To our knowledge, this is one of the first randomized controlled trials showing that daytime sleepiness improved by additional 1.7 score points (to follow-up ESS values of 5.5 ± 3.9 in −SC + DPS vs. 7.6 ± 4.1 in SC) and (auto)CPAP therapy adherence increased by additional 70.1 min/d (338.8 ± 106.8 in SC + DPS vs. 268.7 ± 122.1 in SC) significantly and to a clinically relevant extend in patients with severe OSA when using a digital patient support tool in combination with an (auto)CPAP device compared to standard care alone after 12 weeks. Though other studies such as Malhotra et. al. [32] recruited a large sample size, their evidence level is low as they are not randomized controlled. The difference between the two groups illustrates an area between two health conditions (lower normal vs. higher normal daytime sleepiness) that might make a difference for a single patient`s well-being. Especially patients with an ESS Score > 10 at baseline are at risk to fail in normalizing the ESS Score during the first 12 weeks of therapy [30]. Additional improvement of daytime sleepiness through the digital patient support might have a significant positive effect on patients’ daily life like improved concentration, the experience of alertness or being less sleepy while driving. Therefore, it has the potential to strengthen the therapy effect in the decisive initial therapy phase.

Obtaining a frequent and long nightly PAP device use is mandatory to achieve clinical benefits by PAP therapy, such as enhanced functional outcome, reduction in blood pressure, or improvement in memory [3,4,5,6]. Our hypothesis that a digital patient support tool would improve adherence was confirmed and is in line with previous investigations [22]. With the prototype of the mobile app, even an increased improvement could be achieved.

The impact of telemonitoring on improvement of adherence was already demonstrated in former studies: Aardoom et al. found in a meta-analytic review of 18 RCTs with eHealth interventions that on average the improvement in therapy adherence was about 0.5 h [22]. Bouloukaki et al. [33], showed a significant improvement of 90 min/d CPAP usage and 3 points in ESS score after 24 months as a result of additional visits, telephone calls and education in 3100 OSA patients. Hwang et al. showed a significant effect on 90-day CPAP use in a 4-arm randomized controlled trial with 1455 OSA patients of 60 min/d after 90 days with a web-based OSA education and automated patient feedback in addition to standard care [19]. However, telemedicine-based patient education alone resulted in no significant adherence improvement. Furthermore, no differences in ESS or the FOSQ-10 score in any intervention study arm compared to standard care was observed. Bouloukaki et al. [33] concluded that additional video education session (15 min), a lecture from the sleep clinic’s registered nurses (10–15 min), phone calls by the nurse on the 2nd and 7th day to discuss any concerns, followed by home visits in case of concerns, two additional reviews by the sleep specialist, invitations to discuss therapy barriers, experiences, concerns, fears and beliefs, etc. is a time-consuming care beyond measure and difficult to integrate into clinical routine, the author discloses that the “intensive intervention entailed an additional cost of 30% above the cost of the standard intervention” [33]. Munafo et al. [34] showed in a prospective study that a web-based automated telehealth-messaging program reduced the time expenditure for healthcare professionals by 59%, compared to standard care, with maintaining adherence and effectiveness.

Considering former studies and our current outcomes concerning the impact of telemonitoring on (auto)CPAP therapy results, the use of the app prototype alone does not constitute the reason for altered ESS and FOSQ scores in our study results. Rather, the improvement is a consequence of increased (auto)CPAP use. As Antic et al. [30] already confirmed, indicators for subjective daytime sleepiness, functional status, and sleep-related quality of life, such as the ESS and FOSQ scores, are dose-dependent on CPAP treatment, with greater improvements in more-adherent patients. Regarding a presumable relation of (auto)CPAP adherence and ESS score, our results show an association of (auto)CPAP adherence with a reduction of ESS Score and are thereby in line with the meta-analysis of Li et al. [35]. However, the impact of factors such as age, BMI, baseline ESS, etc. is at least equally strong. Therefore, a prediction of residual daily sleepiness would have to include several factors and the identified discrimination threshold of less than 299 min adherence as predictor for residual sleepiness is not likely to be transferrable to patient cohorts with different characteristics. To obtain significant results regarding the correlation between adherence and follow-up ESS, a larger sample size is needed.

The significantly improved ESS and the trend in the FOSQ score compared to standard care in our study indicates that additional digital patient support has a positive impact on daytime sleepiness and functional status in adults. This might be reducible to our carefully designed automated feedback that considered OSA patient characteristics (well-documented in the thesis “PSI theory and adherence in nightly PAP therapy” of the psychology student M. Michalzyk) and the impact of the possibility to set individual and reachable personal goals on human decisions [36]. In addition, the personal contact patients made during home therapy phase might have affected their outcome. When looking at the number of phone contacts or on-site appointments, the group using the app prototype contacted the sleep therapist more frequently. However, considerably less than 1 contact per patient was registered on average and 12 of those contacts (made by 9 patients) are referrable to the study design or immature technology of the prototype as participants did not receive the welcome email or reported problems with data transmission by modem. This even might have led to frustration during the early therapy experience of these patients but apparently not to a degree of significant negative impact on their therapy adherence. However, initial technical difficulties cannot be generally ruled out in the future. Nevertheless, problems with data transmission by modem become obsolete by providing the possibility to use Bluetooth. A missing welcome email becomes irrelevant outside the study, while installation problems still could pose a small risk. The more frequent contact of the SC + DPS group even might have had a positive influence on these patients` adherence due to personal interaction with healthcare professionals and a possibly related motivating effect. In addition, the experience of solving a problem-like getting transmission of data started might have had an inspiring influence on SC + DPS group. As contacts due to therapy use or missing therapy benefit were mainly made by the SC + DPS group, the examination of the automated therapy feedback might have been the more relevant factor than the contact itself. Adherence of the subgroup with contacting patients is with less than 2 min not higher than in the total cohort (306.1 ± 126.9 compared to 304.5 ± 120.4) and over two-thirds of the population had no personal contacts at all. No difference was found in age, BMI, diagnosed AHI and oAI and in baseline ESS and FOSQ score between both groups. In addition, follow-up data shows no difference in the applied therapy pressure. Therefore, it is far from likely, that either the number of personal contacts nor other patient-related factors are the cause for increased adherence and possibly related improvement of ESS in the SC + DPS group.

The trend of lower leakage is due to the app use, as the application is designed to monitor and support the user’s therapy on a daily basis by providing personalized, schema-guided feedback, electronic questionnaires on potential problems and links to explanations and videos on therapy and the handling of therapy equipment. Hence, automated feedback on acute mask leakage may help the patient to check and improve mask fit or suggest a mask change. It is, therefore, no surprise that more than twice as many responses regarding mask problems were recorded in the standard care group. With the background, that CPAP pressure can be associated with an increased risk of unintentional leakage [37], it needs to be mentioned that in our study no difference between both groups was found in the used therapy pressure after 12 week follow-up.

No difference in the AHI/oAI follow-up was identified between SC and SC + DPS. The cause might be that both groups were already well-adjusted during titration night and no readjustment by the sleep center had to be recommended for the SC + DPS group by the app prototype.

Furthermore, a mean residual AHI of 4.1 n/h in the standard care group indicates how effectively the (auto)CPAP therapy devices used in this trial treat sleep apnea in clinical routine without additional digital support. The clinical benefit of those therapy devices even without additional DPS is also clearly recognizable as after 12 week adherence is high and ESS and FOSQ score improved significantly, even though patients showed only moderate daytime sleepiness at baseline. Furthermore, therapy termination rate was particularly low in both study arms which might be the result of the high level of care provided which is confirmed by patient satisfaction with the PAP therapy in general and with focus on therapy instructions and support.

Personal digital support in terms of an email-based mobile app prototype was well-accepted by PAP patients and worked reliably. In addition, the DPS was consistently used as a supportive tool and not as a replacement of conventional therapy support, as all patients continued standard therapy care while using the DPS.

In comparison with previous reports like Munafo et al. [34] or Malhotra et. al. [32], our study shows better improvement in adherence and consequent reduction in daytime sleepiness. We found no difference in age, BMI, diagnosed AHI and baseline ESS score compared to these reports. The better results might be due to a more intuitive or attractive handling of the tool. In addition, differences in comorbidities, which were not particularly investigated in this study, might be a reason for the success of our digital patient support tool.

All subjects followed the identical initiation of therapy and standardized therapy follow-up procedure according to clinical routine. SD and SD + DPS group had the same education session held by a respiratory therapist about OSA and its consequences, proper use and maintenance of the PAP device, mask fitting, and therapy and study expectations. All patients were provided with the same (auto)CPAP device types.

This study was designed as a prospective, randomized and controlled trial. After the enrollment, participants were randomized to either SC or to SC + DPS group. Consideration of the baseline data shows that randomization was successful. No difference was found in age, BMI, diagnosed AHI and oAI and in ESS and FOSQ score between both groups. In addition, follow-up data shows no difference in the applied therapy pressure. The drop-out rate was particularly low, as only six subjects in the SC and six in the SC + DPS group were excluded from the study, what most probably is a result of the excellent care and education provided by the sleep lab and homecare provider personnel. This is also reflected by generally low therapy termination rates and high adherence values in both groups, posing a significant challenge to the DPS tool to show an additional positive effect.

Through our eyes the homogeneity of the groups concerning demographic, baseline and follow-up data are seen as a huge study strength as none of the factors mentioned above can be the cause of the significant differences in adherence, ESS score, and duration of stable respiration between groups. In fact, the automated feedback seems to have a strong impact on these parameters.

Malhotra et al. [32] showed the potential contribution of a patient management app (Res Med “My Air”) to improve adherence, using big data and propensity score matching. This was not a randomized study, and thus prone to possible selection bias; those who were highly motivated to treat with a high level of education might have used this application selectively. Our current study, despite being small in scale, has the strength of being randomized, which clears this bias.

Munafo et al. [34] showed that a web-based automated telehealth-messaging program involving healthcare providers to manage patients reduced the time expenditure by 59% with similar results in adherence and effectiveness compared to standard care. To our knowledge no randomized controlled trials exist that show improved adherence and daytime sleepiness based on an only patient-managed automated coaching tool compared to standard care like our current trial. The automated monitoring and feedback of sleep status, identification of problems and presentation of solutions lead to patient empowerment and psychological reinforcement.

The current investigation had limitations that must be considered.

First, the cohort of participants consisted of patients diagnosed and treated in the same center and with severe OSA only. Hence, the transferability of the study results to patients with mild or medium OSA is limited. Participants were predominantly male and obese, which also impedes the transferability to other patient groups. However, distribution of sex and BMI is in line with those in similar studies [16, 38,39,40] indicating that the patient population in the current study is representative for the use of telemedical support. Furthermore, the distribution of sex within the two groups was identical (12 women and 38 men each).

Second, the observation period of the study was only 12 weeks, although the fact that adherence decreases within the first months is well-known [8, 10, 41]. However, there is no evidence that the change in adherence of the two groups would differ over a longer period of time.

Third, time expenditure per patient and per study arm was not determined. However, in the staff's assessment, which was verbally communicated, the SC + DPS group did not generate a higher time expenditure compared to standard care group. Evidence is the number of phone contacts or on-site appointments. If the feedback due to technical problems with the app prototype is not taken into account, the PAP therapy of the SC + DPS group has generated about the same amount of follow-up contacts. Nevertheless, it is impossible to distinguish how much time the individual study groups generated per response.

Another limitation is that the unknown influence of home environments (e. g. relatives helping with setup/operation of the app). This was not considered in previous studies and is difficult to measure.