The MinDag app for symptom monitoring and feedback to patients and clinicians during the initial treatment of bipolar disorder – a feasibility study

Abstract

Background

The app “MinDag” (MyDay) was developed as a tool for monitoring mood, symptoms and illness-relevant behaviour in the initial treatment of bipolar disorder. Digital self-monitoring may provide patients and clinicians with valuable data for tailoring treatment interventions. This study aims to evaluate the practical use and clinical implications of integrating MinDag in the early treatment of bipolar disorder from the perspectives of both patients and clinicians.

Methods

The MinDag app includes six content modules covering mood, sleep, functioning/activities, substance use, emotional reactivity, and psychotic symptoms. Patients were asked to use the app for six months, and automated feedback based on the app registrations was delivered to the patients’ clinicians biweekly. The study involved quantitative evaluations completed by patients (n = 20), as well as interviews with patients (n = 7) and clinicians (n = 2).

Results

Overall, the patients reported that they felt that MinDag was safe, relevant and easy to use, although technical difficulties such as too many automated reminders and need for reinstallations were reported. The patients appreciated the potential for increased awareness of their mental health, but expressed a desire for direct access to their data. Clinicians found the visual reports and feedback useful for tailoring treatment, even though the alert system for high-risk variables needed refinement. The quantitative evaluations indicated a positive general reception, with suggestions for improvement in usability and accessibility.

Conclusion

The MinDag app shows promise as a tool for enhancing the treatment of bipolar disorder by facilitating self-monitoring and providing actionable data to clinicians. However, technical issues and the need for direct patient access to data must be addressed. Development of digital tools to support the treatment of bipolar disorder and other mental health conditions is resource demanding, and there is a need to clarify criteria to establish proof of concept to guide the selection of tools for upscaling and implementation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40345-025-00382-x.

Background

Bipolar disorder (BD) is a severe mental disorder affecting 1–2% of the population (Moreira et al. 2017). Forty to fifty% of patients experience a recurrence within a year after an illness episode (Radua et al. 2017). Illness self-monitoring is believed to potentially prevent relapses in BD and thus improve outcomes (Morriss et al. 2007). Detailed monitoring of affective symptoms and other relevant parameters is also considered important to capture the large inter-individual symptom- and course variation observed and could increase our general understanding of the disorder (Bauer et al. 2004). Typically, self-monitoring involves tracking mood and sleep (Chan et al. 2021), but it can also include recording life events and/or behaviour believed to influence the course of the disorder (Faurholt-Jepsen et al. 2013).

Illness monitoring might support relapse prevention and recovery in BD in several ways. For the patient, it appears to enhance the awareness of, or changes in, symptoms including “early warning signs” (Depp et al. 2016), as well as to improve coping and disease management through lifestyle adjustments (Colom et al. 2009). In a clinical context, detailed illness monitoring can provide the clinician with information about the status of the patient, which may enable early relapse-intervention and enhance the optimization and selection of treatment strategies (Baldassano 2005). Illness monitoring is, therefore, an important ingredient in several well-established psychosocial therapies for BD, such as cognitive behavioural therapy (Miklowitz et al. 2008) and group psychoeducation (Colom and Vieta 2006). Traditionally, self-monitoring has been based on paper-and-pencil mood diaries, but digital app-based methods have been developed during the last decades with the increasing availability of smartphones, reducing recall bias and increasing adherence (Faurholt-Jepsen et al. 2016). These methods are generally usable and acceptable by patients (Faurholt-Jepsen et al. 2019a, b). Hence, digital monitoring may provide valid, fine-grained, and real-time information about symptoms and other relevant factors to inform more prompt, targeted, and personalized treatment.

Digital monitoring typically involves feedback to the patients and often to their clinicians and may be effective in improving outcomes, both as a self-management tool and through emphasizing important areas to focus on in treatment (Williams and Pykett 2022). However, the efficacy of digital illness monitoring is still under evaluation, and two recent meta-analyses have come to mixed conclusions. Based on three studies, the first paper reported that smartphone-based interventions (including mood monitoring) significantly reduce manic symptoms, with a trend in the same direction for depressive symptoms (Liu et al. 2020). The second identified 13 studies where the primary focus was on user engagement rather than the effect on illness outcome measures (Anmella et al. 2022). Of the thirteen studies, seven RCT’s (four of which incorporated illness self-monitoring) were included in meta-analyses of efficacy for several outcome measures, such as perceived stress, readmissions and affective symptoms. While all studies reported positive user-engagement indicators, none of the efficacy meta-analyses were statistically significant. The authors concluded that the studies and interventions reported are highly heterogeneous and that the effect of specific elements (such as mood monitoring) cannot be determined yet (Anmella et al. 2022). Thus, the research field is still in an early phase, affecting the speed with which digital interventions can safely be translated into clinical practice (Azevedo Cardoso et al. 2024).

To advance the digital monitoring field, several unresolved challenges are ahead. Some concern the technical properties of the digital infrastructure and the tools themselves, whereas others are related to the perceived usefulness and applicability in real-world settings. The difficulty with establishing the putative usefulness of digital monitoring may be associated with several issues. For a start, there is a need to standardize tools, measures and procedures (Faurholt-Jepsen et al. 2019a, b) whilst also personalising them to fit the setting and the needs of the patient, as well as minimizing potentially adverse effects (Depp et al. 2016). In addition, limited access to necessary technology resources and competence, including secure data solutions, may further complicate the development of digital monitoring. Concerning the implementation in real-world settings, it is necessary to investigate the application of digital tools in routine clinical practice to evaluate their efficacy. However, clinical use is contingent upon the willingness of both the patient and the clinician to test the tools and adopt new procedures. For the clinician, the daily schedule is often pressured, with back-to-back patient consultations yielding limited flexibility and extra time.

Consequently, new tools and procedures must be experienced as helpful and time-saving to be incorporated. For the patient, a balance between the time and effort spent using the tool as intended and the experienced benefit the tool has on the treatment received is essential. Support infrastructure may be necessary, particularly in early implementation phases, but is frequently not sufficiently provided. Indeed, implementing new interventions is a science of its own, which is often neglected or underestimated, reducing the success rates of both evidence-based and innovative approaches (McGinty and Eisenberg 2022).

In the current study, we aimed to test the usefulness and acceptability of MinDag (“MyDay”), an app for illness monitoring in BD developed by our research group. The app was initially developed as a research tool for data collection to better understand intra- and inter-individual symptom variation in BD (2022). The app is part of the research protocol of the BIP-DPS study, a naturalistic and longitudinal study with a specific focus on affective lability/emotional reactivity, substance use, sleep, and functioning in BD. The study is conducted in a clinical secondary care unit for patients receiving their first outpatient treatment for BD, and all patients treated in the unit are asked to participate. Further, we aimed to evaluate if MinDag was experienced as a useful addition to clinical practice by both the patients and their primary clinicians. Three focus areas were of specific interest: (A) Practical experiences associated with app-based illness self-monitoring, (B) Experiences with the use of MinDag as a means for increasing awareness of symptoms and symptom triggers, and (C) Experiences with using output from MinDag in clinical treatment sessions. To explore and shed light on these areas, we used data from self-report evaluation forms supplemented by interviews with patients and clinicians. Finally, the current study describes the available infrastructure during the implementation process. In addition, we discuss what we consider necessary elements for success when implementing new digital tools in clinical practice, based on our experiences with the MinDag project thus far.

Methods

Setting and participants

The MinDag app is used as a tool for prospective data collection at the Bipolar Unit at Nydalen District Psychiatric Centre (DPS), Oslo University Hospital (OUS). The Bipolar Unit is comprised of psychologists and psychiatrists specialized in the assessment and treatment of BD, and research is integrated in the services provided. The research activities are conducted in collaboration with the Section for Clinical Psychosis Research, OUS, which was part of the former Norwegian Centre for Mental Disorders Research (NORMENT). Patients aged 18 to 65 years with a suspected episode of mania/hypomania or newly diagnosed bipolar spectrum disorder are treated in the Bipolar Unit, serving a catchment area of 138.000 inhabitants representing all socio-economic strata of Oslo. The purpose of BIP-DPS is to investigate illness mechanisms in the early phases of BD, where digital phenotyping, including the use of MinDag and actigraphy, is part of the study protocol. The BIP-DPS study has been approved by the Regional Ethics Committee (REK 2018/208) and is based on informed consent.

For the current sub-study, anonymous self-report evaluation forms (n = 20) were collected from patients who used the MinDag app and participated in the early stage of the BIP-DPS study from January 2020 – September 2022. The forms were anonymous in order to ensure that the participants could give feedback as freely and openly as possible. Only patients taking part in the BIP-DPS study were offered to use the app. Unfortunately, we do not have an exact number of how many were invited to take part, but in general, 40–50 patients are referred to the Unit per year. Of these, approximately 30–40 are deemed eligible for the study, but not all are offered to participate for several reasons such as too complex and/or continuously acute clinical presentations, short treatment periods, ambivalence towards treatment in general etc. A limited number of those invited to participate refuse to do so.

The patients in the current study were individuals with bipolar I, II, or not otherwise specified disorder who had recently started their first treatment in the Bipolar Unit. The diagnosis of bipolar disorder was verified for all patients based on the Structured Clinical Interview of DSM 5 Clinician Version (First et al. 2020). The mean age of the patients was 32 (range 21–54; SD 10.8), and 11/20 were female (55%). Of these, seven (three females, four males) agreed to participate in further interviews to elaborate on their experiences and give suggestions for improvement to supplement the information obtained via the evaluation forms. In addition, two clinicians from the Bipolar Unit; a psychiatrist and a clinical psychologist, who had followed many patients through the completion of the study period were interviewed about their experiences with the app.

The MinDag app

The MinDag smartphone app was developed by a project group from the Section of Clinical Psychosis Research, OUS. The group consisted of psychologists, psychiatrists, and a user representative, and collaborated with a team of application engineers and visual designers from the Web application group at the IT Department (USIT) at the University of Oslo (UiO). The project group developed the content and framework of the app, while USIT was responsible for programming and applicability. Data collected from the app is PGP encrypted and transferred to a secure server (Services for Sensitive Data, TSD) after each completed task. The data is stored as csv files on a dedicated server in TSD. For more details about the app setup, data security- and management, please refer to Bjella et al. 2022. The version of the app used in the current study covered six selected content modules that are important for the course and outcome of bipolar disorder: mood, sleep, functioning/activities, substance use, emotional reactivity, and psychotic symptoms. The first three modules were logged on a daily basis, whereas the latter three were logged weekly. The background for the choice of content and variables in the app can be found in the MinDag development article by Bjella et al. 2022.

Study infrastructure and description of the feedback system

MinDag was installed on the patients’ private smartphones, and the initial user registration was completed together with their primary clinician or a project coordinator. This was to ensure that the notifications and prompts were set up according to the wishes of the patient, who could decide what days and times were most convenient for answering the different modules. The patients were asked to use the app for six months, but were informed that they could discontinue use at any time. After start-up, the clinician or project coordinator checked in regularly with the patient to ensure there were no technical difficulties or other complications. Feedback based on the responses to the items in MinDag was not presented to the patient directly in the app. Instead, the clinician received visual reports with individually customized graphs fortnightly. Specific variables could be selected to overlap with existing treatment goals, and several variables could be plotted in parallel curves (e.g. degree of sadness and hours of sleep). The time periods displayed in the visual plots could be selected freely and range from one week to the complete period of use. Together, clinicians and patients could use this information in the treatment to understand how different symptoms and/or behaviours were interrelated, decide upon new focus areas, and target interventions accordingly. Please refer to Supplementary material 1 to see an example of the visual reports. In the test period of this project, a preliminary alert system was put in place where high scores on selected variables triggered an alert (the variables are listed below). The thresholds for triggering an alert were conservatively set to avoid too many false positives. This was considered to be acceptable as all of the participants in the study were patients in the Bipolar Unit and thus in frequent contact with their primary clinician. The alerts were checked by a member of the project group and then passed on to the relevant clinician in charge of contacting the patient to check if everything was okay or if any acute measures were needed. The alert system will be refined, adjusted, and developed further for the continuation of the project.

Alert system variables

• Mood variables: high scores (equal to, or above, four on a seven-point Likert scale) at least four out of the past seven days on feelings of sadness, panic, and elation.

• Sleep: equal to, or less than, four hours of sleep per night for the past seven days.

• Psychotic symptoms: two consecutive weeks with scores of equal to, or above, four on a seven-point Likert scale covering hallucinations, paranoid delusions, and two self-selected items of unusual ideas.

• Non-use of the app every day for a week.

Patient self-report evaluation forms

To get systematic feedback from the patients concerning the app’s usability, content, and acceptability, the project group developed a self-report evaluation form (Supplementary material 2), which was handed out to all patients after completing their project period. The form contained questions pertaining specifically to MinDag, in addition to more general questions adapted from the System Usability Scale (SUS; Bangor et al. 2008). There were twenty items in total concerning app use; the first ten items were scored on a Likert scale from one to five (completely disagree-completely agree), whereas the remaining ten had varying types of scoring instructions (please refer to supplementary material for specifics). Further, the evaluation form had several sections for comments where the patients could elaborate on their responses and provide individualised feedback and suggestions for improvement. Data were analysed with SPSS version 29 and are presented in Figs. 1 and 2.

Fig. 1.

Practical use and content

Fig. 2.

Self-monitoring and clinical use

Interviews with patients and clinicians

Interviews were developed and conducted with both patients and clinicians as a supplement to the evaluation forms to get more in-depth accounts of user experiences concerning practical aspects and the clinical potential of using the app. The interview guides cover questions regarding practical use, technical issues, feedback, and experiential aspects of registering potentially sensitive issues. In the clinician version, questions also include the perceived motivation of the patients, the clinical potential and relevance of using the app, as well as suggestions for improvement. Please refer to Supplementary material 3 for the interview guides. Interview responses were audio-recorded and/or written down during interviews. They were further categorised into three evaluation categories. Sample quotes are provided to illustrate examples of the feedback given by the responders.

Results

We have chosen to categorise the feedback from both the evaluation forms and the interviews into three main categories to shed light on our focus areas of interest, namely: (A) Practical use and content, (B) App-based self-monitoring as a tool to increase awareness of own mental health (results presented only for patients), and (C) Clinical use. Due to technical issues causing prolonged periods of missing data, we did not measure adherence rates.

Self-report evaluation forms

A: Practical use and content: The feedback from the evaluation forms indicates that patients generally found the app safe, relevant, and easy to use. However, there were several periods with technical difficulties where the automated reminders were sent out too frequently, or the app stopped working. These technical issues made it necessary to delete and update the app occasionally, including repeating the start-up registration process. The patients’ responses on the items most relevant to usefulness and content are shown in Fig. 1.

B: Self-monitoring and C: Clinical use: Overall, the patients reported positive experiences with participating in the study. Concerning the benefits of using MinDag as a self-monitoring tool in a clinical context, the majority of the patients found the app to be helpful as a part of the treatment program in the Bipolar Unit, as well as a tool to monitor their mental health. Still, about a third of the patients did not find MinDag to be a useful tool in this respect. Comments in the open text fields indicated that this was primarily related to the lack of direct feedback (e.g. curves, data summaries) based on the registrations in the app. In terms of the putative value of using the feedback from the app actively in the treatment process to personalise interventions, this appeared to be contingent upon consistent and structured use of the visual reports in treatment sessions.

The responses of the patients on relevant items of the form are shown in Fig. 2.

Interviews: patients

Input that contributes to shed light on the three focus areas of interest is highlighted below (quotes from the patients in italics):

Focus area A: Overall, the patients reported positive experiences with the use of MinDag. They found the app easy to use and stated that registering data through the app quickly became a part of their daily routine: “It was okay to use the app every day, it became a habit”. It is worth noting that despite positive experiences, logging data was considered to be a bit of a hassle by some patients: “It was interesting to see data from week to week, but sometimes it was tedious to register everything every day”. Several patients commented that using the app made them feel like they contributed to important research: “It was great! I would like to participate more!”; “Proud to help with gaining more knowledge about bipolar disorder”. The technical issues with the app were a concern for many and influenced the user experience and motivation: “The app did not work in an optimal way and there were lots of technical issues that made it hard to use”; “Participation in the project in itself was positive, but all of the technical problems made me lose interest”.

Focus area B: Using self-monitoring to increase awareness of symptoms and-triggers appeared to be of value to the patients: “It was a good source for increasing knowledge about my own patterns and the effect of interventions”; “It has made me more aware of sleep and other things in my everyday life”. However, for two patients, this awareness was not always helpful: “Gives better overview and contributes to coping, but I also started to look for symptoms, can make it worse”; “For me, personally, it was not beneficial to think about how I am doing every day”.

Focus area C: The experiences with using output from MinDag in treatment sessions were mixed. Some patients found the visual reports hard to read and thus of little value: “The graphs were impossible to read even with the help of the therapist, they were not straightforward”, whereas others appeared to find them more intuitive and consequently helpful: “The graphs made me reflect over feelings and become more aware”. Several of the interviewed patients pointed out that it would be beneficial to access registrations and reports directly from the app and not only through the clinician: “It would be good if the data from the app were more easily accessible”.

Interviews: clinicians

The two clinicians interviewed were selected because they had referred most of the patients to the study. The purpose of the interviews was twofold: to get input on the technical and practical aspects of administering the start-up and follow-up of the patients using the app (focus area A), as well as potential clinical implications of making use of the visual reports and alerts during the treatment course to tailor and personalise treatment (focus area C). Quotes from the clinicians are highlighted in italics.

Focus area A: Both clinicians found that giving information about how to use the app and the start-up was uncomplicated: “Explaining the purpose of app-use was straightforward”; “The start-up was relatively intuitive”. In line with the feedback from the patients, the technical difficulties were noted as deleterious to the user experience and something that affected motivation: “There were many technical challenges. The patients were determined to use the app, but they got annoyed after too many push-notifications”. Concerning the practical technicalities of follow-up during the project period, one of the clinicians pointed out that time is limited and everything that is added to the schedule, such as printing out the visual reports, is a potential obstacle: “It was difficult to remember to print out the visual reports in advance of the sessions. And then there are lots of things happening, many topics that need to be covered in each session, so it’s easy to forget to use them”. Further, both clinicians mentioned that the visual reports and graphs had room for improvement when it comes to readability: “Needs work on the graphic and visual representation of the data; show change – through colours, figures, visualisation of trends”; “The reports are not necessarily easy to digest”.

Focus area C: The clinicians were positive about using the information gathered through app-based monitoring in treatment and indicated that this added value in evaluating and personalising interventions: “Yes, it’s useful to use the visual reports in the treatment. It provides a sort of structure to the conversation concerning how things have been over the past weeks. If you see big changes, then that is a good place to start, it puts a kind of spotlight on something that has been going on. It is useful if you are working specifically with something, sleep for example. The visual reports make you take something seriously, they underline a problem more closely”; “If the visual reports confirmed something that we knew, or showed something new – a discovery, that was exciting and motivating”. The need to improve access to data and graphs was emphasised by the clinicians as well: “The patients want to be able to check their registrations and see the reports in the app – this would be beneficial for us also, time-wise”. There was further a suggestion for improvement that might add to the clinical relevance of the app: “If variables are “red” – if there was an alert, they could be accompanied by advice/help texts around how to improve sleep, mood, tiredness. This could be very relevant – a reminder of the importance of self-care”.

Discussion

In this study, we assess the use of the app MinDag as a tool for symptom monitoring and feedback to patients and clinicians during the initial phase of treatment of patients with newly diagnosed BD. The results suggest that MinDag is an acceptable tool for digital data collection and clinical use, with a limited burden on the patients and high tolerability, as the patients were able to use the app for several consecutive months. However, a positive user experience is contingent on minimal technical difficulties as well as consistent and stable functioning of the app. From a clinical perspective, the results indicate that for those who actively used the feedback from the app in treatment together with their clinician, the app was of value and had potential as an add-on to more traditional methods of evaluation, enabling further personalisation of interventions. The supplementary feedback obtained through the interviews was well aligned with the information gathered from the evaluation forms. Since this is not a controlled or blinded study, we cannot exclude e.g. response bias in the data, but we attempted to minimize this risk by developing the survey in collaboration with a lived experience expert to ensure neutral wording and coverage of all relevant aspects of experience with the app. We also invited all study participants to take part in the survey.

Focus area A: practical use and content

Although the patients were positive about the usefulness and acceptability of MinDag in terms of safety, relevance, and ease of use, the periods with technical difficulties, including errors with the automated notifications, took their toll on the overall user experience and created some annoyance. In particular, the occasional need to delete and reinstall the app may have contributed to missing data registrations and periods of non-use. The negative impact of technical challenges was reported in the feedback from the evaluation forms as well as the interviews and indicates that system stability is a precondition for successful implementation in clinical practice. In other words, there is little room for start-up issues and the functionality of digital tools should be tested thoroughly in pilot studies before large-scale implementation in the clinic.

Focus area B: self-monitoring

Some patients experienced that increased awareness of symptoms and symptom triggers through self-monitoring could be negative, and this is an important issue that is likely to generalise to other digital self-monitoring tools. Indeed, increased attention towards symptoms because of frequent self-monitoring has been hypothesized as an explanation for the increased affective symptoms seen in at least two randomized controlled trials (RCTs), and consequently, caution is warranted (Faurholt-Jepsen et al. 2020, 2021; Palmier-Claus et al. 2021). To remedy the risk of adverse effects, clinicians must address the putative challenges associated with daily recordings of symptoms in the start-up conversations with their patients before app use is initiated. As pointed out to the patients in the current study, it should be possible to discontinue app use at any time, and the threshold for reporting any distress associated with illness monitoring should be low. On the other hand, RCTs have also found digital interventions in BD to be associated with reduced perceived stress and rumination and increased quality of life. Thus, the benefits and costs balance appears complex and is most likely person-specific. To decrease the risk of negative experiences, apps should be flexible with possibilities for personalisation regarding both the frequency of registrations and the choice of modules. Also, using a combination of active self-report and passive digital phenotyping is likely to be the most acceptable and efficient way of monitoring illness activity over time.

Focus area C: clinical use

For the MinDag app to be clinically useful with respect to personalising treatment interventions and optimising patient care, the systematic use of visual reports and feedback is fundamental. Improving accessibility and presentation of the data is likely to make patient consultations more time-efficient and increase patients’ ownership of the data, something which might boost motivation to use the input provided actively. In addition, successful implementation of symptom-monitoring apps as part of regular clinical practice may require stronger scientific evidence, in parallel with repeated clinical experience that such monitoring indeed improves clinical outcomes for the majority of patients.

Limitations

Although a sample size of 20 is not uncommon in digital feasibility studies, the small sample is a limitation of the current study. Further, we are unfortunately not able to report on the exact proportion of participants with complete project periods as the data from the evaluation forms is not linked to the app use data for each participant. Finally, the interviews with patients and clinicians were initially conducted to supplement the information obtained through the evaluation forms. Hence, a strategy for analysing the feedback from the interviews with qualitative methods (i.e. thematic analysis) was not developed a priory and could not be devised retrospectively.

Future directions and conclusions

Developing and implementing digital tools for mental health treatment is resource intensive and requires substantial funding, transdisciplinary expertise, as well as time for iterations of development and testing. This is not easily feasible in routine clinical practice where schedules are tight, and the top priority is meeting the patients’ acute needs. Although the clinical potential of digital tools such as apps in treating BD and other mental disorders needs further investigation, boosting the culture and readiness for innovation in hospital environments appears important. To successfully develop tools that can readily be used in clinical settings, integrating technical and clinical expertise into the health services, or through collaboration between tech enterprises and health care professionals, is necessary. Furthermore, to determine what tools should be upscaled and implemented and not, a clarification of criteria for establishing proofs of concept and evidence for profit realization and efficiency is also an important issue for future investigation.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

TVL conceptualised the study and drafted parts of the manuscript together with MCH who also collected data for the study. SRA, ME, SHL and SHG contributed to data collection and coordination of the study, SHG conducted one of the patient interviews, and IM and TB provided significant scientific input to the study. All co-authors contributed to, reviewed and approved the final manuscript.

Funding

The study was supported by grants from the Norwegian South-East Health Authority (20/00189 − 19), and The Research Council of Norway to the NORMENT Centre of Excellence (Grant Numbers #223273/F50 and #287714).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was part of the BIP-DPS which was approved by the Regional Ethics Committee (REK Sør-Øst 2018/208) and the Safety Representative at Oslo University Hospital, and is conducted in line with the Helsinki declaration of 1975 (as revised in 2008 and 2013). All participants gave written consent to participate and could withdraw from the study at any time.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.