With origins in the early 1990s and the inception of devices such as the IBM Simon Personal Communicator, early smartphone devices offered untethered mobile telephony, augmented by a humble suite of modular apps to extend basic phone functionality, hence the smart moniker in smartphone [1,2]. Nearly 3 billion people worldwide now use smartphones [3]. The release of the Apple iPhone in 2007 and subsequent competing Android smartphone offerings from Google and other vendors saw the emergence of platform-specific app stores, offering downloadable apps for a myriad of purposes [4]. Of the estimated 4.5 million apps available in the Google and Apple app stores, a million collectively pertain to health, fitness, nutrition, and well-being in general [5,6]. A subset of 300,000 of these apps may be regarded as bona fide mobile health (mHealth) apps, some of which may be potentially prescribed to patients for the diagnosis or treatment of health conditions [7]. Acknowledged by the World Health Organization in 2011, mHealth is defined as medical and public health practices supported by mobile devices, such as mobile phones, patient-monitoring devices, personal digital assistants, and other devices such as wearables [8]. More than 500 million people worldwide are believed to have downloaded one or more mHealth apps to their mobile phone [9].

Given the potential for health improvement arising from the availability and utility of built-in sensors in billions of smartphones worldwide, the purpose of this systematic survey was to identify smartphone mHealth apps using built-in sensors, offered by three popular contemporary international curated mHealth app repositories, and to assess which health conditions these apps address and whether they are available across different platforms [39,43,45].

This survey, conducted in October 2019, considered all mHealth app listings in the NHS Apps Library, AppScript, and MyHealthApps–curated mHealth app repositories (Figure 1) [46-48]. These libraries were selected as examples of government-funded (NHS Apps Library) and privately funded curated mHealth app repositories (AppScript and MyHealthApps) [4,15,36,41]. The latter two privately funded libraries differ in that MyHealthApps incorporates patient reviews in the curation process, whereas AppScript uses a proprietary scoring process [22,41,44].

All apps addressing health conditions using built-in mobile phone sensors to generate health data were identified using the publicly accessible search interfaces offered by each repository. As no search criteria were offered by these sites for filtering and identifying sensor-based apps, manual screening of the descriptions of all individual apps listed by each website was conducted by the lead researcher to screen for the use of built-in mobile phone sensors. Curated library descriptions for identified apps were inspected to categorize the purpose of the app as solely diagnostic, therapeutic, or a combination of both. Diagnostic apps were defined as those that identify the nature of a health condition, in contrast to treatment apps, which offered features for health condition management. Health conditions addressed by included apps were classified based on the description provided by each repository.

Apps using external or add-on sensors and nonsmartphone wearable device apps were excluded from this survey, as external components may impose additional cost, complexity, or excessive battery consumption, potentially reducing availability or accessibility to smartphone mHealth users [49]. Propeller is an example of an asthma therapy coaching app using an external Bluetooth sensor, which was excluded from this study [47]. Exercise, general activity, and accessibility apps were also excluded, as only apps used in diagnosis or treatment of specific health conditions were in scope of this survey. Runkeeper is an example of a GPS running tracker designed for monitoring exercise but excluded from this study, as no specific health condition was indicated for its use [47].

Mobile phone operating systems supported by included apps were noted, assessing the availability of these apps by users of the Apple iOS and Google Android phone types. The availability of included apps on advertised platforms was confirmed by following links advertised by each repository to inspect the Apple and Google app store app listings for apps. Apps were not downloaded or tested. App listings included as in scope by the lead researcher were then reviewed by the research team.

Low numbers across result groups precluded rigorous statistical analysis. Descriptive statistics were used where appropriate to illustrate results and to allow comparison between different libraries proportionate to respective library size.

A total of 1200 apps listed in the three selected curated mHealth app repositories were identified (Figure 1). Of 1200 apps, 780 nonduplicated apps were screened for eligibility. A total of 18 mHealth apps using built-in smartphone sensors were found in the three repositories surveyed. These represented 1.1% (8/699) to 3% (2/76) of the total app count across respective curated libraries (Table 1).

Details of smartphone mHealth apps using built-in sensors included from each respective curated mHealth library in this survey are presented in Multimedia Appendix 1. Listings include app type (ie, diagnostic, therapeutic, or both), sensor type used, app name, description, mobile phone operating system, and health concern addressed by each app.

Half (9/18, 50%) of all apps inspected were offered solely for use on the Apple iOS platform, with a further 11% (2/18) dedicated to the Android operating system (Table 2). Only about one-third (7/18, 39%) of the identified apps across all surveyed libraries were available for cross-platform use on Apple iOS and Android operating systems. The MyHealthApps repository offered the greatest cross-platform app availability, with 5 of the 8 (63%) identified apps in this library compatible with Apple iOS and Android operating systems. Most AppScript listings identified (6/8, 75%) were compatible only with Apple iOS.

Almost one-fourth (4/18, 22%) of all included apps were dedicated entirely to the diagnosis of health conditions (predominantly available in MyHealthApps), whereas half were solely treatment oriented (Table 3). The AppScript and MyHealthApps libraries offered comparable numbers of combined diagnostic and therapeutic apps using built-in sensors, whereas more apps dedicated to treatment were available in AppScript compared with the other libraries.

Camera (7/18, 39%) and touch screens (6/18, 33%) were the most frequently identified smartphone sensors used (Table 4). Microphones and accelerometers (and mobile phone speakers) were found to be less frequently used sensors in the identified apps. No GPS-based mHealth apps were identified in this survey. MyHealthApps offered more camera-based apps than the other libraries combined, whereas AppScript listed more apps using touch screens and microphones.

Smartphone cameras assessed pulse rate using photoplethysmography in an anxiety treatment app (Beat Panic), a respiratory therapy app (HeartRate+ Coherence), and a cardiac app (Instant Heart Rate). Beat Panic and Heart Rate+ Coherence are examples where smartphone pulse rate sensing is a secondary function to support a main therapy, namely, anxiety management and breathing exercise, respectively. Camera images were also used for automated skin cancer diagnosis (SpotMole) and in capturing images for dermatological diagnosis (UMSkinCheck, iDoc24, and MyPso).

In addition to capturing responses to speaker-generated tones in audiology testing, touch screens were used in vision training (Vision training 1 and Visual Attention Therapy Lite), neurological tremor assessment (pdFIT and Dexteria), and anxiety management (Chill Panda and Antistress Chromotherapy). Microphone sensors were used in several respiratory therapy apps (Breathing Zone, SnoreLab, and SnoreMonitor SleepLab). The use of a mobile phone accelerometer sensor was identified in a single app for neurological tremor assessment (LiftPulse).

Respiratory (4/18, 22%), dermatological (4/18, 22%), neurological (3/18, 17%), anxiety (3/18, 17%), and visual health (2/18, 11%) were the predominant health concerns addressed by the identified apps (Table 5). MyHealthApps and AppScript libraries listed more apps addressing a wider range of health conditions than the NHS Apps Library. The AppScript repository presented more apps for respiratory-related conditions, concerning snoring (n=2) and breathing exercises (n=2). Both the apps using built-in sensors in the NHS Apps Library addressed the management of anxiety.

Mobile phone cameras are employed in addressing the broadest range of health issues (Table 6). For example, skin cancer assessment camera apps are available in MyHealthApps and AppScript repositories. Diagnostic pattern-matching algorithms analyze acquired camera images of skin lesions in one app in the MyHealthApps library, whereas another from AppScript captures photos for later analysis by a physician. General dermatology apps using smartphone cameras to capture images are listed in the MyHealthApps library. In addition, two apps for the assessment of tremor were identified in the AppScript and MyHealthApps libraries, using the touch screen to assess touch accuracy in Parkinson disease symptom assessment and accelerometer sensors to detect tremor-induced movements, respectively.

Curation activities offered by third-party mHealth libraries, which are underpinned by medical device regulation, contribute to informing and protecting mHealth consumers. In this study, we surveyed three popular curated libraries regarding a specific subset of mHealth apps, namely, those using built-in mobile phone sensors for diagnosis or treatment of health conditions. Key aims of this survey included determining app availability, mobile phone operating system compatibility, intended purpose (diagnosis or therapy), types of sensors employed, and the range of health conditions where built-in smartphone sensors are used. First, this survey yielded a relatively small number of apps across the libraries examined, with differences found in the number of apps available between libraries. Second, more apps were available for the users of Apple iOS smartphones than for those of Android devices; cross-platform availability differed between the libraries surveyed. Third, the majority of apps offered treatment and combined diagnosis and treatment, with a smaller proportion offering dedicated diagnostic functionality. Fourth, cameras, touch screens, and microphones were the most frequently used mobile phone sensors in these apps. Finally, the range of health conditions addressed by these apps included respiratory, dermatological, anxiety, and neurological conditions.

Searching for apps related to particular health topics or medical concerns pose challenges for health professionals and consumers alike. Search engines, such as Google and Bing, which index available apps based on keyword search algorithms, often yield large volumes of uncurated search results for a given health topic [35,36,50,51]. Although the Apple and Google app stores categorize submitted apps for more focused searching (eg, Health and well-being), those searches can still return an overwhelming result list of indeterminate quality [4,35]. Search engines and app stores display star ratings and reviews to indicate the popularity of given apps, but these may not be reliable measures by which listed mHealth apps can be trusted [22,45]. In addition to high-level categorical grouping, third party–curated mHealth libraries offer more detailed subcategories and lists for specific health conditions and medical specialties.

No studies could be found that quantify the prevalence of mHealth apps using built-in sensors in curated mHealth app libraries. A 2016 review of health care–related apps available from the Google and Apple app stores identifies 80 clinical or health care–related mHealth apps for diagnosis or health monitoring [35]. Of the apps described in this review, mobile phone cameras are the most frequently employed sensor type, with camera images used by some apps for dermatological and ophthalmological diagnosis. Camera imaging is also employed in apps for blood flow monitoring by means of photoplethysmographic monitoring of pulse rate and estimation of blood pressure [35]. Emergent problems with blood pressure estimation received wide publicity when found to be unreliable in the case of at least one app [32]. In a number of health care contexts, app availability has been termed volatile, where apps may be removed from app stores in response to the revision of the underlying evidence base of an app or for medicolegal reasons [52].

Critics highlight a lack of transparency in standards applied to the screening of submitted apps before inclusion and hosting in popular app stores and search engines, resulting in mHealth app offerings, which may vary in quality or safety [7,45,51]. Measures of mHealth app quality have been developed (but not widely applied), including the (now defunct) Happtique Health App Certification, EU Kitemark certification, Intercontinental Medical Statistics (IMS) Score, and Mobile Application Rating Scale (MARS) [39,51]. For example, MARS evaluates mHealth app quality in five areas: aesthetics, functionality, engagement, information quality, and subjective quality [39,53]. Curated mHealth app libraries offer trusted sites for health consumers to select mHealth apps, constituting more detailed and specialized search portals than the aforementioned search engines and app stores [4,13]. Varying degrees of (proprietary) vetting are conducted to assert the safety and efficacy of curated apps, thereby imbuing search results with trust; detailed app scoring methodology and the incorporation of app quality measures, such as MARS, into the vetting process are not disclosed by these sites [40,54].

In contrast to a million health and well-being apps on offer to mobile phone users from popular app stores, only 18 mHealth apps using built-in smartphone sensors are identified in this survey, representing 1.50% (18/1200) of all mHealth apps collectively offered by the three curated libraries examined here (Table 1). A key consideration in the curation process is that medical device regulatory requirements may preclude listing of some apps in these libraries to prevent harm to app users [50]. Active mHealth apps (ie, those using sensors to gather health data) may be at greater risk of causing negative health impacts because of potential harm from inaccurate or incorrect data, demanding more rigorous curation and potential exclusion from curated libraries [23]. Regulatory authorities may require the assessment and accreditation of mHealth apps that offer diagnostic or therapeutic recommendations or those that transform the functionality of the mobile phone into that of a medical device [23,54]. Some sensor-based mHealth apps may use sensors as a secondary or supporting measure and thus not be regarded as medical devices per se [23]. Overall, two such examples are identified in this survey: anxiety and breathing exercise apps that use camera sensors for pulse detection as a secondary or indirect health data measure.

The respective smartphone market shares for Apple and Android devices are comparable [55,56]. In contrast, not all the apps identified in this survey are available across both popular mobile phone platforms, potentially disadvantaging some mHealth consumers. Half (9/18, 50%) of the apps identified in this survey are dedicated solely to Apple iOS, a further 11% (2/18) specific to Android, and only about one-third (7/18, 39%) available for both operating system platforms (Table 2). Apple iOS device manufacture is controlled solely by Apple, with relative homogeneity in hardware components such as sensors potentially offering app developers more stable or predictable target platforms for app development [56]. In contrast, Android devices may originate from multiple hardware vendors with disparate (sensor) hardware components, potentially adding complexity to the development of apps catering for a wider range of target device hardware and sensors [55,56].

Half of the identified mHealth apps (9/18, 50%) offer dedicated treatment features, with further about one-fourth (4/18, 22%) dedicated to diagnosis (Table 3). The imperative to seek (traditional doctor-patient) medical consultation regarding definitive diagnosis, potential risk of self-misdiagnosis, and regulatory restrictions may all contribute to the smaller proportion of purely diagnostic sensor-based apps offered by the libraries surveyed [57]. Dermatology and skin cancer diagnostic apps using the mobile phone camera constitute 4 of the 5 dedicated diagnostic apps identified.

Cameras and touch screens are the most frequently used sensors in the identified apps, followed by microphones and accelerometers. Apps using camera sensors are most prominent in the MyHealthApps library, whereas AppScript lists more microphone and touch screen apps (Table 4). Notwithstanding contemporary research studies regarding the use of GPS for activity tracking in mental health conditions such as bipolar disorder and general depression, no examples of translating this research into GPS-based sensor apps were found in any of the libraries surveyed [20,26]. Anxiety therapy was the sole focus of both apps identified in the NHS Apps Library. Respiratory concerns were the most frequently addressed health conditions in the AppScript library, whereas apps related to dermatology and neurological conditions were more prevalent in the MyHealthApps library (Table 5). Cameras are employed in a wider range of health conditions compared with other sensors (Table 6). Photoplethysmography is used to measure heart rate and heart rate variability in three camera apps, whereas photo capture for later inspection by a clinician is offered by two dermatology and skin cancer diagnostic apps. A single camera app performs steganography (pattern matching) for skin cancer diagnosis.

This survey found that mHealth apps using built-in sensors for diagnosis and treatment represented but a modicum of all apps found in the curated mHealth libraries examined. The nature and rigor of the curation process go some way to explain this observation, including the constraints of regulatory requirements for software deemed as medical devices. This may also help explain the smaller proportion of dedicated diagnostic apps observed in these libraries. Some health consumers may be disadvantaged by differences in the availability of apps on competing mobile phone platforms. Cameras, touch screens, and microphones were used most frequently in the surveyed apps. A limited range of health concerns were addressed by the surveyed apps.

Further efforts are needed to increase the availability of ubiquitous, low-cost mobile phone sensor technology in curated lists to assist with health conditions.