Case Study · Clinical Intelligence

More TMS candidates reached treatment

In a multi-site interventional psychiatry practice, OutcomesAI helped more of the strongest TMS candidates move from psychiatric evaluation to treatment by finding overlooked evidence already documented in the EHR.

Results

Documented clinical need became an actionable treatment queue

Before OutcomesAI, 65 of every 100 strongest TMS candidates reached treatment. With OutcomesAI, 88 of every 100 did. That is 23 more patients per 100 reaching a treatment they may otherwise have missed.

Treatment initiation improved among the strongest TMS candidates from 65 of 100 to 88 of 100, adding 23 more treatment starts per 100 candidates

The longest waits fell the most

The typical wait improved, but the major change was at the tail. Patients who used to sit longest were identified, reviewed, and moved toward treatment faster.

Wait-time results showing the largest improvements among the slowest TMS candidates

Problem

Strong candidates were hidden in ordinary documentation

Treatment-resistant depression, failed medication trials, functional decline, contraindications, patient interest, location, and follow-up timing were documented across psychiatric evaluations and EHR data. The information existed, but no team could read every evaluation and assemble it consistently at practice scale.

Objective

Find the right patients sooner, without disrupting care

The goal was to identify treatment-resistant depression patients who might qualify for TMS but had not yet surfaced, with concise evidence clinicians could verify quickly inside the existing workflow.

Goals

What success required

Improve outcomes

Increase the share of strongest candidates who reached treatment.

Sharpen decisions

Show clinicians the evidence behind each rating, not just a score.

Increase efficiency

Reduce review time while making every evaluation reviewable.

Protect safety

Keep contraindications and safety concerns visible before outreach.

Avoid disruption

Use existing EHR documentation with no new forms or clinician workflow changes.

Preserve judgment

Support professional decision-making; clinicians make the final call.

Results summary

  • More of the strongest candidates reached treatment. Before OutcomesAI, 65 of every 100 strongest candidates reached treatment. With OutcomesAI live, 88 of every 100 did—23 more patients per 100.
  • An impossible task became a daily working queue. Thousands of psychiatric evaluations analyzed automatically across 12 offices, with a ranked, filterable worklist used daily by the Chief Medical Officer, clinicians, care coordinators, and office managers.
  • The strongest candidates were easier to prioritize. Clinicians received a ranked worklist with the evidence needed to decide who warranted review first.
  • Review became faster and more transparent. What previously took several minutes across scattered EHR screens became a one-to-two minute evidence review.
  • The slowest patients moved sooner. The typical wait improved modestly, but the biggest gains came at the tail: patients who used to sit longest were identified, reviewed, and moved into treatment faster.
  • Care delivery was not disrupted. Providers documented as usual. The EHR remained the source of truth. Clinicians retained the final treatment decision.

Clinician Adoption

"Before OutcomesAI, no one could read every evaluation. Now I open my own panel and see the strongest candidates, the sentence from the note behind each rating, and when the patient is next in. I decide from there."

Clinical leader, multi-site interventional psychiatry practice

Why it matters beyond TMS

TMS is the validated proof case. The broader product is behavioral-health Clinical Intelligence: finding documented care opportunities across specialty pathways such as Spravato/esketamine, ketamine, substance-use, psychotherapy, and PTSD.

Explore Clinical Intelligence

See what’s hiding in your own notes