Personalization strategies in digital mental health interventions: a systematic review and conceptual framework for depressive symptoms

Introduction Personalization is a much-discussed approach to improve adherence and outcomes for Digital Mental Health interventions (DMHIs). Yet, major questions remain open, such as (1) what personalization is, (2) how prevalent it is in practice, and (3) what benefits it truly has. Methods We address this gap by performing a systematic literature review identifying all empirical studies on DMHIs targeting depressive symptoms in adults from 2015 to September 2022. The search in Pubmed, SCOPUS and Psycinfo led to the inclusion of 138 articles, describing 94 distinct DMHIs provided to an overall sample of approximately 24,300 individuals. Results Our investigation results in the conceptualization of personalization as purposefully designed variation between individuals in an intervention's therapeutic elements or its structure. We propose to further differentiate personalization by what is personalized (i.e., intervention content, content order, level of guidance or communication) and the underlying mechanism [i.e., user choice, provider choice, decision rules, and machine-learning (ML) based approaches]. Applying this concept, we identified personalization in 66% of the interventions for depressive symptoms, with personalized intervention content (32% of interventions) and communication with the user (30%) being particularly popular. Personalization via decision rules (48%) and user choice (36%) were the most used mechanisms, while the utilization of ML was rare (3%). Two-thirds of personalized interventions only tailored one dimension of the intervention. Discussion We conclude that future interventions could provide even more personalized experiences and especially benefit from using ML models. Finally, empirical evidence for personalization was scarce and inconclusive, making further evidence for the benefits of personalization highly needed. Systematic Review Registration Identifier: CRD42022357408.


Review Question
Are Digital Mental Health Interventions for Depression static (= every patient is going through the same content in the same order) or adaptive/ dynamic (different content and order for patients)? If DMH Interventions are adaptive/ dynamic, how are content and ordering tailored to the individual patient?

Databases: PubMed, PsycINFO and Scopus
Search will be performed between 01.09.2022 and 31.10.2022. Only publications published in English in peer-reviewed journals and conference proceedings will be included.
Additionally, reviews identified fulfilling the inclusion criterion defined below will be used as additional source to identify relevant literature. For this, the cited literature in these reviews will be scanned for missing papers not included in this review yet. An equivalent search strategy will be used for the other databases.

Search Strategy Link
-

Participants/population
For this review, studies will be selected based on intention of the intervention and not the actual tested population. Therefore, an intervention designed for insomnia but tested with a population of depressed individuals would be excluded, while an intervention targeting depression being tested in e.g. a sample of university students will be included. The following markers will be used to determine the intended use of an intervention: 1. Language: A sentence like 'Intervention for depression' would indicate inclusion, 'Intervention for depression and anxiety' exclusion 2. Used questionnaires: The use of an anxiety questionnaire as the main outcome would indicate exclusion.

Intervention(s)/Exposure(s)
Digital Mental Health interventions are those that deliver care via digital channels such as smartphones or the internet. DMH interventions can be guided (including contact with a clinician), self-guided (no contact with a clinician) or blended (DMH intervention integrated with other forms of treatment, such as psychotherapy).
This review does focus on interventions intended for use in the direct treatment of symptoms, excluding those that are solely focussing on prevention, relapse prevention, post intervention care or passive monitoring of symptoms.
Additionally, interventions need to deliver content and/or exercises and not be solely based on digital human support.
Finally, this review focuses on interventions designed specifically for depression and depressive symptoms.
Therefore, interventions that target depression and anxiety, or any other comorbid disorder are excluded. Additionally, interventions targeting a specific sub-symptom of depression (e.g. rumination) are excluded. Finally, studies on interventions targeting a specific subtype of depression, such as prenatal depression are excluded.

Comparators:
None 21. Type of Studies: 1. Study addresses internet/smartphone based intervention 2. Intervention is specifically and exclusively targeting depressive symptoms and/or MDD, which is measured by an established diagnostic questionnaire (scores in established clinical questionnaires or interviews such as the PHQ or HAMA that indicate at least mild depressive symptoms, following standard cut off criteria) and/or MDD (self-assigned or diagnosed according to DSM-IV TR / DSM-5 /ICD-10 diagnostic criteria).
3. Intervention is not targeting solely a specific subsymptom of depression (e.g. rumination). Neither is the intervention targeting a specific subtype of depression, or designed for prevention, monitoring or post-intervention-care.
4. Empirical Study with original data. 5. Publication in a peer-reviewed journal 6. Published in 2015 or later 7. In English language 8. Intervention is designed for Adults aged between 18 and 70, and therefore not exclusively for adolescent or elderly people. 9. Full Text available 10. Protocols and Secondary Data Analyses are excluded.

Context:
As this review is interested in the interventions described in the articles selected, in a second step interventions need to be extracted from the selected articles. As different studies might investigate the same intervention or single studies compare more than one intervention, the extracted interventions will likely differ in form and number from the selected articles. The following criteria will be applied to extract interventions from the selected studies If several articles use the same intervention 1. The newest article will be used to determine all datapoints outside of the main variables, such as duration of the intervention.
2. All articles will be used to determine the variable of interests. In case of opposing information, the newer article will be used, but the discrepancy mentioned in a comment.
3. If there is more than one intervention evaluated in one article, they will be included as separate interventions for additional analysis, as long as every intervention is clearly distinct from each other and fulfills the inclusion criteria

Main Outcome(s):
Is the intervention static or adaptive? This is determined by the description of the intervention in the methods section. Variation will be coded for the following subdomains. For the coding all available papers on the intervention are used and inconsistency handled as described at 23.
In absence of information on a variable aspect in a subdomain, this domain is coded as being static. If variability/tailoring/personalisation is mentioned but not explained, authors of the study will be contacted and asked for clarification with a 4 week window to respond and one reminder after 2 weeks.

Additional Outcome(s)
It will also be investigated, whether there are any studies including direct comparisons of adaptive and non-adaptive interventions. It will be reported 1. How much of these studies exist 2. Whether the superiority of adaptive interventions was proven 3. The effect sizes.

Data Extraction (Selection and Coding)
Study Selection: The titles and abstracts of the results of the database search will be checked and studies selected as described at 22.
Duplicates will be removed. For a random subsample of 100 studies, this will also be done by a second researcher, in order to calculate interrater reliability. In case of an insufficient interrater reliability (<0.9), the second reviewer will do the whole filtering process, and disagreement between the two researchers will be discussed and decided by consensus.
In the next step, distinct interventions will be extracted from the selected articles and data extracted as described at 23. and 24. This will be done by 2 reviewers independently and disagreement solved by consent.
The method section of all papers on an intervention is used to determine the main outcomes, as described at 23. Also, linked additional material will be included as well. No additional research will be performed on characteristics of the intervention, with exemption of contacting of authors as described at 24.
The following variables will be extracted:

Name of Intervention
Year (

Risk of bias (quality) assessment
The amount of authors responding to our clarification request will be reported.

Strategy for Data Synthesis
The following results will be reported: 1. Amount of studies being variable in % (overall as well as per subdomain).
2. Type of variability in %.
3. Source of variability in %. 4. Variability by type of intervention 5. Narrative summary of most common mechanisms of introducing variability.

Analysis of subgroups or subsets
Interventions will be subdivided by type of digital mental health intervention, i.e., self-guided, guided and blended. As this type influences the potential techniques to introduce variability into the program, this appears relevant to be coded and analyzed.

Type and Method of the Review.
Systematic Review. Mental health and behavioral conditions