Automation and Mental Health: Algorithms for Patient-Therapist Matching

How to cite

Chugunova, Marina; Chatterjee, Chirantan; Wang, Lucy Xiaolu; Ghosh, Mainak; Singhal, Abhay. 2022. "Automation and Mental Health: Algorithms for Patient-Therapist Matching." AEA RCT Registry. AEARCTR-0009202. https://doi.org/10.1257/rct.9202-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Marina Chugunova* , Chirantan Chatterjee , Lucy Xiaolu Wang , Mainak Ghosh , Abhay Singhal

* Primary investigator

Registry ID
AEARCTR-0009202
DOI
https://doi.org/10.1257/rct.9202-2.0
First registered
Last updated
Status
Completed
Trial dates
2022-02-01 – 2022-03-08
Countries
India ()
Registry record
http://www.socialscienceregistry.org/trials/9202

Abstract

In this project, we study the role of a human as opposed to an algorithmic mediator in compliance decisions of the users. We work in partnership with one of the largest mental health apps in India. The app enables matching between users in need of mental health support and therapists who are able to provide it. When users register in the app they are offered to undergo a psychological assessment that would improve the suggested therapist match. In this experiment, we vary if the prompt to undergo an assessment comes from a human mediator or from the system. A priori it is not clear if making the involvement of algorithm salient would increase or decrease the take up of the assessment in the context of mental health. Given the shortage of psychologists and high demand for mental health support, technologies such as mental health mobile apps are a powerful force in delivering mental health support to those in need. The results of our study are therefore of high applied importance and through a better match with a therapist can improve the welfare of many.

Experimental design

The study is conducted in partnership with a large mental health app in India. From the start of the trial all the new users are randomized either into a Human or in an Algorithm treatment arms. The treatments differ subtly in wording used to offer the prompt to undergo the assessment. The treatment either displays the involvement of a human or an algorithm in offering the prompt. As a main outcome variable we consider if the user took the assessment. Taking the assessment was a desirable action that would increase the quality of the match but it did not exclude users from communicating with a mediator and ultimately receiving the match with the therapist.

Design details

Randomization method
By a system upon registration of a user
Randomization unit
Individual user
Clusters
at least 600 users
Observations
According to our power calculations, the sample size of 600 people (300 per treatment arm) will allow us to detect an effect size of d=0.3 with alpha of 0 .05 and power of 0.95. The final number of observations depends on the industry partner.
Arms
at least 300 per treatment arm
Primary outcomes
user’s decision to undergo the assessment after the prompt
Secondary outcomes
attrition rates; time needed to complete the assessment; user's decision to open the report after completing the assessment

Cited by

No citation data available for this registration yet.

Keywords