The use of cell phones by community health workers and other medical practitioners in low-income countries has been promoted as a potential revolution for health systems development. This “mHealth” revolution has been seen as an opportunity to develop diagnostic, treatment and surveillance networks wirelessly, to build mobile apps allowing remote nurses and doctors to provide higher-quality care to rural patients even in places without a hospital or well-functioning health clinic. Several foundations are now offering grants to build and distribute phone applications that will offer everything from prescription drug advice to epidemic surveillance tools. But is mHealth really going to improve health outcomes? Or is it just another technological bomb thrown at poverty and poor infrastructure?
Globally, about 3.1 billion people used mobile phones in 2007; that’s nearly half the planet. The greatest growth during the last decade has occurred in Asia, the Middle East, and Africa. In many of these continents, mobile phone subscribers outnumber fixed-line telephone subscribers, particularly as countries leap-frog over the traditional development step of planting land-lines and rely instead on building wireless communication towers and Internet-based businesses.
In theory, this access to the Internet and specifically to mobile phone technologies should advance healthcare delivery. Doctors were one of the first groups, in Western countries, to adopt the original smartphone precursor–the Palm pilot (remember that attractive little gadget?)–and use it routinely to look-up prescription drug dosing or drug interactions. Now medical offices commonly use tablets or connect smartphones to servers to lookup medical records, display radiology images and send communications between providers and patients (on a typical day, I receive a half-dozen emails from patients, and login to our electronic medical record via an Internet browser to get test results or refill prescriptions). But why should this technology be restricted to the United States or other wealthy countries?
When posed with this question, a number of foundations have sponsored mHealth projects for poorer regions, such as designing mobile phone applications that can help roving community health workers to determine a diagnosis or send a message to a distant physician for advice. Such technologies have also included providing education and medication adherence reminders directly to patients. But while this sounds good in theory, the results in practice have been surprisingly mixed, for reasons we’ll discuss.
A number of mHealth initiatives have been commonly cited as being models for poor countries. In Mexico, Vidanet and CardioNet have been popular mobile text messaging educational applications, through which people receive messages encouraging them to exercise and eat healthy, or to reduce their risk of HIV infection. A related series of HIV-focused mHealth applications deployed in sub-Saharan Africa check-in with patients via text message to monitor how they’re feeling. A health worker responds to patient concerns or if a patient fails to respond. Similar mHealth apps send text message reminders to patients to take their prescribed treatments. Another set of applications, such as ResultsSMS, connect patients to their medical records, in order to disseminate test results and remind patients of their follow-up appointments.
There is no doubt that the reach of such patient-centered programs has been extensive. An HIV educational text messaging program in India, for example, reached 9 million handsets in just over a year. It’s not clear, however, whether any health benefits resulted from these reminders; so far, most studies have been limited to proximal end-points (like whether people received the messages and contacted a provider if they were concerned) rather than hard end-points like mortality. Some randomized trials to test hard outcomes are in preparation. Early studies have typically found no clear disease control or mortality benefit.
Nevertheless, the provider-centered apps have found greater obstacles to adoption than patient-centered ones. Mobile phone apps for village health workers were found to have limited use in a number of trials. Extensive apps involving remote data collection and feedback went essentially unused in several pilot studies among different regions. And direct patient-provider communication apps seem to have faltered in Asia and elsewhere.
The limits to Nokia
According to a series of surveys and a new data repository created by the World Health Organization (WHO), the differences between successful and unsuccessful mHealth programs seem to be determined in part by their target audience and the underlying habits of healthcare providers and patients being served.
The picture of mHealth that is being peddled by many foundations is that community healthcare workers or patients themselves will use smartphones like an iPhone to communicate with remote doctors and decide how to pursue treatment for a condition in a remote setting. Even more far-fetched is the idea of continuously monitoring patients using biosensors or triggering epidemic outbreak detectors at the WHO and CDC when patients report symptoms of a killer disease through their smartphones. Some have even suggested that a pregnant woman without a midwive could use her smartphone to safely deliver an obstructed baby.
These ideas would make a great episode of MacGyver. But the capabilities of most mobile phones in remote settings are generally limited to plain text messaging. Web browsing, GPS navigation, and email through smartphones is not as well developed in most low- and middle-income countries; where it is available, it remains expensive and slow, and easier to perform through laptops than through phones. Pedometers or more advanced body monitoring is likely to remain limited to over-eager health nuts or research studies.
The technological infrastructure will likely improve in most countries, but what is less likely to improve is the degree of treatment access facilitated by the technology among the poor. What studies have been done on mHealth initiatives seem to show is that those people who need care the most are the least likely to have a mobile device; hence, those patients who are most informed about their health, and least likely to be unhealthy, are those who are accessing these health-related text messages and pill reminder systems. Mobile penetration at present is less than 10% in the world’s rural regions, and the large increases being seen in places like India and southern Africa are generally among the middle- to upper-class urban populations.
More access doesn’t mean more productivity
Among healthcare providers being provided applications, a second problem has been that offering new tools does not necessarily result in greater healthcare access or better quality of healthcare for patients. There are a series of humorous ethnographic studies of Internet use among workers, showing that having people gain access to communication technologies during work may not make them more productive, but instead just increase their likelihood of surfing the nytimes homepage. In the rural healthcare context, even if people are not distracted by the technology and use it to help them work, we still need physical people to see patients, whether or not they have a phone on hand or a paper manual to help them diagnose a rash or infection.
The common theme in a number of mHealth studies is there is a disconnect between developers of mobile apps and public health or healthcare experts who know what is relevant and useful for improving healthcare delivery. For example, a recent mHealth app was promoted to an infectious disease clinic to help remind providers about a particular type of vaccination; the creators of the app were ill-informed, not realizing that the patients at an acute infectious disease treatment clinic, by definition, would almost entirely be ineligible for the vaccine. A similar problem was found in diabetes self-management apps for patients; a recent review found that most of these apps gave advice that didn’t correspond to evidence-based medical guidelines.
What seems to be most commonly found in mHealth assessments is that highly-motivated clinics and skilled clinical ‘geeks’ who are already intrigued by technology and concerned with improving infrastructure to include electronic medical records and good communication tools are adopting the mHealth applications–typically applications to help with the dosing of pharmaceuticals (like ePocrates), organizing of healthcare data (e.g., patient charts), and transmission of clinical questions to outside experts. But the counterfactual question is whether a mobile phone is needed at all for these tasks, as opposed to laptops and the Internet, especially if a highly motivated healthcare worker could just use a pocket pharmacopea or reference manual rather than an expensive-to-produce mobile app. That is, getting providers cheap access to the Internet and low-cost subscriptions to key websites like UpToDate (the main repository of updated medical guidelines) may be more useful than creating designer applications.
Conversely, where there is little motivation or incentive to adopt new mobile apps, healthcare workers are not generating the momentum to do so. This problem is being called the “Google Health” phenomenon, after the flopped attempt by Google to introduce an electronic medical record system on the Internet. Google thought that by creating a simple, freely-available medical record system, it would be widely adopted. While that may benefit public health, and be a great app, the political and social barriers to widespread adoption were too high for seemingly little gain, or hidden gains that are not always material and obvious to patients or providers. There was so much incentive from for-profit firms to build billing systems (systems that keep track of how to charge for a visit) rather than good medical records systems (which keep track of medical conditions and their management) that no one adopted the Google approach, and most hospitals and clinics still use billing-based records systems that have a few notes attached for the doctors, rather than a dedicated system focused on reducing medical errors and communicating well between offices and hospitals. The incentives are misaligned with need.
An alternative is to provide the same investment dollars to traditional health resources, e.g., textbooks/handbooks for providers, just as Partners in Health did during the rise of HIV and resurgence of drug-resistant TB. While low-tech, the manuals and handbooks required average healthcare workers to attend sessions and be directly supervised in their training and care, which mHealth technologies do not, by default, offer the opportunity for.
Trying to centralize the decentralized usually fails
mHealth apps are also being increasingly used for research studies. Many of these apps are unlikely to single-handedly produce significant mortality benefits, but they are really novel data collection tools. Systems like AESSIMS (a disease reporting app) and EpiSurveyor (for outbreak investigation) have been ported into broader databases like HealthMap, to keep track of disease reporting. But there is much concern that these systems will have inherent surveillance biases. That is, it’s not immediately clear that a cluster of mobile phone reports necessarily indicates an epidemic; as the ambulance drivers used to say, if we took as 911 calls as an indicator of real crimes, then the anxious little old lady who calls the police every night after seeing a dark shadow will be falsely interpreted as living in the highest-crime area of the city. Hence, these tools remain interesting epidemiological experiments, rather than being useful yet for real outbreak investigation or routine disease monitoring.
This is not to say that mHealth technologies are useless or won’t be widely adoped; they are just likely to be widely hyped but used by individuals in limited contexts as another tool for good practice rather than a panacea to global health problems. Like most technologies, they’re generally concentrated in the hands of those who already have resources, organized electronic health initiatives, and motivated and skilled staff. It will be difficult to construct a massively-well-coordinated universally-acceptable health communication infrastructure via mobile phones to provide rural healthcare in poor settings. More likely, especially as smartphones get cheaper, already dedicated health professionals will gain access to smartphones and use their own favorite apps like prescription drug databases or other quick references, and email with expert colleagues or have their patients subscribe to simple text-message-based reminders to take their pills. But it’s unlikely that we’ll see mHealth generate mass mortality benefits in the near future. Our joy of experiencing and creating new technologies may just outpace our need for them, or direct us towards the most-fun-to-use technologies rather than the most necessary ones.
Sanjay Bansu, MD, PhD, is a public health epidemiologist. His blog, EpiAnalysis, is a forum for public health epidemiologists who study global health data, healthcare policy, economics, and sociology.