mHealth App Prescription in Australian General Practice: Pre-Post Study

Introduction

More than 350,000 apps exist in the Medical and Health and Fitness categories in major app stores [1], with downloads and revenues in the billions [2]. Their popularity and potential to influence health-related behaviors make their integration to medical practice imminent [3]. Pragmatic studies of app prescription in primary care have been emerging around the world with varied interventions and results [4-6]. To assist the integration of apps into clinical practice, mobile health (mHealth) app repositories have been created, including the National Health Service App library in the United Kingdom [7], Health Navigator in New Zealand [8], and other private entities such as AppScript [9] and the Organization for the Review of Care and Health Applications [10].

Given the potential of mHealth apps to help improve the self-management of chronic conditions, we explored their value in general practice. Previously, in an overview of systematic reviews, we explored the possibility of simple integration of mHealth apps into the general practice setting and proposed a concept of “prescribable” mHealth apps. These were defined as proven effective (that is, shown to help achieve measurable clinical improvements in patients’ conditions), in addition to being standalone and currently available in the app stores [11].

We also explored the potential barriers to app integration in Australian general practice [12]. Patients expressed their preference for doctor-recommended apps; however, doctors were overwhelmed by the sheer number of available apps and faced 2 major barriers: not knowing of many prescribable apps and the lack of trustworthy source to access such apps. To address these barriers, we developed a brief implementation intervention. Objectives of this study were to explore the feasibility of app prescription by general practitioners (GPs) and to evaluate the effectiveness of an implementation intervention to increase uptake of app prescription.

Methods

Study Design and Setting

We employed a single-group, before-and-after design. Our study was conducted in the Australian general practice setting. Ethics approval was obtained from the Bond University Human Research Ethics Committee (#OB00017).

Participants and Recruitment

GPs currently working in Australia at least 2 days a week were eligible to participate in our study. Information about the study was distributed at 2 annual national GP conferences (GPDU2018 and GP18) and posted to a closed Facebook group called GPs DownUnder. Recruitment occurred from June 2018 through November 2018, and data collection occurred from September 2018 until May 2019. Upon completion of the study, GPs were thanked with Aus $50 gift cards.

Intervention

There were two parts to the intervention. First, prescription pads for 6 apps were developed (Figure 1). These apps were chosen because they address conditions relevant in general practice, have either direct trial evidence (This Way Up: Managing Depression, St. Vincent’s Hospital Sydney Ltd [13]; Tät – Pelvic floor exercises, Umeå University, Sweden [14]; Lose-It!, FitNow Inc [15,16]; CBT-i Coach, US Department of Veteran’s Affairs [17]) or indirect evidence from trials of similar apps (Smiling Mind, Smiling Mind Pty Ltd [18-20] and Quit Now: My QuitBuddy, Australian National Preventative Health Agency [21]). The apps also had to have stable content, were created or backed by trustworthy not-for-profit organizations, and were available for both Apple and Android phones. Five of the apps were freely available, and one (This Way Up: Managing Depression) had a one-time purchase price of Aus $59.99. The cost of apps was not an exclusion criterion as it will help assess if cost is a barrier to app prescription.

The app prescription pads had individually numbered pages with a tear-off design. Each app prescription page included the app’s full name and logo, download instructions, space for the patient’s name, the reason for prescription, and a disclaimer. Prescription pads were assembled onto an A4 display stand and mailed to participating GPs. A letter outlining the study timelines and procedures along with a short introduction to each app was included in the shipment.

The second part of the intervention was aimed at enhancing uptake. Short videos (2 minutes) demonstrating the content, functions, and features of the apps in detail were created for each app. A YouTube link to the video randomly selected for each participant was emailed following their second month’s reporting.

Our study aimed to change the prescribing behavior of GPs. Evidence suggests that behavioral interventions are more effective and sustainable when guided by behavior change techniques. Our prior research helped to identify the target behaviors [12]. We based our intervention on the Capability, Opportunity, Motivation, and Behavior model [22]. Capability to prescribe apps was addressed through the list of evidence-based apps and the introductory videos demonstrating the content, features, and function of the apps; opportunity was enabled through the purposefully designed stand with the prescription pads; and motivation was harnessed through the GPs’ expressed interest in the study that demonstrates their belief that app prescription would be a good thing to do [23].

‎View this figure

Procedures

At the beginning of the study, participants signed consent forms and answered the prestudy questionnaire via the web-based SurveyMonkey tool (SurveyMonkey Inc, San Mateo, CA). The survey collected demographic data, contact details, current app prescription rate in the preceding 2 weeks (self-reported in ranges: 0, 1-5, 6-10, >10 times), and level of confidence around app prescription.

The official commencement dates for the study were recorded as the date that each participant reported they started using the pads. Every 4 weeks following commencement, participants were asked to send a photo of the prescription pads electronically to the research team to provide details of the number of prescriptions for each app within that month. If participants took leave from work, the reporting dates were adjusted to allow for a full 4-week reporting period.

Qualitative semistructured interviews (10-15 minutes) were conducted and audio recorded at the end of the study, either face-to-face or by telephone, to gather feedback on the intervention. GPs were asked about their knowledge of other apps and relevant resources outside the study, including the Handbook of Non-Drug Interventions (HANDI) project by the Royal Australian College of General Practitioners, which includes a number of mHealth apps. Interviews were transcribed verbatim, coded by the lead researcher (OB), and thematically analyzed to determine the feasibility of the interventions, barriers, and solutions to the scalability of the intervention to Australian GPs. The thematic analysis was done in consultation with a second author (TH).

Sample Size

Prior data [12] indicated that the difference in the response before and after is normally distributed with a standard deviation of 10 and a baseline mean of 2 apps prescribed per month per GP. We calculated that we needed 24 participants to have 80% power (with α=.05) to detect an increase of app prescription by at least 6 per month. Taking attrition into account, we planned to recruit 30 GPs for the study.

Outcomes

Data on app usage were collected for the 2-week period prior to study commencement and then every month for 4 months. The primary outcome of the study was the number of app prescriptions dispensed in total, as an average per month, per GP, and per GP per fortnight. We calculated the median number of apps recommended by a GP per fortnight using the following formula:

m=l + (w(n/2-c))/f

where l is the lower limit of the bin (range) containing the median, w is the width of the bin, n is the total population, c is the cumulative count (frequency) up to l, and f is the count in the median bin.

Prestudy raw numbers are provided in Table 1 (m=1.7 [1 + (5(39/2-17))/19]). Poststudy numbers are given in the Results section (m=4.1 [1 + (5(39/2-0))/31]).

Secondary outcomes were confidence around prescribing apps (measured on a 5-point Likert scale; prestudy and poststudy); the number of intervention video views and their impact on the subsequent prescription numbers; and attrition rate. In addition, the acceptability of the interventions to GPs and their feedback on the interventions were explored in semistructured interviews. Descriptive statistics were used to report the frequency of app use at each time point and confidence in app prescription. Qualitative data were analyzed thematically.

To conduct the overall analysis of the effect of video exposure on prescription rates, the 6 separate outcomes (1 for each app) were considered as one overall global outcome (individual monthly counts were not aggregated). Initially, a Poisson model was fitted with the (categorical) explanatory variables specified: the month (1 to 4), exposure to the video (yes/no), video (1 to 6), and interaction between exposure and video. To account for the 24 repeated measures collected for each GP (4 timepoints by 6 apps), a random intercept was fitted. Overdispersion was assessed using generalized chi-square divided by degrees of freedom. Due to evidence of overdispersion for the Poisson model (generalized chi-square/degrees of freedom=2.13), a negative binomial model was fitted and showed no evidence of over-dispersion (generalized chi-square/degrees of freedom=0.98).

Results

Overview

A total of 40 currently practicing Australian GPs were recruited for this study. One GP dropped out before the beginning of the study, and 3 GPs dropped out after the second and third data collection due to relocation and change of jobs. The full 4-month study was completed by 36 GPs, and we analyzed the data as intention-to-treat (ITT). The median age of the participants was 40 years, the median length of time in practice was 8.5 years, and participants worked a median of 4 days a week (Table 1).

Prescriptions

In total, 1324 app prescriptions were dispensed over 4 months, with a mean of 331 prescriptions a month. Figure 2 illustrates the number of individual app prescriptions within the monthly totals. The number of apps prescribed per GP per fortnight increased from an imputed prestudy median of 1.7 to 4.1. Overall, the Smiling mind app was the most frequently prescribed (533/1324, 40%), followed by CBT-i Coach (242/1324, 18%), Managing Depression (167/1324, 13%), Lose-It! (155/1324, 12%), Quit Now: My QuitBuddy (134/1324, 10%), and Tät Pelvic floor exercises (93/1324, 7%).

‎View this figure

Figure 3 illustrates the distribution of the total app prescription per GP. According to the ITT analysis, a median of 30 apps (range 6-111 apps) was prescribed per GP over the 4 months. Every GP prescribed at least one app per fortnight, 31 (31/39, 80%) GPs prescribed 1-5 apps, 7 (7/39, 18%) prescribed 6-10 apps, and 1 GP prescribed more than 11 apps. The GPs’ confidence around prescribing apps doubled from a mean of 2 (not so confident) before the study to 4 (very confident) at the end of the study: 0/39 not confident at all; 1/39 (3%) not so confident; 12/39 (31%) somewhat confident; 25/39 (64%) very confident; 1/39 (3%) extremely confident.

At the end of the study, the My QuitBuddy app video was viewed 8 times; the Smiling mind, Managing Depression, and Lose-It! app introduction videos were viewed 9 times each; the Tat-Pelvic floor exercise video was viewed 19 times; and the CBT-i Coach video was viewed 21 times. We were not able to track whether every GP watched the video sent to them. The effects of exposure to app videos are shown in Figure 4. Only two of the app videos had a significant effect on the subsequent app prescription numbers following the exposure to the video: Smiling Mind app prescription increased from 3-4 times per month to 6 times per month, and Lose-It! app prescription increased by one time. The full analysis is provided in Multimedia Appendix 1. A global test for the interaction between exposure and video showed strong evidence of heterogeneity (P<.001) indicating the treatment effects were different across the 6 apps. Therefore, we did not report an overall effect of the videos.

‎View this figure

‎View this figure

Qualitative Interviews

As per the ITT analysis, 39 GPs were interviewed at the conclusion of their participation in the study. Participants expressed their overall experience of prescribing apps as overwhelmingly positive. They liked the size of the prescription pad, the information included on it, ease of use, and integration into the workflow, with the most useful feature identified as the visual cue aspect of the stand. They also liked the short length of the videos, yet felt they contained sufficient details about the apps. Most GPs reported not downloading and interacting with the apps themselves. Although most reported having watched the allocated video, many did not recall the contents during the poststudy interviews.

Two of the 6 study apps were well known to the GPs: 28/39 GPs were already familiar with Smiling Mind, and 12/39 GPs were already familiar with Managing Depression. They had been recommending these apps to their patients even before the study and appreciated having a formal prescription to give out during the study. Among the other apps that GPs recommended, mindfulness and meditation apps (Calm, Headspace) were common. Mental health–related apps were the most frequently prescribed, and all GPs reported that the overall number of apps they prescribed is a reflection of the demographics of their patients and the prevalence of conditions encountered.

GPs reported that they might have prescribed the weight loss and pelvic floor exercise apps more frequently. Instead, they habitually referred patients to dietitians and physiotherapists or to programs and tools already compiled as the first line of intervention. None of the GPs, except for one, had watched, read, or received any other app-related content apart from the study intervention. Knowledge of HANDI was low, especially that apps were included in some HANDI entries. However, upon learning this, they all agreed that HANDI would be a reliable evidence-based app repository for GPs in Australia. The main barriers and facilitators to app prescription in general practice are shown in Table 2 along with illustrative quotes.

Discussion

This study demonstrated that it may be possible to increase the uptake of mHealth app prescription by providing an implementation intervention in the Australian general practice setting. The results demonstrated a total of 1324 app prescriptions by 39 GPs over 4 months and positive feedback from GPs about the intervention. The fortnightly number of apps prescribed per GP more than doubled compared to the prestudy level. However, identified barriers to app prescription uptake were poor knowledge of prescribable apps and insufficient familiarity with the apps to foster confident prescribing habits. Participants identified a need for a reliable prescribable app repository, preferably integrated with their electronic medical systems, and consistent and persistent messaging to increase the knowledge and familiarity of such apps.

The variation in the total individual tally of apps prescribed by participants may reflect differences in their personal digital propensity and flexibility in altering prescribing behavior. The reduction in the monthly app prescriptions after the first month could be related to the timing of the second and third reporting for about half of the participants. These occurred during the Christmas, New Year, and summer holidays in Australia, during which acute conditions dominate GP visits more than chronic conditions, which were the focus of the apps in the intervention.

The app explanation videos had varying effects on app prescription numbers. The results from the qualitative interviews showed that app prescription numbers are primarily dependent on the patient cohort and the prevalence of the conditions for which the intervention apps were intended. Thus, the short explanatory videos were informative but unlikely to be sufficient to influence complex behaviors such as prescribing. Perhaps, it would be more beneficial if video introduction and instructions for mHealth apps were developed for patients and given as part of the app prescription.

This is the first study to test the feasibility of an intervention to increase app prescription in Australian general practice. The overall attrition rate was low, and we analyzed the data as ITT, including those who dropped out of the study. Limitations include lack of access to electronic medical record data of the GP clinics to correlate the prevalence of conditions with the frequency of app prescription within the patient cohort. We aimed to recruit a sample of GPs representative of the national GP cohort; however, our participants’ median age of 40 years was younger than the national average of 50-55 years. Other limitations include a single-group pre-post study design, possible volunteer bias of the participants, and short time frame (4 months). Ideally, a randomized controlled trial should be conducted to establish the long-term effectiveness of the intervention with a large and representative sample for a longer duration. Due to the restrictions of available time and resources, we were unable to achieve this. Future studies should also opt for an electronic version of app prescription to improve sustainability and scalability. Another limitation is the analysis of qualitative data by a single researcher; however, the qualitative data result was a small part of our secondary outcome to primarily answer if the intervention was acceptable and feasible for practicing GPs.

There are few comparable studies of app recommendation in a primary care setting. A trial for an app prescription platform, AppSalut, in Spain involved 32 doctors who made 79 app recommendations in 5 months. Of the three apps they used, a medication adherence app was the most prescribed [4]. It sends the prescribed app to patients as text messages and can monitor and receive data on patients’ use and adherence to the system. In the United States, the Cambridge Health Alliance network of primary care clinics implemented a mental health app dissemination program, in which they evaluated mental health apps, selected 7 apps, and recommended these 7 apps in 12 primary care clinics [5]. Similar to the finding of our study, app prescriptions for anxiety and stress were the most frequently prescribed. An Australian study tested the feasibility of integrating mHealth apps into dietetic practice by asking 5 dietitians to use one chosen app for 12 weeks [6].

All of these studies provided training to the participating health care professionals to educate them about the study apps as well as the electronic systems they needed to use. The qualitative feedback from our participants also included the need for such training. However, because GPs often report being overworked, time-poor, and inundated with different information and offers, it would be challenging to organize out-of-hours training involving many GPs or train dedicated personnel to visit GP clinics during lunch hours, which was suggested by the GPs as a solution. The scalability of such an intervention would pose funding and logistical challenges.

One way to promote the sustainability and scalability of mHealth app integration into clinical practice is to provide an electronic repository of vetted and curated apps for health care professionals. In Australia, the Victoria Department of Health [24], Black Dog Institute [25], and HANDI project by the Royal Australian College of General Practitioners [26] offer small repositories of mHealth apps, but these organizations function under different jurisdictions with no national guideline in place. GPs in our study emphasized the need for a nationally accessible repository of a select few prescribable apps that are relevant to general practice that is safe, reliable, and easy to navigate.

We found that mHealth app prescription is feasible in a general practice setting in Australia by addressing previously identified practical barriers to mHealth app prescription. Our implementation intervention was effective in increasing app prescription. However, the future of app prescription depends on efforts to increase GPs’ knowledge of prescribable apps as well as a dedicated trustworthy app repository for GPs.