Digital psychiatry is an exciting field and includes efforts to enhance diagnosis and monitoring of mental illness using passive sensing (eg smartphone sensors). But the literature is rife with small N studies focusing on single diagnoses. That's a huge problem for two reasons... 1/ - ThreadSky

aidangcw.bsky.social • 29 days ago

Digital psychiatry is an exciting field and includes efforts to enhance diagnosis and monitoring of mental illness using passive sensing (eg smartphone sensors).

But the literature is rife with small N studies focusing on single diagnoses.

That's a huge problem for two reasons...

1/

Comments

tervoclemmensb.bsky.social•29 days ago

Thought I was reading just a perspective/viewpoint thread and then it end with a paper! Looks awesome--excited to read.

aidancscott.bsky.social•29 days ago

I recently conducted a lit-review, "use of real-time PREMs in adolescent inpatient psychiatry."

Finding quality studies was an incredible pain!

What's the disconnect holding back larger studies in Digital health, passive monitoring, Dx?

aidangcw.bsky.social•29 days ago

First, small N studies are unreliable. There's a good chance they model sampling error, and that seems to be the case given how variable the results are across studies. Also, passive sensing work is often exploratory and includes many predictors...many variables should have many observations.

2/

aidangcw.bsky.social•29 days ago

Second, studying individual diagnoses at a time is a dead end. That's just not the way psychopathology is put together. There's too much overlap between diagnoses to make the results of any single diagnosis very clear.

The overlap needs to be studied alongside specificity.

3/

aidangcw.bsky.social•29 days ago

Our new paper (led by @whitneyringwald.bsky.social and w @vizecolin.bsky.social) tackles both of these problems. We use large N (~550) and measure broad domains of psychopathology using the new HiTOP-PRO, alongside the EARS system for passive sensing. This first paper looks at individual diffs

4/

aidangcw.bsky.social•29 days ago

We use latent variable models to examine bivariate associations among broad domains of psychopathology and passive sensing features, as well as what each of these domains shares (p-factor) and has unique in terms of associations. Here are the main figures.

Preprint link: https://osf.io/preprints/psyarxiv/3tyjp_v1

5/

aidangcw.bsky.social•29 days ago

We find interesting patterns of associations bw domains suggesting they offer distinct profiles in passive sensing features...and we also find that when considered together there is a great deal of overlap such that unique associations are much fewer.

Check out the figs and paper for details!

6/

aidangcw.bsky.social•29 days ago

Here's the link to the preprint again: https://osf.io/preprints/psyarxiv/3tyjp_v1

Enjoy!

7/

bwnelson.bsky.social•29 days ago

Very excited to read this one!

Comments

Posting Rules

Reply