Acceptability has been defined as “the demonstrable willingness within a user group to employ information technology for the tasks it is designed to support” [49]. Focus groups are a qualitative method to collect in-depth information about participants’ opinions, attitudes, and beliefs around a specific topic [50] and are thus suited for acceptability studies. In this phase, we will conduct 6 focus groups onsite at WSFT to measure patient and stakeholder acceptability and preferences of using EHRs to identify at-risk patients and send them electronic messages that promote behavior change mobile apps.

Participants, both patients and hospital staff, must be at least 18 years of age. Patients must self-identify as smokers and/or individuals who regularly consume alcohol. We will recruit participants who drink alcohol at any level (not just risky levels) to avoid alienating patients who do not recognize they are drinking at risky levels or those who do not wish to identify as such. We will attempt to balance participants by gender and age.

Topic guides will be based on Perceived Attributes of eHealth Innovations [51] and the diffusion of innovation theory [52], which can explain and predict an intervention’s acceptability and potential for adoption [51,53]. Specifically, we will focus the topic guide on 3 constructs central to the Perceived Attributes of eHealth Innovations and diffusion of innovation theory—relative advantage, complexity, and compatibility. These 3 attributes were selected for their alignment with the larger construct of acceptability (our Phase 1 focal point), and we will focus equally on these 3 attributes in terms of question intensity and analysis. To the extent that participant discussions are responsive to these attributes, we do not intend to prioritize one attribute over another.

We will conduct 3 focus groups with patients: one with patients who smoke tobacco, one with patients who drink alcohol, and a third with patients who smoke tobacco and drink alcohol. In addition to questions about the acceptability of the electronic message system, we will ask patients about their preferences for message modality (email, SMS, online patient portal, other), opinions about the content of messages (see Multimedia Appendix 1 for sample messages), and any potential unanticipated consequences.

We will also conduct 3 focus groups with hospital staff: one with health care professionals who undertake lifestyle screening, one with EHR administrative staff (eg, information analysts, information technologists, communications staff), and a third with senior managers in the eCare team. In addition to questions about the acceptability of the messages, we will ask hospital staff to identify any technical, legal, privacy, ethical, or medical challenges or concerns they foresee in implementing the proposed system within the hospital’s EHR.

We will recruit patients via the WSFT website and through a volunteer organization that supports research studies at the hospital. Hospital staff will be recruited via intranet, email newsletter, word of mouth, and a monthly corporate briefing. Patients and staff members who express interest in the focus groups will be emailed an information sheet and a consent form. Participants must sign the consent forms prior to any research activity. Participants in the patient focus groups will complete a brief, anonymous, demographic form similar to the demographic characteristics data recorded in the EHR (see Phase 2). Patients will be offered a £25 (US $34) voucher as a token of appreciation.

Each focus group will include approximately 6-8 participants and last approximately an hour. All focus groups will be audio-recorded. A professional transcription service will transcribe the audio recording verbatim, and all personal identifying information will be removed.

Qualitative coding software NVivo will be used to organize the themes in the transcripts [54]. Focus groups will be coded by one researcher, and coding will be cross-checked by a second independent researcher. Researchers will meet regularly to discuss the coding and any discrepancies will be resolved collaboratively. Framework analysis will be used to analyze the data, guided by the Perceived Attributes of eHealth Innovations [51]. We will also use inductive analysis where themes will be identified from the data [55]. Data from the focus groups will inform messaging decisions in Phase 3. For example, results from the focus groups will determine whether messages are sent via SMS, email, or patient health portal and the framing of the message content in order to maximize acceptability and minimize any concerns.

Feasibility assesses whether the methods and procedures of a proposed study will work before implementing it on a large scale [56]. We will work with hospital information analysts to identify the proportions of patients who smoke, drink alcohol at risky levels, or both in the hospital’s EHR. Obtaining information about the patient population will inform the planned segmentation in Phase 3. Specifically, we will mine the EHR for the following key data fields: risk profile, contact information, demographic characteristics, chronic or past health conditions, and screening recency. The risk profile is categorized as (1) exclusive smokers, (2) exclusive risky alcohol drinkers, and (3) dual smokers and drinkers. Contact information includes an email address, a mobile phone number, or a record of access to the online patient portal. Demographic characteristics include gender, age, ethnicity, and level of economic deprivation (Index of Multiple Deprivation using post code). Chronic or past health conditions include hypertension, high cholesterol, heart disease, chronic kidney disease, stroke, chronic obstructive pulmonary disorder, asthma, diabetes, dementia, cancer, arthritis, schizophrenia, bipolar and/or other psychosis, and depression and/or anxiety. Screening recency is the date on which tobacco and alcohol use status was most recently assessed.

We will use descriptive statistics to tabulate the frequencies and percentages of patients with valid data in each of the aforementioned key fields. Availability of contact information, demographics, health conditions, and screening recency will be stratified by risk profile. Distributions will be compared with chi-squared tests and logistic regression analyses. No personally identifiable information nor patient-level data will be available to the study team. The study team will provide table shells to the hospital-based information analysts, who will aggregate the data onsite and return deidentified results to the study team, consistent with hospital policies. Findings from Phase 2 will inform the viability of the proposed 3 risk profile groups to be examined in Phase 3. A sample size of 383 will be needed for each risk profile to produce a 95% confidence interval of ±5%, assuming a population proportion of 50% and a population size of 100,000 [57]. While we plan to conduct Phase 3 regardless of the true EHR population size as revealed in Phase 2, this statistic provides a benchmark for interpreting results. Furthermore, although not an outcome for the proposed study, findings from Phase 2 will be shared with hospital staff to identify any potential gaps in completeness of records and to suggest potential changes to hospital protocols to improve their completeness.

The reach of a public health intervention can be defined as “The proportion and representativeness of individuals who are willing to participatein a given initiative, intervention, or program.” [58]. In this phase, we will send proactive electronic messages to at-risk patients who have visited the hospital in the past year, encouraging them to access behavior change apps that support smoking cessation and alcohol reduction. The primary outcome will be the click rate on the link embedded in these messages. We will tailor messages to each participant’s risk profile (ie, smoker, risky drinker).

All identified smokers and risky drinkers for whom contact information is available will be eligible to receive messages, except for patients <18 years old; patients on the End of Life Pathway; pregnant women, as their data are held in a separate database and governed by separate hospital policies; and patients who have opted out of receiving messages from the hospital.

The final content and modality of the messages will be based on the acceptability and feasibility findings from Phase 1, and the final number of participants will be based on the feasibility findings from Phase 2. Eligible patients will be sent messages containing a link to a webpage hosted by the NHS where participants can download a free behavior change app relevant to their risk profile. These apps include “SmokeFree” and “Drink Free Days,” both of which are promoted by Public Health England as part of their “OneYou” campaign [45,46]. The apps are available for iOS and Android devices.

A data privacy impact assessment has been performed to assess the risk posed by processing patients’ data in order to send communications promoting behavior change apps [59]. Each participant will receive at least one initial message and another reminder message if they do not click the link in the initial message; the total number of messages will be determined by the results of Phase 1. To capture and track click rates, each message-embedded link will contain a tracking code that is unique to its recipient. Information technologists at the hospital will use the tracking codes to identify which participants clicked through to access the NHS webpage.

Approximately 1 week after the intervention messages have been sent, participants will receive a message inviting them to complete an anonymous follow-up survey. The survey will collect information on their experience of receiving the message, to further inform acceptability. As in Phase 1, survey items in Phase 3 will be based on the Perceived Attributes of eHealth Innovations and will examine 3 acceptability constructs: relative advantage, simplicity, and compatibility [51]. Patients who complete the survey will be entered into a prize draw to win £250 (US $338) high street vouchers.

With the exception of the primary outcome (ie, click rate of the link), all study measures are defined within the WSFT EHR system. For behavioral risk status, smokers are defined by a single item labelled “Does the patient smoke?”, and risky drinkers are defined by the hospital as those with an Alcohol Use Disorders Identification Test–Consumption version (AUDIT-C) score between 5 and 10 [60]. Demographics include gender, age, ethnicity, and Index of Multiple Deprivation. Health conditions include hypertension, high cholesterol, heart disease, chronic kidney disease, stroke, chronic obstructive pulmonary disorder, asthma, diabetes, dementia, cancer, arthritis, schizophrenia, bipolar and/or other psychosis, and depression and/or anxiety. The primary outcome for Phase 3 is the click-through rate of the links embedded in the messages. In other words, the outcome is the proportion of participants who click the message link to access a behavior change app, with the numerator being the number of individuals who clicked the link and the denominator being the number of individuals who received the message.

We will use logistic regression analysis to model the likelihood of clicking on the message-embedded link within each risk group, with demographics and health condition statuses entered as covariates. The logistic regression approach will support comparison of effect sizes as the relative risk, a metric that is familiar and easily interpretable across the health sciences. Additional analyses will explore the click rate stratified by risk group and sociodemographic characteristics to investigate health disparities, if any.

We determined a priori that 5% would be a clinically meaningful proportion of users clicking on the message-embedded link. This value was selected based on feasibility determined by prior research [39]. Although relatively modest in absolute terms, 5% of the total number of smokers or risky alcohol drinkers within the NHS EHRs would represent many people. We hypothesize that 5% is attainable based on similar studies in which patients identified through an EHR responded to a quitline referral [39,43] or proactive telephone calls offering telephonic coaching [41] and on email marketing campaign industry standards of click-through rates of 2%-5% [61]. This includes those whose mobile phone numbers are invalid (eg, message undeliverable) and therefore unable to receive the messages.