In what has become known as P4 or precision medicine [1,2] a major goal of medical research and applied health care is the evolution from a reactive treatment approach toward medical care that is predictive, preventative, personalized, and participatory.

This approach is particularly relevant in the field of psychiatry, as the imprecise nature of psychiatric nosology, in part due to the heterogeneity of clinical populations, complicates the identification of vulnerable groups and effective treatments [3,4]. Affective disorders such as major depressive disorder exemplify this problem. Only approximately one-third of patients with moderate to severe depression respond to the first treatment attempt with medication [5,6]. This leads to a prolonged illness duration for nonresponders, which is associated with worse overall health outcomes and significantly higher costs to the health care system [7,8]. Precision psychiatry could help alleviate this problem by predicting (P1) the occurrence of depression as well as individual disease course and preventing (P2) unfavorable outcomes such as chronification and suicide by personalizing (P3) treatment plans according to individual risk and resilience factors.

Preliminary attempts have been made to achieve the prediction of disease course and treatment outcome of major depressive disorder through the use of predictive modeling approaches [9-13]. These are a first step toward the identification of biomarkers for depression, which may ultimately inform clinicians who is at risk for relapse or a particularly severe outcome and would benefit from more invasive interventions such as electroconvulsive therapy [14]. However, previous work relies on extremely homogeneous study populations that are carefully selected according to strict inclusion and exclusion criteria. Any predictive models using such data are of little value when findings are to be generalized to clinical reality—highly diverse inpatient populations [15,16]. In contrast to the aforementioned data from homogeneous, well-characterized study samples, data that are routinely gathered in clinical practice are highly heterogeneous, unvalidated, and often not standardized or are inaccessible for predictive analysis [17].

In order to achieve the long-term aims of precision medicine in major depressive disorder treatment, the implementation of a standardized data collection routine in naturalistic environments from real clinical populations is needed. While high temporal resolution data on sleep and activity levels can be tracked with smartphone- or wearable technology–based solutions [18], differentiated data on the patients’ mood or affective state—the core features of major depressive disorder diagnosis—are needed for precise and valid models of affective disorders that accurately reflect the disease and treatment course. As this information needs to be provided by the patients themselves, much emphasis needs to be put on the participatory (P4) aspect of precision medicine in psychiatry. Patients thus need to be engaged and participate actively by contributing either self-report measures at regular intervals or by participating in clinical interviews or ratings. As external assessments are time-intensive and require clinical training, the use of self-rating scales, which can be completed by patients independent from the presence of a researcher or clinician, might be preferable. Recent evidence suggests reasonably high agreement when comparing patient self-reports with diagnostic clinical interviews [19], which supports their use in clinical practice. Previous studies also found the incorporation of self-report measures of symptom severity into routine care to foster engagement between patients and health care professionals and enhance care delivery in various fields of medicine [20]. In psychiatric populations with affective disorders however, questions can be raised as to the patients’ ability and motivation to provide such data, considering the lack of energy as well as cognitive impairments that define major depressive disorder during an incapacitating episode requiring inpatient treatment [21,22]. The difficulty of recruiting depressed patients for randomized controlled trials has been well documented [23], although some investigations revealed that patients did report positive attitudes toward research participation when they felt they were contributing meaningfully to the advancement of major depressive disorder treatments [24]. It remains unclear how the collection of standardized patient reports throughout the treatment course would compare in inclusion rate to the usually much more time-intensive and elaborate study protocol of a randomized controlled trial. It remains equally unknown whether certain patient subgroups may be systematically less willing or able to provide such data regularly, either due to their symptom severity or other disease-specific or sociodemographic factors, which may constitute exclusion criteria in randomized controlled trials.

Another point to consider when striving to make health care truly participatory is that assessments should preferably be collected in a digital format, as digitization allows for quick data analysis and, ideally, feedback for patients on their personal outcomes [15,25]. In general, digital solutions outperform paper-and-pencil questionnaires in practicality, acceptability, and completeness of data across studies in different fields [26-28]. Digital data collection with tablet computers, specifically, is well accepted among psychiatric patients [29]. However, previous investigations with patient reports in psychiatry only included a single assessment or pre–post comparisons as opposed to tracking individual symptom levels throughout the duration of their hospitalization [27]. The feasibility and acceptability of such a study protocol in psychiatric populations remains therefore hitherto unclear.

We established a system of longitudinal digital data collection that gives patients the opportunity to participate actively in providing data concerning their mood and symptom levels throughout the course of their inpatient treatment for an affective disorder via tablet computers. This study assessed whether affective disorder inpatients are willing and able to participate in repeated digital data entry throughout the treatment course. We additionally examined whether age, gender, symptom severity, and global functioning systematically co-vary with the feasibility and acceptability of such research efforts in clinical populations with affective disorders. We furthermore aimed to validate self-report measures of depression severity with the use of an external assessment performed by a clinical interviewer.

With this study, we demonstrated the feasibility and acceptance of repeated digitized assessment of risk and symptom profiles in affective disorder inpatients. Our results indicate that participatory medicine can be achieved in patients with affective disorders, as they are willing and able to contribute self-report measures throughout the duration of their inpatient treatment. This study, therefore, provides important insight into the possibility of routinely collecting longitudinal data in real-world clinical cohorts that may guide the way toward personalized psychiatric care.

During an assessment period of 1 year, we achieved an inclusion rate of 66.5% of patients with a diagnosed affective disorder. This rate is similar to those reported from other investigations performed on the general population and nonpsychiatric patient groups, which indicates that major depressive disorder or bipolar disorder symptomology does not constitute a barrier toward participation [37,38]. Adherence to assessments was high, and the dropout rate after initial participation was very low. Exclusion by the clinician or researcher and refusal to participate by the patients themselves were largely not due to symptom severity or cognitive impairment but for organizational reasons or a general disinterest in the study, which mirrors reasons for nonparticipation in research from nonpsychiatric populations [39]. It also confirms previous research on the generally high level of acceptance of longitudinal self-reporting technology in patients with bipolar disorder [40,41].

A vast majority of patients who did participate were able to enter data independently and did not encounter technical difficulties. More importantly, we found no association between symptom severity at baseline and the amount of required support in handling the equipment or prolonged data entry times. This is in line with previous investigations on the feasibility and acceptability of digitally based assessments in psychiatric populations [29]. Moreover, we were able to demonstrate the validity of patient reports with the use of an external measure of depression severity performed by a clinical interviewer. The level of agreement between self-reported depression severity with the BDI and the external rating based on the HAMD was comparable to findings from the literature [42] and indicated high validity of the digital self-report measure. This is an especially promising result for the implementation of patient-reported data collection technologies into routine documentation as well as its use for research purposes because it indicates that little to no additional personnel resources are required in order to gain valuable longitudinal data throughout the course of treatment. Moreover, the digital implementation of such assessments will allow for more accurate data collection and immediate data storage and analysis [15]. In the future, such an infrastructure of digital data collection could be used to communicate treatment outcomes and visual representations thereof directly to patients. Similar approaches have been found to improve communication between patients and health care providers [43] and would also constitute an improvement in the participatory aspect of precision medicine.

Although our results generally support the feasibility of longitudinal digital data collection in affective disorders, a few systematic recruitment and accessibility issues must be addressed. Despite the fact that no association between symptom severity and performance during data entry in our participating sample was detected, 6% of all potential participants were excluded beforehand due to reduced cognitive ability or clinicians’ concerns over their acute mental distress. Although this embodies only a small percentage of our sample, this result suggests that a systematic exclusion of more severe cases may not be avoidable. Nevertheless, it may be worthwhile to critically consider such cases individually, as it has been shown that carers overestimate the amount of distress patients are put under during research participation [24].

We, furthermore, found a statistical trend for women to be more likely to refuse or be excluded from the study than men. Although this finding did not survive the correction for multiple comparisons, it is nevertheless meaningful to consider potential reasons this gender disparity may occur, as it may again be due to symptom severity. Although the reasons recorded at the time of exclusion do not suggest symptom severity is the main factor, evidence does suggest that women are more likely to seek psychiatric treatment and report more severe symptoms [44,45]. As we do not have data on the symptom levels of the excluded group, this question cannot be answered with certainty.

Regardless of gender, the factor that impacted both the amount of required assistance and time during data entry was older age. Older adults found it more difficult to handle the tablet-based assessments and took longer to complete them. However, the percentage of participants who needed assistance was comparatively small, and even those who did require assistance were able to complete assessments regularly, which suggests that their difficulties with handling the equipment did not stop them from participating. Although our study did not assess subjective attitudes toward technology, previous studies [38,46] found that digital methods of data collection are well-accepted even among older adults. It can also be expected that technological literacy will rise in older populations over the years, as smartphones and tablets are becoming increasingly ubiquitous, which will alleviate the difficulties for this specific age group in the future. Studies [47,48] also show that, although older adults lag behind in digital literacy, such competencies can be acquired through social support. Nevertheless, our findings suggest that options for support of older participants should currently be offered to not systematically exclude technologically less well-versed patients from participatory care.

In addition to older age, there was also a trend for lower global functioning to be related to slower data entry pace; however, it was not associated with the amount of required assistance. Even though this association was not upheld when correcting for multiple comparisons, this finding should also be critically discussed. It suggests that patients with a generally lower level of global functioning take longer to complete assessments but are still able to do so independently. Moreover, the added time expenditure does not lead to participants dropping out of the study, indicating that the slower entry pace is tolerable and not a barrier that would keep lower functioning patients from participating in such research.

Furthermore, patients who are not proficient in the language spoken by their health care providers are a systematically disadvantaged group in psychiatric care who could not be included in this investigation. At equal or greater levels of need, persons with an immigration background are known to seek mental health treatment less often and are less likely to report favorable treatment outcomes [49,50]. This suggests that the inclusion of marginalized populations would be of great importance especially when investigating individual risk factors for affective disorders on the way to precision psychiatry. In fact, digitally assessed self-report measures present the opportunity to get detailed, standardized assessments despite language barriers, as questionnaire measures can be made available in every language. The app that we used for data collection supports the implementation of multilanguage assessments [51] and could therefore be used in future investigations in order to also reach and assess non-German speaking clinical populations. This would provide a wealth of standardized, quantifiable information about patients of diverse cultural backgrounds that could guide treatment but also assist in identifying suitable interventions for clinical populations with specific ethnic or cultural differences and risk or resilience factors.

To the best of our knowledge, this study is the first naturalistic investigation to incorporate repeated, digitally assessed patient reports throughout the course of inpatient treatment in affective disorders. Overall, the acceptability and feasibility of such study protocols within the clinical routine is high while required resources remain comparatively low. Patients are willing and able to provide data at regular intervals and are not systematically disadvantaged by the severity of their affective symptoms. Future implementations should keep gender, age, and cultural factors in mind when approaching patients and offer assistance with any technological equipment as needed.

In conclusion, this study is a first step in demonstrating that the participatory aspect of precision medicine can be achieved in psychiatry. In the future, the information gathered routinely through patient reported assessments could be combined with other potential data sources such as fitness trackers and information gained from electronic health records [25,52-54]. This may pave the way for data-driven predictive models that could, in the more distant future, be used to predict and prevent the occurrence of affective disorders, as well as facilitate the identification of individual risk profiles. Even without the use of predictive modeling, self-reports and the direct exchange of such information between patients and clinicians might improve treatment [43]. Overall, such advances in psychiatry will be invaluable as personalized treatments tailored to such individual risk factors may lead to much shorter and less frequent hospitalizations, which would equate to more cost-effective treatments and a pronounced reduction in patient suffering.