This pilot study was conducted at the University Women’s Hospital in Heidelberg, a perinatal center of the highest level, performing over 2300 deliveries per year. Pregnant women hospitalized due to high-risk pregnancies were asked to participate in the study. The criteria for eligibility included age of 18 years or older, fluency in the German language, a gestational age of ≥24 and ≤34 weeks, and the ability to access to the internet. Women were not eligible to participate if they were expecting multiples. In total, 90 inpatients were asked to participate, of whom 68 women agreed. The reasons for not participating included lack of interest, scheduling conflicts (clash of dates due to lack of time), or severe pregnancy complications. After enrolling in the study, every participant was provided with a tablet and free wireless internet service and access to the eMBI course (Figure 1). The app was designed and developed by an interdisciplinary team of gynecologists, psychologists, and midwives by the Institute for Women's Health Tuebingen, Germany. The study participants took part in a brief 1-week course of mindfulness through an eMBI based on the mindmom app, which is currently being examined in a prospective randomized trial (trial registration DRKS 00017210). The RCT aims to examine the clinical effectiveness and cost-effectiveness of an eMBI in a sample of pregnant women during the third trimester of pregnancy who were screened positive for emotional distress according to the Edinburgh Postnatal Depression Scale (EPDS). The screening is administered while attending ambulatory prenatal care either with a registered gynecologist or at one of the study centers. The participants were randomized in 1:1 into the intervention (eMBI) or control (usual treatment) group.

The original intervention was used prenatally in an established concept of 8 weekly 45-min sessions [39]. In this pilot study, the time frame was tightened to a 1-week version, including a total of three 45-min modules on mindfulness (Table 1). The pilot study was carried out to optimize the user-friendliness of the mobile app by making appropriate adjustments. In addition to the questionnaires, semistructured interviews were conducted. Feedback from the participants on user-friendliness and suggestions for improvement (semistructured, qualitative evaluation) were taken into account in the final concept of the content and structure of the mobile app. We chose to use an inpatient population as the convenience sample for a population at heightened risk for mental health conditions and to explore whether eMBIs could represent accessible mental health resources to support hospitalized patients during pregnancy.

Through mediation of psychoeducational content and cognitive behavioral therapy-related approaches, the app teaches participants how to deal with stress, pregnancy-related anxiety, and symptoms of depression. Thus, it promotes the autonomy of the mother-to-be regarding the upcoming birth and during the initial days and weeks with the newborn. The app contains instructional videos and audio files, interactive worksheets, and a personal “skills box” to collect exercises, videos, and texts, which the participants found helpful. After the 1-week course, participants had the opportunity to continue accessing the exercises. All the participants received scientifically validated information about pregnancy and birth via a pregnancy counselor. The main topics were the physical changes in pregnancy, the birth process, pain relief during birth, bonding between parents and child, breastfeeding, and tips on the formalities related to birth. No financial compensation was offered to the participants. All questionnaires were completed digitally prior to and after the 1-week course. In addition, we gathered medical and sociodemographic data that were double-checked against the hospital records (Table 2). Data were collected via the mindmom app based on electronic patient reported outcomes.

To assess the psychometric data on depression and symptoms of anxiety, the following instruments were used: EPDS, State-Trait Anxiety Inventory (STAI), and Pregnancy-Related Anxiety Questionnaire (PRAQ).

All analyses were conducted using the SPSS software (IBM Corp, v. 24.0.0.0). Since the distributions of the self-report scales did not deviate from the normal distribution (P>.18 in Kolmogorov-Smirnov-test), we chose a parametric analysis strategy for the parametric variables. Due to scale-specific amounts of the missing values, the number of valid cases varies between the analyses. We considered sociodemographic data (eg, age), medical data (eg, gestational age), and the self-report data (eg, EPDS) for this procedure. The missing completely at random test results were not significant (χ²₃₄₁=334.4, P=.59), indicating that missing values were at random and that subpopulations were representative for the total sample [48].

First, the descriptive statistics of the sample characteristics in the relevant sociodemographic, medical, and self-reported variables were reported according to the levels measured. Second, the frequency of the cases scoring above the EPDS cutoff (>9) were compared to the expected frequency (11.1%) according to the review of Bennet et al [1] by using the chi-square test. The same analysis was used to determine the frequency of the women scoring above and below the STAI cutoff (>40) for equality. Third, the EPDS, STAI, and PRAQ-R scores were tested for a decline between the first and second measurements by using two-tailed t tests for paired samples. Finally, high versus low treatment compliance was tested for differences regarding the symptom levels by using t tests for independent means. The two-sided critical α-error was set to α=.05. Due to the exploratory nature of the analyses, the α errors were not Bonferroni-adjusted. To estimate the effect sizes, we computed ω² = χ²/N for chi-square tests and ω² for t tests. ω² is a population-based estimator of the explained variance. ω²=0.01 or ω²=0.01 is interpreted as small effects, ω²=0.09 or ω²=0.06 as medium-sized effects, and ω²=0.25 or ω²=0.14 as large effects [49].

In total, 68 hospitalized pregnant women were included in the study, 39 of whom completed the full 1-week electronic mindfulness course. The most common diagnoses of the participants are listed in Table 3. Of the 68 participants, the information for 1 participant regarding the diagnosis for hospitalization was missing; therefore, the information of 67 participants is provided in Table 3.

A total of 29 participants who completed the baseline visit were lost to follow-up. Thus, the overall completion rate was 57% (39/68). The nonparametric sociodemographic characteristics of the study population are summarized in Table 4. The parametric demographic, medical, and self-report data are summarized in Table 5.

Table 6 shows the observed and expected case numbers for the EPDS, the STAI-S, and the STAI-T cutoffs at both assessments. The chi-square tests were highly significant for the EPDS at both assessments (P<.001) with large empirical effects (ω²=0.778 at first and ω²=0.908 at second assessment). With 39% (26/67) of the participants at the first assessment and 41% (16/39) of the participants at the second assessment scoring above the cutoff (>9), there were significantly more participants with depressive symptoms in the sample than expected (11.1%). The chi-square tests were also significant for the STAI-S at both assessments (P=.008 at first and P=.04 at second assessment) with medium-sized empirical effects (ω²=0.105 at first and ω²=0.111 at second assessment). With 66% (45/68) of the participants at the first assessment and 67% (26/39) of the participants at the second assessment scoring above the cutoff (>40), there were significantly more state anxious participants in the sample than expected (34/68; 50%). The chi-square tests were not significant for the STAI-T at either assessment (P=.22 at first and P=.63 at second assessment). With 42.6% (29/68) of the participants at the first assessment and 46.2% (18/39) of the participants at the second assessment scoring above the cutoff (>40), the number of trait anxious participants was not significantly higher in the sample than expected (34/68; 50%). The power to find large effects in this analysis was sufficient for large effects (ω²=0.25, 1–β=.98 at the first and 1–β=.87 at the second assessment). However, medium-sized (ω²=0.01, 1–β=.70 at the first and 1–β=.47 at the second assessment) and small effects (ω²=0.01, 1–β=.13 at the first and 1–β=.10 at the second assessment) cannot sufficiently be ruled out.

Table 7 summarizes the descriptive and inferential results of the paired sample t tests on the changes between the first and the second assessment in the subsample of 39 participants who were eligible at both assessments. No significant change was found in the EPDS, STAI-T, or PRAQ-R scores (P>.20). The power to detect large (ω²=0.14, 1–β>.99) and medium-sized effects (ω²=0.06, 1–β=0.86) was sufficient for these analyses. At the same time, with a power of 1–β=0.23, small effects (ω²=0.01) cannot be ruled out. However, the STAI-S scores significantly declined for these patients between the first and the second assessments (P=.03) with a small effect of ω²=0.051 (ie, 5.1% of the variance of the changes between the first and the second assessment can be explained by the passing of time).

To examine the differences between the high and low app engagement regarding the symptom levels at the second assessment, we defined a group of participants that completed more than 50% of the 1-week course modules. Furthermore, we dummy coded noncompliant as “0” and compliant as “1” for >50% of all the 3 modules. There was no significant difference between high (8.96 [SD 4.30]) and low app engagement (8.30 [SD 4.05]) regarding EPDS scores at the second assessment (t₃₇=–0.49, P=.62). Moreover, there were no significant differences regarding the STAI-S scores (t₃₇=–0.60, P=.55) or the STAI-T scores (t₃₇=–0.60, P=.55) at the second assessment between high (STAI-S: 44.82 [SD 10.56]; STAI-T: 39.26 [SD 8.81]) and low app engagement (STAI-S: 42.85 [SD 9.79]; STAI-T: 37.63 [SD 8.27]). The power to find large effects in these analyses was insufficient for large (ω²=0.14, 1–β=.68), medium-sized (ω²=0.06, 1–β=.33), and small effects (ω²=0.01, 1–β=.09). Consequently, the effects of any size cannot sufficiently be ruled out.

However, a significant, medium-sized difference regarding PRAQ-R scores was observed at the second assessment (t₃₇=2.03, P<.05). If participants completed more than 50% of all the modules, they had significantly lower PRAQ-R scores (18.74 [SD 4.49]) than those with low app engagement (22.54 [SD 6.90]) at the second assessment. Overall, 7.4% (ω²=0.074) of the PRAQ-R score variance at the second assessment can be explained by the group variable “high vs low app engagement.”

In our study, the prevalence of depression and anxiety among hospitalized high-risk pregnant women was found to be high. At the baseline assessment, 39% (26/67) of the study participants achieved EPDS scores above the cutoff value for a minor/major depression and 66% (45/68) of the women had high levels of anxiety as measured with the STAI-S. Therefore, our results reinforce the findings of previous studies on depression and anxiety disorders in perinatal populations demonstrating regularly higher EPDS and STAI scores than the expected ranges in nonpregnant populations [50]. In our study, we used a meta-analysis including 19,284 patients for comparison that reported EPDS scores ≥10 in 11.7% and 11.1% of the women in the second and third trimester, respectively [1]. Regarding anxiety symptoms during pregnancy, recent reviews found pooled prevalence rates of 15%-23% across trimesters [5,51]. The study selection of these reviews was restricted to samples of pregnant women recruited through general obstetric/prenatal units. By contrast, our sample included hospitalized high-risk pregnant patients, in whom prevalence rates are considerably higher.

When compared to recently published studies with inpatient samples, we found similar tendencies for anxiety and depression levels during pregnancy. Regarding antenatal depression assessed via the EPDS, a study in Singapore found that the rate of major depression in a sample of high-risk pregnancies (11%) was higher than that in an unspecified obstetric risk cohort (4.3%) [14]. Dagklis et al presented similar results with high depression rates of 24.3% and 28% in 2 different high-risk pregnancy unit samples [52,53]. The scores in our sample exceed the aforementioned study results considerably, with up to 41% (16/39) of the participants reaching or exceeding the EPDS cutoff value. In a recent study examining depression, anxiety, and attachment among women hospitalized in an antepartum unit, screening identified over one-third (36%) of the participants to be at risk for depression (EPDS score ≥10) and almost half (47%, 46/98) reported elevated state anxiety (STAI-S ≥40) [54]. Likewise, Barber and Starkey showed that hospitalized pregnant women had significantly higher state anxiety levels, with 47% of the women showing STAI-S scores above the cutoff compared to normative data from nonpregnant samples [55]. These results are in line with our sample, with STAI-S scores >40 for 66% (45/68) of the participants.

In direct comparison, it becomes clear that hospitalized pregnant women represent a vulnerable group, showing rates of anxiety and depression more than three times greater than those reported in nonclinical samples. Despite the high rates of anxiety and depressive symptoms, a surprisingly low number of participants were receiving mental health treatment. In inpatient obstetric settings, where access to individual psychotherapy is often extremely limited, eMBIs could provide an easily accessible support option. Incorporating low-threshold eMBIs not only in hospital care but also in outpatient practices could minimize the stigma of starting mental health treatment.

For our sample, we decided to choose a cutoff value of >40 for STAI and a cutoff value >9 for EPDS, which is in line with that reported in recent studies on the use of EPDS and STAI as valid screening tools for depression and anxiety during pregnancy. Tendais et al found optimal cutoffs for STAI-S as 40 and EPDS as 9 during pregnancy [56]. Regarding the accuracy of the EPDS in identifying depression and other mental disorders, Howard et al reported a likelihood ratio of 9.8 for the EPDS [3]. The fact that the EPDS also performs well in screening for depression and anxiety in high-risk pregnant women was confirmed by Thiagayson et al, who recommended further psychiatric assessments for women with a score ≥9, as found in our study [14]. For the STAI, Barnett and Parker recommended cutoffs of high (≥40), moderate (32-33), or low (≤25) anxiety on the basis of the mean trait scores of a sample of 94 primiparae [57]. Grant et al found that a cutoff >40 for both state and trait scales yielded optimal sensitivity (80.95%), specificity (79.75%), and positive predictive value (51.5%) to determine cases of anxiety in the third trimester of pregnancy [46]. In our sample, there were 59% (40/68) primiparae, and the mean gestational age was 30.14 weeks; thus, the values compare well. In all, the EPDS and STAI prove to be effective screening tools that are frequently used in studies including high-risk pregnant women.

The latest reviews about eMBI during pregnancy suggest that web-based interventions targeted at improving mental health may be beneficial during the peripartum period. However, the findings and their generalizability are limited both by the heterogeneity of the interventions and study designs and by methodological limitations. Pooled results of non-RCTs reporting outcomes on anxiety, depression, and perceived stress showed a significant benefit for the mindfulness group, but this review found no differences between the mindfulness intervention and control groups in RCTs [58]. Another review showed significant reductions in depression, anxiety, and stress by means of preanalyses and postanalyses, each with small to medium effect sizes [31]. However, between-group analyses failed to find any significant postintervention benefits of MBIs in comparison to control conditions.

In our pilot study without a comparison group, a significant reduction in the mean state anxiety levels was found after completing the 1-week eMBI (P<.03). Recent studies including other MBIs during pregnancy found similar effects with significantly lower STAI scores (P<.001) after interventions such as yoga, music therapy, or progressive muscle relaxation [50,59]. Compassion-focused therapy represents another promising approach for effectively treating depression and anxiety in perinatal populations [60]. The equivalence of compassionate mind training on the constructs of self-confidence, inadequate self-criticism, and self-compassion and the superior performance of compassionate mind training in reducing depression and anxiety scores in the latest studies suggest that compassion-focused therapy offers additional benefits beyond the current gold standard of cognitive behavioral therapy [61,62].

As one of our key results, we found a small, yet significant reduction in the mean state anxiety levels after completing the 1-week electronic course of mindfulness (P<.03). Concerning engagement with the app, participants who completed more than 50% of all the 3 modules showed lower scores for the PRAQ-R at the second assessment (P<.05). Our sample population showed an overall completion rate of 57% (39/68). We decided to consider an individual completion rate of more than 50% of the overall module in at least 2 of 3 course modules as compliant, as similar cutoffs have been used before. For instance, previous studies on mindfulness-based programs have considered participants who completed at least 50% of sessions as having fulfilled an adequate minimum amount of the course [30]. Hence, our compliance rate is comparable to that reported in similar studies, which reported even lower compliance rates from 21% to 35% [30,63]. Regarding the nature of this very specific study population, the majority of the dropouts can be explained by hospital discharge before the second assessment. Women who are no longer considered to be at high risk of preterm delivery might not see any reason to continue the eMBI, as they have already experienced symptom relief due to the end of hospitalization. Nonetheless, the overall dropout rate of the course was high (29/68, 43%). The reasons for dropout included discharge from the hospital, actually giving birth prematurely, severe stress and concerns, or adverse pregnancy complications. Studies have shown that although face-to-face mindfulness courses for pregnancy have demonstrated good levels of adherence to the course with completion rates of over 85% [39], adherence after online courses tended to be more difficult. The reasons for low levels of motivation might include rare contact with the treatment team/researchers or a lack of social support from the other participants [30].

As our pilot study design is not an RCT, the changes in the mean state anxiety levels after the 1-week course of mindfulness cannot be clearly attributed to the eMBI. Urech et al [35] assessed the efficacy of a web-based cognitive behavioral stress management training in women with preterm labor and found no significant differences in the psychological parameters between the intervention and the control group (based on distraction). They hypothesized that the psychological well-being improves automatically during the course of pregnancy since participants in both groups showed lower stress-related psychological levels [35]. However, another study suggested that differences in the mean change of the STAI scores between the intervention and control groups were not solely due to feelings and symptoms changing as gestation progresses [50]. Considering the length of the eMBI, most studies report an average duration of 6-8 weeks of intervention [23,39]. In comparison, our 1-week electronic mindfulness course may seem rather short. However, a course duration of 8 weeks would be unsuitable for inpatients. Newham et al already proved that even a single session of yoga reduced both the subjective and physiological measures of state anxiety (STAI-S and cortisol), and this class-induced reduction in anxiety remained at the final session of the intervention [64]. Likewise, studies examining the immediate effects of an intervention reported significantly lowered STAI scores after a single session of complementary therapy-based interventions. The mean post-intervention scores were consistently significantly lower than the baseline intervention scores after all single-session interventions [50]. These facts underscore our results and demonstrate that short-term interventions can be beneficial, especially for those who have the greatest need. However, the small sample size limits the statistical power to detect the effects as well as the generalizability. The large number of statistical tests used in this study increases the global α error; thus, random effects cannot be ruled out sufficiently.

Peripartum anxiety and depression are highly relevant health issues in hospitalized pregnant patients. Despite high prevalence rates and the patients’ need for regular mental health support, only a very small proportion receives adequate treatment. In this pilot study, we could show that our 1-week course of mindfulness may be an effective brief intervention that could have the potential to positively influence childbirth anxiety and pregnancy-related stress among high-risk patients. eMBIs may provide an easily accessible and effective means to support pregnant women in managing anxiety and stress during their inpatient stay and beyond. RCTs with confirmatory analyses in larger samples are needed to confirm the effectiveness of mindfulness in promoting perinatal mental health and to study the potential for harm in mindfulness-based programs.