Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0603
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-07-31T06:09:01.779615Z
status: published
lens: g_space
sector: health
headline: Imminent opioid overdose risk prediction using classical and machine learning survival models following first recorded opioid-related diagnosis: a prospective cohort study from the <i>All of Us</i> research program
dek: Background: Identifying patients at risk of opioid overdose in healthcare settings is critical, yet evidence on predictive models and their performance to predict imminent opioid overdose remains limited. Objective: We compared classical and Machine Learning (ML) survival models to predict 30-day overdose risk following a first opioid-related diagnosis to determine whether algorithmic complexity improves clinical decision support. Methods: We conducted a prospective cohort study using longitudinal Electronic Hea…
gain_title: Survival models including machine learning approaches predicted imminent 30-day opioid overdose following a first opioid-related diagnosis with C-indices up to 0.745, identifying prior overdose and opioid misuse as strongest predictors for proactive clinical intervention.
problem_title: (none)
trace_subject: (none)
gain_reading: Survival models including machine learning approaches predicted imminent 30-day opioid overdose following a first opioid-related diagnosis with C-indices up to 0.745, identifying prior overdose and opioid misuse as strongest predictors for proactive clinical intervention.
gain_evidence: Test-set C-indices ranged from 0.708 to 0.745, with Weibull AFT and Survival SVM performing best | Patients with prior overdose or opioid misuse diagnosis represent a high-yield target for proactive clinical intervention
problem_reading: (none)
problem_evidence: (none)
quick_read: Using longitudinal EHR data from the All of Us Research Program, a prospective cohort study tracked adults after a first opioid-related diagnosis to predict overdose within 30 days. Classical and machine learning survival models were tested, with 560 overdoses observed among 14,737 individuals and best test-set C-indices between 0.708 and 0.745.

The findings suggest simple interpretable models may be sufficient for imminent risk stratification since added algorithmic complexity did not meaningfully improve prediction, while highlighting prior overdose and misuse as high-yield intervention targets. Uncertainty remains about generalizability beyond All of Us EHR data and performance in high-risk strata where calibration details were less emphasized.
limitation: 
tag: Evidence-backed gain
key_points: Prospective cohort study used longitudinal EHR data from All of Us Research Program v8 covering May 2018-October 2023. | Among 14,737 adults with first recorded opioid-related diagnosis, 560 incident overdoses occurred within 30 days. | Compared Cox proportional hazards, Weibull AFT, and four ML survival models: Elastic-Net Cox, Random Survival Forest, Gradient-Boosted Survival Trees, Survival SVM. | Prior overdose had adjusted hazard ratio 5.77 and opioid misuse 3.72, consistently driving risk across models.
rundown: Researchers followed 14,737 adults from the All of Us Research Program after their first recorded opioid-related diagnosis for 30 days, observing 560 overdoses for a cumulative incidence of 3.8%. They compared classical Cox and Weibull AFT models against four ML survival models.

Prior overdose (aHR 5.77, 95% CI 4.64-7.17) and opioid misuse (aHR 3.72, 95% CI 2.94-4.70) were the strongest predictors across all models. Test-set C-indices ranged from 0.708 to 0.745, with calibration acceptable and predicted risks closely aligned with observed risks in low and moderate strata.
sources:
- peer_reviewed | The American Journal of Drug and Alcohol Abuse | https://doi.org/10.1080/00952990.2026.2697750 | 2026-07-30
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

056ae17939d0e3d631a2d6b4313fcaf6a7b7eea24dfc6d41291a5dede8b801f4

previous

0000000000000000000000000000000000000000000000000000000000000000