REAL-WORLD LONGITUDINAL EVALUATION
22,382
Patients represented in the source dataset.
A multi-site evaluation examined repeated HEAL opioid education assessments from 2019 to 2026. Findings describe associations over time; analytic samples varied by outcome.
KEY FINDINGS
Three signals observed across longitudinal follow-up.
These are modeled trajectories from the technical report. They are not simple raw before-and-after comparisons and do not establish causation.
01
OPIOID RISK TOOL
5.54.4
Modeled ORT scores declined over time.
The mixed-effects ORT analysis included 1,478 participants and showed a downward trajectory across follow-up.
Time coefficient: B = -0.0252 per month; p < .001.02
COMPLIANCE & AGREEMENT
12-15%2-5%
Flagged compliance responses decreased.
Modeled probabilities declined across unadjusted and adjusted analyses, indicating fewer flagged responses over time.
Approximate modeled probabilities; analytic samples varied by model.03
NALOXONE PREPAREDNESS
35-40%25-30%
Preparedness gaps decreased over time.
Modeled probabilities of flagged naloxone preparedness responses declined consistently across analyses.
Lower flagged-response probability indicates improved preparedness.Chart endpoints are approximate model-implied values presented in the study materials. See the technical report for full model specifications, confidence intervals, and outcome-specific sample sizes.
WHAT WAS EVALUATED
Risk, adherence, knowledge, behavior, and preparedness.
- Opioid Risk Tool and related screening scores
- Knowledge and education responses
- Compliance and treatment-agreement responses
- Misuse and unsafe-behavior indicators
- Naloxone awareness, access, and readiness
METHODS AT A GLANCE
Built for the complexity of routine clinical data.
The full technical report contains the complete cohort construction, model specifications, tables, figures, and sensitivity analyses.
Real-world source data
Data came from routine HEAL OpEd use across multiple clinical sites and time periods, not a controlled trial.
Longitudinal structure
Repeated records were linked with a hierarchical matching approach because the source files lacked one universal patient identifier.
Mixed-effects models
The analysis used longitudinal models designed for repeated observations, irregular follow-up, and incomplete responses.
Sensitivity analyses
Key results were retested after excluding lower-confidence matches and applying alternative follow-up restrictions.
INTERPRET WITH CARE
What this analysis can - and cannot - show.
The results support a careful discussion of longitudinal associations. They should not be presented as guaranteed outcomes or proof that the program alone produced the changes.
- 01
Observational analysis without a control group; it cannot establish that HEAL alone caused the observed changes.
- 02
The number of participants or records analyzed varied by outcome and model.
- 03
Patient matching was reconstructed because source files did not contain one universal patient identifier.
- 04
Follow-up timing was irregular, questionnaires were sometimes incomplete, and some measures were self-reported.
SOURCE MATERIALS
Read the complete analysis.
The technical report is the primary source. The four-slide file is a concise visual summary of selected modeled findings.
44PAGES
PRIMARY SOURCE · VERSION 2 · 2026
Real-World Longitudinal Evaluation of HEAL Clinical’s Opioid Education Program
Methods, cohort construction, statistical models, tables, figures, findings, sensitivity analyses, and limitations.
Access full report Short access form required
04SLIDES
VISUAL SUMMARY · 2026
Longitudinal Study Summary
A four-slide overview of selected modeled trajectories for ORT scores, compliance responses, and naloxone preparedness.
Research status: These materials are presented as an independent technical analysis and study summary. HEAL does not describe them here as peer-reviewed or as evidence of NIH endorsement.
EVIDENCE STANDARD
Association, not a guarantee.
The evaluation observed changes over time in several opioid-related risk, compliance, and preparedness measures. Individual results may vary, and the analysis does not establish that HEAL alone caused the observed changes.
Access the technical report