Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1629
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-16T22:03:23.702846Z
status: published
lens: p_space
sector: health
headline: Predicting remission following repetitive transcranial magnetic stimulation in treatment-resistant depression: a comparison of parsimonious and ensemble machine learning models
dek: Repetitive transcranial magnetic stimulation (rTMS) is resource-intensive, and prognostic models may estimate remission probability after treatment in treatment-resistant depression (TRD). Whether algorithmic complexity improves remission prediction over a parsimonious clinical model remains unclear. In a retrospective cohort of 263 adults treated with rTMS who met a baseline data-availability criterion, we compared a simple logistic regression model using age, sex, and baseline MADRS with tuned multivariable re…
gain_title: (none)
problem_title: Richer feature sets and ensemble methods gave no significant gain in remission discrimination, and discrimination was only moderate, with ROC-AUCs of roughly 0.69 to 0.71.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: Richer feature sets and ensemble methods gave no significant gain in remission discrimination, and discrimination was only moderate, with ROC-AUCs of roughly 0.69 to 0.71.
problem_evidence: richer and ensemble strategies did not demonstrate improved remission discrimination | mean calibrated ROC-AUCs ranged from 0.685 to 0.711
quick_read: Researchers retrospectively analysed 263 adults treated with repetitive transcranial magnetic stimulation for treatment-resistant depression. They compared a basic logistic regression using age, sex, and baseline depression severity with tuned regression models and SuperLearner ensembles built on increasingly detailed feature sets. Twenty-seven percent of participants remitted.

The results suggest that predicting who will remit after this resource-intensive treatment may not require complex algorithms, since a three-variable model performed on par with richer ones. Discrimination was only moderate, though, and the authors say the findings do not show equivalence or that the models would work in other settings.
limitation: The cohort was retrospective and only internally validated, and the authors state that the results do not establish equivalence or transportability to other settings.
tag: Evidence-backed problem
key_points: Of 263 adults, 71 (27.0%) reached remission, defined as a follow-up MADRS of 10 or below. | Across 12 configurations, mean calibrated ROC-AUCs ranged from 0.685 to 0.711. | In eight paired comparisons against the simple model, ROC-AUC differences ranged from -0.008 to 0.018, and none was significant after BH correction. | Models predicted MADRS response less well than they predicted remission.
rundown: The simple reference model used only age, sex, and baseline MADRS and reached a ROC-AUC of 0.707. A MADRS-only benchmark scored 0.708. Feature sets for the more complex models ranged from routine clinical variables to questionnaire scores, symptom dimensions, and individual items.

Evaluation used 10 repeats of stratified 10-fold cross-validation. Random-forest imputation gave similar remission discrimination.

In a decision-curve analysis at a 25% threshold, the simple model and a SuperLearner model using symptom dimensions yielded 6.0 and 7.1 net true-positive decisions per 100 compared with treating everyone.
sources:
- peer_reviewed | Journal of Affective Disorders | https://doi.org/10.1016/j.jad.2026.122602 | 2026-10-08
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

9251d185ada083653a80fdb96cb0e67a2466b3602c3ccea327cda5bc5daf2983

previous

0000000000000000000000000000000000000000000000000000000000000000