This investigation was part of a pragmatic observational study of the effects of mobile-based self-management for urinary incontinence. The app Tät® was designed for self-management of stress urinary incontinence in women. It includes information, lifestyle advice, PFMT exercises with graphical support, reminders, and a statistical function. The graphical support consisted of a moving bar, indicating the contraction time and intensity the user should aim for. The efficacy of Tät® was demonstrated in an RCT [6], after which the app was made freely available in Swedish from the App store or Google Play. Subsequently, authorized translators have translated the app into English, Finnish, German, Spanish, and Arabic, and the app has been released in those languages as well. The overall aim of the pragmatic study was to describe the use and effectiveness of the app when it was used under real-life circumstances.
Upon downloading the app, the user was informed about the study of the app when freely available, and was invited to participate by answering a short baseline questionnaire. Participants who were still using the app after 3 months were asked to complete a follow-up questionnaire. The study information and the questionnaires were also translated by authorized translators and appeared in the language that had been chosen when users downloaded the app. Both questionnaires were anonymously submitted to a secure research database. The questionnaires were assigned a unique identification number generated within the app, such that they could be correctly paired but not traced back to a specific mobile phone, telephone number, or any other personal data. Hence, all study data were self-reported, and we had no ability to confirm any of the submitted information.
Data were collected from participants who had downloaded the app in all available languages between 16 January 2018 and 1 June 2019. Another article describes this population, as well as the characteristics of completing participants and their effect on incontinence symptoms [18].
In this present study we only included those participants who had answered the PGI-I within 89–135 days (Fig. 1). Questionnaires submitted after more than 135 days were considered to reflect a more sporadic opening of the app rather than the continuous use that we aimed to study, or a reluctance to participate in the study and hence postponing the response. Follow-up time of shorter than 89 days was considered to indicate a technical error as this should not have been possible.
Further inclusion criteria were self-stated female sex, age of ≥18 years, and urinary incontinence defined according to the validated International Consultation on Incontinence Questionnaire-Urinary Incontinence Short Form (ICIQ-UI SF). Participants were considered to have urinary incontinence if they reported both leakage on the ICIQ-UI SF question “How often do you leak urine?” and if they reported any amount to the ICIQ-UI SF question “How much urine do you usually leak?” Exclusion criteria were current pregnancy or post-partum (<3 months) at baseline or follow-up.
Ethical approval and informed consent
The Regional Ethical Review Board, Umeå University, granted ethical approval to both collect the data (number 2014–389-32 M) and perform specific analyses in this study (number 2017–405-32 M). After reading the provided information about the study, it was optional to complete the questionnaire. Submission of the questionnaire was considered to indicate consent to participate, and these users were included in the study. No reimbursement was given.
Questionnaires at baseline and 3-month follow-up
Upon first opening the app, participants answered a baseline questionnaire about age, educational level, and current residence, as well as the validated questionnaire assessing urinary incontinence symptoms (ICIQ-UI SF) [19]. Age was registered as the stated age at baseline, and then divided into four categories: <30 years, 30–39 years, 40–49 years, and ≥ 50 years. To report their current residence, users answered a multiple-choice question with answers ranging from metropolitan to rural area, and could state their country of residence. The variable education was then recoded into a dichotomous variable: university education or not, because there were very few participants who had had 9 years or less of education. Questions were also asked about current pregnancy or recent delivery (<3 months) at baseline and follow-up. Three months after answering the baseline questionnaire, users were asked to complete the symptom score ICIQ-UI SF again, to estimate their frequency of training over the last 4 weeks and app usage since they downloaded the app, and to rate their improvement using the validated questionnaire Patient Global Impression of Improvement (PGI-I) [20]. The questions about PFMT frequency and app use had five different answer categories, as reported in Tables 1, 2, 3, which were not altered in any way for the analyses.
Table 1 Baseline and follow-up characteristics analyzed for association with any and great improvement (N = 2,153) Table 2 Univariate analysis of baseline and follow-up factors (N = 2,153) Table 3 Factors associated with any and great improvement (N = 2,153) The questions were considered as factors potentially associated with treatment outcome and were thus analyzed for possible association. Age was analyzed as a continuous and categorical variable in the univariate analyses. Education level, residence, and frequency of PFMT and app usage were analyzed as categorical variables. The total ICIQ-UI SF score was analyzed as a continuous variable.
Treatment outcome
We analyzed the selected factors to determine their impact on two different levels of treatment outcome: any improvement after 3 months and great improvement after 3 months. These were defined as follows:
-
1.
Any improvement at follow-up was defined as giving an answer to the PGI-I questionnaire that indicates improvement that is “a little better,” “much better,” or “very much better.”
-
2.
Great improvement at follow-up was defined as rating one’s condition as “much better” or “very much better” according to the PGI-I questionnaire.
These outcomes were chosen for two reasons: first, to ensure a patient-centered approach and second, the definition of great improvement has been used in other studies [10, 13, 17]. Owing to the large sample size, we chose to analyze the impact of the factors on two levels of improvement to better describe the impact of the factors and thus avoid drawing conclusions from randomly significant factors.
Statistical analysis
The factors were first assessed using descriptive analysis to summarize and describe demographics and baseline characteristics. Data were further analyzed using the Pearson Chi2 or Fisher’s exact test, as appropriate, and univariate logistic regression to assess possible associations with outcome. Next, regardless of univariate associations, the factors were all included in a multivariate logistic regression model. We analyzed separate multivariate logistic regression models for each outcome variable. For both outcomes, the following factors were entered into the multivariate regression model: age, educational level, current residence, total ICIQ-UI SF score, type of incontinence, frequency of PFMT, and frequency of app use. Factors were stepwise removed according to significance level until only significant variables remained (p ≤ .05). Age was included as a categorical variable in the multivariate analyses, as it better described the relationship with the outcome. Throughout the analysis, age was included in the models to control for possible age differences.
As the app is used to support PFMT, the variables “frequency of PFMT” and “frequency of app use” were analyzed for correlation and the models were also analyzed with these variables separately to find which variable better describes the relationship with the different levels of improvement.
As the questionnaires could only be submitted if all the questions were answered, there was no missing data from the completing participants. All statistical analyses were performed using SPSS version 26 software.