As a volunteer in a trial of mobile health technology, I can attest that it's incredibly cool to pick up your iPhone, fire up an application to monitor your heart rate and rhythm, and then beam your ECG reading to a cardiologist halfway around the globe. As a physician-scientist, I also know that cool technology is not necessarily synonymous with good science or sound health practices and that therein lies a challenge.

The use of cell phones and wireless sensors to gather and access health data has grown quickly in recent years. Popular mHealth apps are used for counting calories, gauging nutrition, tracking workouts, calculating body mass index and quitting smoking. These worthy efforts pale next to the potential of mHealth to aid in medical research and health care.

Mobile devices offer remarkably attractive low-cost, real-time ways to assess disease, movement, images, behavior, social interactions, environmental toxins, metabolites and a host of other physiological variables. Many mHealth technologies could be put to highly innovative uses in biomedical research; at the same time, biomedical research could help build the foundation of evidence that so many mHealth applications currently lack.

Because mobile devices are miniaturized and require little energy to operate, they have the power to bring the research laboratory to the patient in ways never before possible. For instance, clinical trial participants can avoid the inconvenience of visiting research facilities, writing down their daily activities or wearing clunky monitors. Scientists would also get higher-quality data because written diaries and questionnaires about exercise, diet, pain, and so forth, are notoriously unreliable. Real-time continuous biological, behavioral and environmental data can greatly improve understanding of the underlying cause of disease. Combining mHealth data with GPS data could also lead to early detection and warning systems for outbreaks of illnesses related to environmental exposures or infectious agents.

Wireless sensors could help scientists keep track of sleep patterns at home, instead of their having to rely on lab-based studies or self-reporting. Doctors could monitor blood pressure during daily activities, which is when it matters most, rather than in a clinic. Washable tattoos embedded with nanosensors could take blood glucose and sodium readings for transmission via a smartphone.

To make all this happen, health researchers, technology developers and software designers must pull together to find ways of evaluating new technologies. The National Institutes of Health is working to build the interdisciplinary research capacity needed to establish an evidence base for the benefits and risks associated with mHealth technologies.

Maintaining privacy and security of health data is a challenge that calls for research. How do we protect trial participants and ordinary consumers without adversely affecting research and quality of care? Who will set rules for privacy of mHealth data? Who will provide protections if privacy is breached?

We must also learn how people are actually using mHealth in their everyday lives. I suspect that, right now, the majority of users are much like me, treating their new apps as gee-whiz toys rather than as valuable tools for improving their health. I am convinced, however, that the real potential of mHealth lies with much more committed users, such as the children with type 1 diabetes who took part in a yearlong, case-control study of wireless technologies to monitor and manage blood glucose levels. That study, published in Diabetes Care, showed that youngsters who used the automated system had significantly better glycemic control and diabetes self-management skills than those who did not. Now that's an mHealth moment worth getting excited about.

SCIENTIFIC AMERICAN ONLINE
Comment on this article at ScientificAmerican.com/jul2012

7 Comments

1. 1. john97john 12:55 PM 6/21/12

   June 20, 2012
   
   
   Dear Dr. Collins,
   
   In your lead article on pg. 16 of the Scientific American July issue, you note that (mHealth) applications need evidence-based information if they are to be trusted and used. Abu-Mostafa in his machine-learning paper in this same issue notes that the NetFlix $1 million prize data-mining competition was won with a 10.6% improvement in the statistical accuracy that NetFlix could make in predicting what movies their customers would like next.
   
   Accuracy in evidence-based information is the overall need in medicine. Abu-Mostafa and the world's best statisticians could only produce a 10.6 % accuracy improvement result using statistics in that $1 million, world wide NetFlix competition. I would like to urge NIH to set a standard for accuracy in your evidence bases, of something like 90% at least, for medical data.
   
   My perspective, enhanced by Abu-Mostafa's paper, is that statistics will never achieve even 90% accuracies, regardless of the size data that is generated. In genetics with sequenced data, something like dimensionalities of 3Exp 9 will be encountered. Statisticians don't like huge dimensionalities. In Supervised Learning applications, an analytic linear-algebra vector-fusion algorithm in combination with Quaternion Algebra appears, in my usage, able to give 100% accuracies when “fitting” smaller data with complex coefficient polynomials of smaller degrees.
   
   FYI: I am the guy who, at your Utah Population D.B. Celebration talk some years ago, asked how one could collect all the environmental data you had said was needed to be collected. Quality of the data is a big issue in accuracy.
   
   Robert R. Johnson
   Prof. Emeritus, Computer Science, U/U

   Reply | Report Abuse | Link to this

2. 2. mHealthTalk 09:58 PM 6/25/12

   I enjoyed your article, Robert, but this one piece caught my attention. "Health researchers, technology developers and software designers must pull together to find ways of evaluating new technologies."
   
   What about the consumer/patient? We need to think of ways to help them find and evaluate useful mHealth apps too, probably through social media that encourages individuals to share their discoveries and experiences with others. These ePatients already use the many discussion forums in websites like PatientsLikeMe.com, but I've found that the active groups are very specific to a particular condition.
   
   Seniors often suffer from multiple medical conditions and physical or cognitive limitations, and that's why I started Modern Health Talk (mHealthTalk.com) -- to focus on tech solutions that help the elderly age-in-place as long as possible. Many of these solutions involve mHealth apps or rely on smartphones or tablets as home health gateways to connect different sensor devices to remote monitoring services.
   
   I write about all of the mHealth technologies you mentioned from a consumer perspective, but I've not yet found the right social media formula for engaging consumers in a healthy dialog of tech solutions they found too.

   Reply | Report Abuse | Link to this

3. 3. IngridMadden 09:48 PM 6/26/12

   I was especially grateful to see the mention of passively collected data in this article. The non-invasive collection of continuous streams of data from smartphone users is still widely unexplored in the field of mHealth. The passive observance of temporal, spatial, and contextual markers of behavior over time has the potential to better inform doctors with real, personalized evidence - and requires little to no effort from the patient. As long as a mobile device is on and always with the user, the embedded tools already built-into smartphones can automatically monitor human rest and activity; thus, making it possible to identify health and disease indicators like never before.
   
   Ingrid Madden
   www.OpenmHealth.org

   Reply | Report Abuse | Link to this

4. 4. gary_johnson_53@hotmail.com 10:53 AM 7/10/12

   I cringe whenever I see BMI mentioned in the Press. According to the BMI, Tom Brady, the quarterback for the New England Patriots is obese. There are far better ways of estimating healthy weight. I built a body fat estimator based on the Navy calculations at http://garyjohnsoninfo.info/FoodFactsHintsTips/BodyFat.html
   You have to measure height and weight and 3 additional size measurements for men and an additional one for women; however, the results are far more accurate.
   
   mHeatlh apps could be a very useful tool. I would advocate that we trade off a little less ease of use for higher accuracy. There is an app that takes your heart rate by using the camera of your smart phone. There are a few simple procedures (example http://www.nemahealth.org/programs/healthcare/heart_rate_pulse.htm#2) which can give you better results.
   
   Combine taking your Body Fat Calculations or manual heart rate estimations with 5 minutes of training or comparing results with an accurate measuring device.
   
   Compare the usefulness and accuracy of the results to the BMI and a camera generated heart rate.
   
   I hope you can see why I cringe.
   

   Reply | Report Abuse | Link to this

5. 5. ramiejohn 02:11 PM 7/11/12

   The comments by Robert Johnson and Ingrid Madden hit on several points that I agree are necessary for a bright future for mHealth:
   
   • “(mHealth) applications need evidence-based information if they are to be trusted and used.”
   • They require a high predictive accuracy (not merely the 10% rate of Netflix).
   • Supervised learning —and artificial intelligence algorithms— improve upon the accuracy provided by statistical approaches.
   • The promise of mHealth is in personalizing medicine, and because medicine is an ‘art’, there must always be a patient participatory component.
   
   Although Robert voices his concern that ‘data quality is a big issue in accuracy’, I suggest that even ‘dirty’ data can be helpful. In the U.S., we report adverse drug events in what is called the MedWatch program, but it is underutilized and reports are often incompletely filled out. The data in this warehouse is used by the FDA to monitor real world drug problems. Until we get, as Ingrid suggests, a passive collection tool, the MedWatch database is the main source for performing pharmacovigilance or post-marketing surveillance. However, John M. Armstrong, Ph.D. and myself (a Ph.D. immunologist) have developed a predictive analytics tool called Medloom that uses an advanced form of artificial intelligence to use this ‘dirty’ data to discover patterns. Not only do we generate profiles of patients who are at risk for bad health outcomes, but we can link this information to electronic health information systems. By cross-referencing patterns found in the real world of drug use with a particular patient profile, we can determine if the patient is at risk and provide that information to the doctor at the point-of-care. In a retrospective study, Medloom had a predictive accuracy of greater than 80%. Medloom also draws strength from the 4 P’s of personalized medicine: Personalized, Predictive, Preventive, and Participatory.
   
   Lately, I’ve embarked on an endeavor to bring this technology to a patient-facing mobile application at MedloomforME.com.
   
   Whether “Medloom for Me” achieves success as a mobile app will depend on support by the patient community. I wish, as does Ingrid, that there exists a central forum for patients to discover and discuss mHealth applications. Such a forum would go a long way towards streamlining adoption of the most effective apps.
   
   Dr. Ramie Leibnitz
   President of Lead Horse Technologies
   

   Reply | Report Abuse | Link to this

6. 6. Dylan Cicero 12:41 PM 7/19/12

   Dr. Collins, you bring up several thought-provoking questions about the legitimacy of mHealth. Indeed, there have been few controlled trials on the efficacy of health apps or connected devices besides the WellDoc’s diabetes study. Likewise, physicians and patients lack an objective way to sort through the thousands of apps in the healthcare marketplace to find the right apps to support clinical treatment. Still, we should not be dissuaded by the lack of clinical evidence in the field, thus far. A rapidly growing mHealth infrastructure suggests that mHealth holds great promise—and is here to stay.
   
   Recently, Happtique, a mobile health app management company, released guidelines for their app certification program. The guidelines, drafted by a physician, nurse, biotechnology scientist, and a patient advocate are meant to ensure high content, operability, security, and privacy for health and fitness apps. Such a process for app certification demonstrates the seriousness of using mobile for health. The guidelines are available for review and public comment(@http://www.happtique.com/wp-content/uploads/App-Certification-Standards-final.pdf).
   
   Happtique has also began a curation process, by which the (over 15,000) health and fitness apps are being cataloged by physician specialty and disease type. @Dr. Leibnitz and @mHealthTalk, this kind of catalogue will provide a simple context for physicians in the same specialty and patients with similar conditions to review apps.
   
   Other firms, too, are helping to prop up the mHealth infrastructure. Cloud computing and data analytic companies are eager to compile information from wearable devices and deconstruct it into meaningful components. Mocana Corporation is developing technologies to further guarantee patient data security transmitted through mobile applications. The Continua Health Alliance has created a standard to ensure the interoperability between connected health devices and healthcare IT. John’s Hopkin’s Global mHealth Initiative is underway, in which they are conducting a series of 49 studies to test the efficacy of certain health apps. And the mHealthZone radio show, powered by Happtique , continues to facilitate dialogue about the latest mHealth developments every Thursday at 12:00 noon (the show can be heard @http://www.mhealthzone.com/).
   
   We are not there yet, but the mHealth revolution is on its way.
   

   Reply | Report Abuse | Link to this

7. 7. MarieRaoulx 05:52 AM 11/19/12

   Hi,
   I'm contacting scientist as your readers :) because i have seen a funny app <a href="goo.gl/eZNwH">Heart Beat Rate</A>. It measures your pulse with the webcam.
   Do you know how it works.
   Thanks in advance for your reply
   Best regards
   Marie
   

   Reply | Report Abuse | Link to this