This single-center prospective study aimed to evaluate the acceptability of the DIDE application (ClinicalTrials.gov Identifier: NCT04659954). Eligible patients were proposed to download and use the application during eight weeks, in addition to their usual treatment in the center (medical appointments, interviews with nurses, pharmacological treatment). The study was conducted in Toulouse University Hospital, France, between January 2022 and December 2022.

Primary endpoint was first defined as the percentage of patients who logged into the application at least once every week. Some participants used the application on a more irregular basis over the study’s timeframe, as they felt the need to: this possibility was in line with existing data regarding mobile health use [26]. We thus modified our primary endpoint, to also include patients who logged into the application at least 10 times in four different weeks during the eight weeks of the study. User logins were recorded in real-time, and later separated in four groups for analysis: user logins during center’s working hours (Monday to Friday, 9 a.m. – 5 p.m.), evenings (every day, 5 p.m. – 10 p.m.) nights (every day, 10 p.m. – 9 a.m.) and week-end (Saturday, Sunday and bank holidays, 9 a.m. – 5 p.m.)

At the end of the study, participants were asked to complete the French-translated System Usability Scale (SUS) questionnaire, widely used in applications development and user experience design, in order to evaluate the usability of the application [27]. This questionnaire comprises 10 items, such as “I think that I would like to use this system frequently”, and aims to measure subjective impressions of a system’s usability, frequently defined as the combination of a system’s effectiveness, efficiency and the user’s satisfaction [28]. Each item is rated using a five-points Likert Scale, and the final score is comprised between 1 and 100.

Participants were outpatients at our SUD treatment center, and were proposed to participate in the study during medical appointments. Eligibility criteria were as follows: presenting an active substance use addiction requiring weekly monitoring of consumption, possessing an Android smartphone or having access to an Android smartphone and having access to an internet connection, being able to read and understand French. Patients with cognitive or psychiatric disorders that may affect their ability to consent, or involved in another protocol involving a modification of the treatment for addiction were excluded from the study. Addiction diagnosis was determined after medical examination and based on DSM-5 criteria [29]. After they accepted to take part in the study, participants could download the application from Google’s Play Store, and had the application activated on their own smartphones. The application was operational during eight weeks from the installation, before its access was remotely revoked. A delay of two additional weeks was possible depending on the end of study visit date.

The application was developed in collaboration with BotDesign, a partner start-up specialized in the development of eHealth solutions (https://www.botdesign.net). The application’s design is based on attachment theory and neurobiology of addiction, in order to help fostering the therapeutic alliance by proposing a system suited to answer the specific needs of patients with SUD. The application is centered around three core-concepts: ecological momentary assessment, messaging system with the healthcare team, and positive social reinforcement.

First, the choice of using ecological momentary assessment, in the form of self-report questionnaires later discussed with a caregiving team, was made for patients to more easily report their psychological state to the healthcare team, despite the possible fluctuation of symptoms [30], and the alteration of circadian rhythms [31]. We moreover considered self-report systems as a way to promote the declarative memory system, mediated by the hippocampus and hypothetically impaired by drugs of abuse to the benefit of habit memory, further reinforcing addiction [32]. Patients were prompted to complete questionnaires either through a conversation with a chatbot using multiple choices of answer, or by selecting a specific questionnaire. The first questionnaire, “Consumption diary”, acts as a self-report tool of the patient’s use of substance or addictive behavior, where patients can increment or decrement a counter through the day for each type of addiction specified within the application (e.g., cannabis, cocaine). The second questionnaire, entitled “Rhythm of life”, revolves around self-report of circadian rhythms of sleep and meals, which are altered in SUD [31]. A “Feeling” questionnaire was implemented, with the aim for patients to specify their overall well-being using a visual scale ranging from 0 to 10, in order to facilitate mentalization of their internal state. Patients also have the possibility, at any time, to report their craving using a dedicated questionnaire, with the option to specify the craving’s intensity, and to send a message to the caregiving team or receive information on available hotlines or forums for addiction. The questionnaires’ content is available in Additional File 1.

Second, the increased accessibility to caregivers was guaranteed by a messaging system, which the patient was systematically encouraged to use. Patients were informed that these messages were not to be considered as an emergency contact. However, they allowed them to leave a trace of their difficulties, so that the team could secondarily get in touch with them to propose appropriate support. Caregivers actively interacted with the patient during our service’s working hours, and showed their own interest in the intervention, which in turn was hypothesized to stimulate patients’ engagement in care. Data gathered within the app moreover served as decision-making support system for the team by providing supplementary clinical information.

Finally, the application provided positive social reinforcement, by using a considerate, valuing language and by displaying various messages to show concern to the user, congratulate or greet them. Combined with the possibility to review the application use with the caregiving team, these elements aimed to renew the patient’s interest in social interactions, hindered by SUD induced hypohedonic state and impairment of social skills [33].

Healthcare professionals in the center were granted a distinct access to the application system, using a dedicated desktop interface, that allowed them to use the messaging system with the patients, and to monitor data from individual questionnaires. The system was presented as a tool to help fostering therapeutic alliance with SUD patients, and was integrated in the center’s daily functioning. The team took additional time during their weekly clinical meetings to review completion of questionnaires, which helped decision-making processes regarding patients’ treatment, and helped them to prioritize patients they felt were facing more difficulties.

In order to maximize patients’ engagement, we used codesign techniques during the application’s development. Prior to the start of this study, patients with SUD were received in the center where the application would later be deployed. Voluntary patients were asked general questions about smartphone possession and usage, and their interest in health applications, before being presented with a prototype of the application, and asked to accomplish specific tasks. Additional feedback on the prototype was gathered during two supplementary sessions, during which one researcher (AS), as well as user experience designers from our partner start-up, asked the users their opinion and ideas on the application, their feedback on the questionnaires and the overall wording of the application’s content, and observed their interaction with the interface. This feedback was implemented in subsequent versions of the application. The application was not modified by users’ feedback during the course of study.

This research received a favorable evaluation from the French Committee for the Protection of Persons Sud Méditerranée I (Marseille, France), and was registered under the internal reference number 20 93.

A distinction was made between categorical (e.g., professional status, housing) and quantitative variables. Percentages were calculated for each categorical variable possible values, while for quantitative variables, a set of descriptive statistical measures, including the mean with the standard deviation, was studied. Additionally, boxplot graphs were used to observe the distribution of quantitative variables.

Confidence intervals with a 95% confidence level were constructed to provide a plausible range of values for the percentages of categorical variables and also for the means of quantitative variables. Due to the small sample size and non-normality of the data, assessed using the Shapiro-Wilk test, specific tests were used to calculate these intervals. The Bootstrap percentile method was employed for quantitative variables, while the Clopper-Pearson Mid P method was used when two modalities were present for categorical variables, and the multinomial method when three or more modalities were present. All analyses were conducted using R Studio - version 4.2.1.