mHealth”.


Ubiquity

The ubiquity of smartphones means we can bring new healthcare technologies to everyone (quickly, too). Such technologies enable communication between providers and patients, behavioral monitoring and change via applications, and data gathering via sensors.

Today there are over 2 billion smartphones. About 4 billion people have a mobile phone of some kind, which is 80% of the earth’s adults. This is a technology that reaches almost everyone in the world:

We have an unparalleled opportunity to help the poor and underserved via mobile technology.

Also, smartphones are not just ubiquitous among people, they are universal within peoples’ lives. We use smartphones more than any other device, and spend ~50% of our day on media and communication. Smartphones are our remote control for the world. They define how we interact with each other, purchase goods and services, and access information. Smartphones even define how we interact with other technologies. Benedict Evans writes how “the smartphone is the Sun and everything else orbits around it” (i.e. smart watches, tablets, thermostats).


Context

Smartphones have sensors that provide rich contextual data about a patient’s life. Using these capabilities to continuously monitor patients is important because of a concept called “hovering”. Asch, Muller, and Volpp write:

…even patients with chronic illness might spend only a few hours a year with a doctor or nurse, but they spend 5000 waking hours each year engaged in everything else — including deciding whether to take prescribed medications or follow other medical advice, deciding what to eat and drink and whether to smoke, and making other choices about activities that can profoundly affect their health.
Source: NEJM 2012

Because of these sensing capabilities, providers can now measure a patient’s blood pressure, heart & respiratory rate, peripheral oxygen saturation, ECG data, blood glucose level, sleep & movement activity, location, and more. The rich lifestyle context from this data could help patients and clinicians work together towards better health goals.

Unfortunately, providers face the challenge of “information overload”. If everything is important, nothing is the priority. This challenge will only grow as we start to incorporate more complex data such as genomics, which is especially relevant in oncology or rare genetic diseases. Sophisticated computation becomes necessary to simplify complex data into actionable insight that helps providers and patients make decisions.


Computation

Computation is critical to make sense of the huge amounts of data from mobile platforms because both these data and the underlying human physiology they reflect are complex. Thanks to advances in internet connectivity, mobile brings incredible computing power to everyone’s phone. This enables new types of healthcare technology.

Researchers at UCSF built an Apple Watch app to track the wearer’s heart rate. On the surface, the simple interface masks complex algorithms that spot warning signs from heart rate. Simply setting off an alarm when heart rate exceeds or undershoots a fixed value would not be useful because each patient has a different baseline level of healthy heart activity. Also, cardiovascular activity differs for an individual depending on their circadian rhythm, other health issues, physical exertion, etc.

Some algorithms are computationally intense and must run “in the cloud” on powerful servers. But because of mobile connectivity, providers can obtain health data from a patient and analyze it on their phone while the server “crunches the numbers” both quickly and invisibly.

Several types of computation are important for mobile health. Signal processing removes outliers and noise from mobile health data, which could hide important trends. Machine learning and statistics enables researchers to classify outputs based on inputs. For example, a machine learning algorithm can estimate the probability of mental illness using heart rate and physical activity as input data.

Computation is not the last step though. mHealth needs to improve how providers engage patients as they take action together. For example, imagine a mobile application to assess risk of dental caries by processing pictures taken by the smartphone. The output of the app should be more than just a probability of a carie. Ideally, the app empowers the patient (and/or provider) to make a decision about care, perhaps by recommending dentists that accept that patient’s insurance. If the patient cannot afford insurance or pay for dental care, the app could guide the patient to a free dental clinic.

Computation is critical for mHealth because it enables the greatest opportunity: to improve how patients connect to the “physical” world.