▶4:02
Opioid Use Disorder (OUD) Solution scores every patient's risk before the treatment decision, then surfaces the protocol-matched next step inside the EMR. Clinicians spend less time assembling the chart, prescribe against real risk instead of instinct, and capture the ancillary services each patient is clinically eligible for.
Run the score to see which factors contributed and how much.
Results from a six-month deployment at Rocky Mountain Pain Solutions, a multi-provider chronic pain practice. The same 1,200-patient dataset became the training foundation for the OUD risk model later qualified on Mayo Clinic Platform.
cliexa collects the data a pain or behavioral health practice already trusts and scores it automatically, so a composite picture of risk exists before the patient is roomed instead of after the visit ends.
Seven instruments, the record, and the pre-visit intake resolve into one score.
An XGBoost classifier trained on real clinical data works together with cliexa's Clinical Rules Engine to return a single risk tier: Minimal, Low, Medium, or High. The rules engine holds the clinical guardrails, and the model reads the pattern across everything the record contains.
OUD risk index
A risk number a clinician cannot interrogate is a number they will override. cliexa shows which instruments, medications, and behaviors moved the score, how far each one moved it, and what changed since the last visit.
COMM rose from 6 to 11 and morphine equivalents crossed 90 per day, so the tier moved from Medium to High between visits.
Once a patient has a risk tier, the follow-up schedule stops being a judgment call made under time pressure. cliexa holds the practice's own monitoring rules and applies them to every patient at every tier, consistently.
Provider visit and urine screen every two months, set by the tier.
A therapist is attached the moment the tier reaches High.
Risk re-evaluated in three months, or sooner if a score changes.
Your protocols, encoded once, applied the same way to every patient.
The tier tells you which services a patient is clinically eligible for. cliexa surfaces those recommendations inside the chart, so the clinician no longer has to remember which patient qualifies for what. This is where a stratified population turns into delivered care.
Each recommendation traces to the tier and the documented comorbidity behind it.
Six months at Rocky Mountain Pain Solutions across 1,200 chronic pain patients, generating structured assessment, medication, toxicology, and behavioral data during active care.
A model built on that dataset predicted opioid risk with 97% accuracy compared with the risk pain and mental health providers calculated themselves.
The University of Colorado Denver's Department of Computer Science and Engineering built a deep neural network on the same clinical data, handed off to cliexa in April 2020. Results were presented to practice administrators through an MGMA-accredited CME session.
cliexa was among the first companies selected, gaining access to de-identified patient data, bias and fairness evaluation, and clinical validation infrastructure.
cliexaAI Opioid Use Disorder Solution was qualified on Mayo Clinic Platform, combining an XGBoost classifier with the Clinical Rules Engine and returning real-time, interpretable four-tier risk scores.
If your specialty is not on this page, we can apply the same predictive work to yours.
That same foundation was applied to chronic kidney disease modeling across 85,000 patient records through Mayo Clinic Platform Discovery. The work demonstrated that the underlying approach, including disease-agnostic reasoning, rigorous validation, responsible AI, and enterprise-ready deployment, could extend well beyond a single condition.
OUD risk prediction is where the architecture was proven. It is not where its potential ends.
Ask us about your specialtyDisease-agnostic reasoning, rigorous validation, responsible AI, and enterprise-ready deployment extended from opioid risk to a second condition.
OUD risk prediction is the use of a clinical model to estimate a patient's risk of developing opioid use disorder before it presents clinically. cliexa scores validated assessment instruments alongside medication history, morphine equivalents and urine toxicology already in the record, then returns a four-tier risk level inside the clinical workflow with the factors that produced it.
cliexa calculates opioid use disorder risk with an XGBoost classifier working alongside a deterministic Clinical Rules Engine. The rules engine holds the clinical guardrails while the model reads the pattern across patient-reported scores, prescribing data and toxicology results. The output is a single risk tier of Minimal, Low, Medium or High, returned in real time.
cliexa captures and scores SOAPP-14, DAST-10, AUDIT, GAD-7, PHQ-9, PDI and COMM. Those patient-reported scores are read alongside medication history, morphine equivalents per day and urine toxicology consistency. These are the instruments pain management and behavioral health practices already use, so the model works from evidence clinicians already trust.
cliexa's OUD risk model predicted opioid risk with 97% accuracy compared with the risk calculated by treating providers in pain and mental health settings. The model was trained on structured data from 1,200 chronic pain patients captured during a six-month clinical deployment, which means the training data came from real clinical care rather than synthetic or retrospective sources.
Yes. Every cliexa risk score arrives with the instruments, medications and behaviors that moved it, ranked by weight, along with the trend since the previous visit and a full audit trail. Pairing a machine learning classifier with a deterministic rules engine keeps the reasoning inspectable, and a clinician reviews and signs every decision.
Qualification on Mayo Clinic Platform means a model has been evaluated against an institutional standard for responsible, production-ready clinical AI. cliexaAI Opioid Use Disorder Solution became a qualified solution in January 2026. cliexa was also among the first companies selected for Mayo Clinic Platform_Accelerate in 2022, which provided de-identified data access, bias and fairness evaluation, and clinical validation infrastructure.
No. The purpose of risk stratification is to match treatment and monitoring to each patient's actual risk. Higher-risk patients receive closer monitoring and behavioral health support, while lower-risk patients avoid unnecessary screening burden. Opioid prescriptions fell 35% in deployment because clinicians could finally see risk clearly, and the clinician retains every prescribing decision.
The risk score is available before the treatment decision is made. Patient-reported instruments are scored the moment they are submitted, so a composite picture of risk exists before the patient is roomed rather than after the visit ends. The tier and its contributing factors surface inside the chart the clinician already works in.
OUD risk prediction is used by pain management practices, behavioral health and addiction medicine programs, primary care groups managing chronic pain populations, and health systems running opioid stewardship programs. It is also relevant to organizations in value-based or risk-bearing contracts, where identifying and documenting risk early affects both clinical outcomes and contract performance.
Bring a de-identified cohort from your pain or behavioral health panel. We will show you how the tiers fall out, what drove each score, and which patients your current workflow is not catching.