Digital health in physiotherapy is an emergent area. It continues to gather speed with its use in both clinical practice and research growing exponentially [1-3]. Many digital modalities, with varied functions, are relevant to physiotherapy practice; these might include the delivery of timely digital patient information and resources through websites and patient portals [4], digital patient assessment using connected Bluetooth and wireless-enabled devices [5], digital models of care (telehealth) supported by video-based consultation [6-8], remote monitoring of patient status through wearables [9,10], and mobile health (mHealth) apps used to prescribe, monitor, and support home exercise programs [1,11].

Mobile apps are one type of digital health modality of particular interest because of the ubiquity of smartphone use and their ability to deliver digital therapeutics [12-14]. They are relatively inexpensive and thus scalable [15]. They are also worthy of scientific attention within the physiotherapy community because of their ability to support interventions targeting several aspects of behavior change, otherwise known as behavior change techniques (BCTs) [16-19]. Facilitating positive behavior change in people with health conditions is often central to the management approach of physiotherapists, aimed at deriving the best possible therapeutic benefit from an intervention. For example, behavior change may be required for a person to follow physiotherapist advice regarding activity pacing or to adhere to exercise and physical activity recommendations [1,16,19,20]. BCTs can be thought of as the components of an intervention that regulate the behavior by altering cause and effect [21]. As defined by Michie and Johnston [22], they may be thought of as the “active ingredients” that facilitate intended behaviors. Apps have the ability to incorporate and deliver numerous BCTs, which can mediate a behavioral target of physiotherapy care [23-25].

The app space continues to grow exponentially, with well over 300,000 health apps now available and over 200 health apps added to major app stores daily [18]. Concurrently, despite a growing evidence base regarding the role of mobile apps in physiotherapy contexts, there remains a dearth of high-quality research using validated and rigorous scientific methods assessing the quality of physiotherapy-specific mobile apps and their behavior change aspects (eg, BCTs used and behavior change potential) [11,25].

The primary study objective was to describe mobile apps, intended for use by patients, that can support physiotherapy care. The secondary objective was to evaluate app quality [26] and any BCTs contained within them [21], and to evaluate the behavior change potential of the apps [24].

The detailed protocol for this systematic review was previously published [25]. As this study is a systematic review of apps (in app stores) and does not involve human participants, ethics approval was not required. The review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement’s systematic review reporting principles, with minor adaptations as relevant for our search of app stores rather than research literature [27]. This method is common and adopted in other similar studies [18,28,29] (Multimedia Appendix 1) [30].

App characteristics are reported as descriptive and categorical data and a proportion (percentage), including app name, data privacy transparency (yes or no), companion app (yes or no), platform (Apple App Store, Google Play, or both), focus of the app (specific condition or region or whether more general), physiotherapy specialty (eg, musculoskeletal, pelvic health, and pediatric), target behaviors, simplified target behaviors, country of origin, developer qualifications (not clear or health care professional [HCP] or non-HCP), app version number, payment method (free or in-app purchases or one-off payment), and cost. Overall mean (SD) and individual domain scores are presented for MARS and ABACUS scores, as well as total BCT identification frequency and BCTs coded per app.

Our initial search identified a total of 1913 apps (Figure 1). This included 1302 unique apps from the Apple App Store, 600 apps from the Google Play Store, and a further 11 apps identified on the websites of professional physiotherapy associations. Overall, 834 were left after duplicates were removed. Following screening of app name, description, and screenshots, a further 790 were excluded. In total, 44 apps were screened fully, of which 19 were excluded. Reasons for exclusion included non-English language, costing more than Aus $10 (a currency exchange rate of Aus $1=US $0.72 is applicable), delivering its own service, irrelevant content, targeted at physiotherapist (clinician use) rather than the patient, being a bespoke or white-labeled app (ie, standard architectures licensed to a business or brand for private use), incompatibility with a current mobile device, and not a stand-alone app. This left 25 included apps for data extraction. Search update was rerun in February 2023, yielding an additional 405 unique apps. Following screening, this yielded a further 10 included apps for data extraction. Thus, a total of 35 apps were examined.

App characteristics are described in Multimedia Appendix 2. Overall, 22 (63%) apps were available on both Apple App Store and Google Play platforms, and 12 (34%) only on Apple App Store. Only 1 (3%) app was solely available on Google Play. Country of origin was clear for 24 (69%) apps, with 6 apps originating from the United Kingdom, 6 from the United States, and 5 from Australia. Regarding price, 26 (74%) were free to download, and 9 had a one-off payment price ranging from Aus $1.49 to $9.99. Additionally, 10 apps offered in-app purchases (including subscriptions). Just under half of the apps (n=16, 46%) were not clear about the developer’s qualifications, a further 17 (49%) were developed by HCPs (ie, physiotherapists, orthopedic surgeons, exercise physiologists, and chiropractors), and 2 (6%) were developed by non-HCPs. Overall, 15 (43%) apps had an obvious physiotherapist companion feature (ie, dashboard, desktop software, or precise login for therapists to access and engage with patients), and 31 (89%) apps had information about privacy and security in the app store. The focus of most apps was general in nature (eg, not condition-specific; n=24, 69%), while 11 (31%) had a specific target (eg, knee rehabilitation and pelvic floor training). This included 3 apps focusing on pelvic health and 1 app each focusing on wrist or hand, knee, hemophilia, core stability, wheelchair-bound individuals, cerebral palsy, osteoarthritis, and back pain.

Table 1 shows app quality scores using the MARS. Individual app quality scores ranged from 2.8 to 4.5, with a mean score of 3.7 (SD 0.4) of a maximum of 5. The apps with the highest MARS scores were Squeezy: CF (4.5), Squeezy (4.4), AllyCare (4.3), Squeezy for Men (4.2), and TeleHab (4.2). Of the appraised MARS domains, “aesthetics” (mean 3.9, SD 0.6) and “functionality” (mean 3.8, SD 0.5) were the highest scoring sections, while “subjective app quality” (mean 3.4, SD 0.8) and “engagement” (mean 3.4, SD 0.5) were the lowest. To the best of our knowledge, only 4 (11%) apps met criterion 19 (has the app been trialed or tested and published in scientific literature), including ReHand, Hand Rehabilitation, Squeezy, Embodia, and PhysiApp. Only 1 (2.9%) app (ReHand, Hand Rehabilitation) has been evaluated in a randomized controlled trial (RCT), showing positive outcomes for physiotherapy patients [34,35].

The most frequently observed target behaviors in the apps included recording information about exercise (n=28, 80%), performing therapeutic exercise (general; n=25, 71%), and communicating with a health professional (n=12, 34%; Multimedia Appendix 2). Other observed target behaviors included performing therapeutic exercise (men’s health, women’s health, hand therapy, knee, and perioperative), connecting with a health professional (make an appointment), connecting with friends, general self-care, and postsurgical self-care.

Coded BCTs by frequency can be observed in Multimedia Appendix 4, with a glossary of BCTs presented in Multimedia Appendix 5. The mean number of BCTs identified per app was 8.5 (SD 3.6). The apps with the highest number of unique BCTs were AllyCare (16), My Exercise Messages (14), ReHand, Hand Rehabilitation (14), and PhysiApp, BlueJay Engage, A Rehab Diary, and TeleHab (all with 13 BCTs identified). The most frequently coded BCTs were BCT 4.1 (instruction on how to perform a behavior; coded in n=32, 91% apps), BCT 1.4 (action planning; coded in n=30, 86% apps), BCT 2.3 (self-monitoring of behavior; coded in n=28, 80% apps), BCT 2.2 (feedback on behavior; coded in n=27, 77% apps), BCT 6.1 (demonstration of the behavior; coded in n=27, 77% apps), BCT 7.1 (prompts or cues; coded in n=25, 71% apps), and BCT 9.1 (credible source; coded in n=25, 71% apps).

Table 2 shows ABACUS scores for each app. The behavior change potential of included apps was a mean of 8.5 (SD 3.1) of a maximum of 21 (range 3-15). Of the 4 domains assessed by the ABACUS, section 1 (knowledge and information: mean score 3.0, SD 0.9) and section 3 (feedback and monitoring: mean score 2.9, SD 1.6) scored the highest, while section 2 (goals and planning: mean score 0.3, SD 0.8) and section 4 (actions: mean score 2.4, SD 1.0) scored the lowest. The apps (n=21) with the highest ABACUS scores included AllyCare (n=15, 71%), My Exercise Messages (n=14, 67%), ExorLive Go (n=14, 67%), A Rehab Diary (n=13, 62%), and both BlueJay Engage (n=12, 57%) and ReHand, Hand Rehabilitation (n=12, 57%).

The study has limitations. The review only analyzed apps in the English language, which limits the generalizability of the findings. Importantly, the search strategy had specific and stringent inclusion or exclusion criteria, which may lead to interpretative bias. For instance, apps had to be directly identifiable as being supportive of care between a patient and their physiotherapist. While several apps in app stores may indirectly support physiotherapy (eg, general exercise or informational apps), these were out of scope. Similarly, our search was based on a specified set of keywords, and thus, retrieval may not have captured all relevant apps. Furthermore, some apps originally identified in the search strategy were not able to be trialed for several reasons such as no free trial readily available, not available in the select region, or no response by developers for limited-time access by the study authors. A similar scenario occurred for apps that were subscription-based. In these cases, free trials were examined where possible, which may have not included full app functionality, thus potentially leading to more conservative quality, BCT, and behavioral potential scores. Regarding app quality, this was not appraised by patients themselves in this study. Should this have occurred, a different version of the MARS, the user-MARS, would have been required [61]. It is possible that patients may achieve a different quality rating for the apps. This is an area worthy of future research.

In addition, our protocol involved apps being trialed for a minimum of 10 minutes [25,29]. Vaghefi and Tulu [62] suggest “... most users tend to withdraw from mHealth apps before the end of the first week.” This raises the possibility that engaging with each app for a limited amount of time can make it difficult to get a complete picture of an app, with certain features and BCTs not being immediately apparent. This was observed firsthand in the case of “My Exercise Messages,” which scored 14/21 on the ABACUS in this study but scored 17/21 when rated by its developers who are intimately familiar with all the features of the entire 24-week app program [50]. The discrepancy is likely due to the fact that the more BCTs become apparent, the longer a user engages with the app over 6 months. A similar issue lies in the possibility that important descriptive data about an app may be missed or inaccurate because we relied on limited data sources. For instance, not all the necessary detailed information about an app can be found in the app itself or within its app store description. In select cases, this information may be identifiable if publications about the app exist. However, our protocol was deliberately designed to more closely mimic how a lay user would likely find and interact with an app through an app store. For these reasons, the findings reported in this study may be conservative for some apps, and it is possible that quality and behavior change scores would be higher with proper in-depth use. We thus urge readers to use caution in interpreting our findings.

Finally, this study focused primarily on app descriptive characteristics, quality, BCTs, and behavior change potential. It was not an aim of this review to evaluate the clinical effectiveness of these apps in changing behavior or their efficacy toward improving patient health outcomes.

Mobile apps available to support patient care received from a physiotherapist are of variable quality and contain relatively few BCTs. The potential for behavior change varied widely across apps. This study has provided the first comprehensive examination of mobile apps specifically supporting the care of patients receiving physiotherapy.