Dr. Rahul Kushwah reports

QSCREEN AI REPORTS INTERNAL VALIDATION RESULTS FOR CORRECTIONAL INTAKE PLATFORM: 100% SENSITIVITY ON SUICIDE RISK AND WITHDRAWAL DETECTION ACROSS 65 CALIBRATED CLINICAL SCENARIOS

QScreen AI Inc. has released internal validation results for its correctional intake platform across 65 calibrated clinical scenarios tested against published instrument thresholds.

By digitizing the high-liability blind spots of correctional intake, QScreen AI transforms a universal institutional risk into a scalable software-as-a-service revenue model with zero-hardware deployment friction. Prior announcements described what this platform is capable of, and today's announcement reports what it achieved.

What the comparison shows

The company ran the same 40 clinical scenarios through a paper-only intake model built on published self-disclosure rates from the correctional health literature. The gap was significant. The paper-only model identified zero of the seven postrelease overdose cases in the cohort. It identified zero of the suicidal patients who had not volunteered the information, and it identified zero impairment cases where the only signal was physiological. QScreen AI correctional intake platform identified all of them.

That gap has a dollar figure attached to it. Postrelease overdose deaths carry a documented societal cost exceeding $1-million per incident per published health economics literature. Seven identifiable cases per 40 booking cohort, each triggering a naloxone protocol and a MAT referral before the patient's release date, is an intervention with real economic value at the one moment in the justice system where it is still possible to make it. It is also a liability question. Every one of those cases represents a documented risk that a paper form cannot defend in court. QScreen AI creates the clinical record and every decision documented, every flag timestamped, every nurse override recorded with a rationale.

The platform

QScreen AI administers nine validated clinical instruments during a structured intake of under 10 minutes on a standard laptop running a standard camera without a need for specialized hardware (no need for a multispectral camera) or an information technology project. The instruments are CIWA-Ar (Sullivan 1989), COWS (Wesson and Ling 2003), Columbia C-SSRS (Posner 2011), PREA-R (Moss and Metzger 2009), MAT Readiness per SAMHSA TIP 63, postrelease overdose risk per Binswanger (NEJM 2007), a benzodiazepine withdrawal clinical rule, a 72-hour deterioration model and an artificial intelligence clinical summary. The nine reported herein are those validated in this study against published thresholds. The full 13-instrument architecture remains available for deployment.

Validation methodology and results

Scenario-based validation against published clinical thresholds is the recognized predeployment methodology for clinical decision support tools under ASME V&V 40-2018. Ground truth was set from published instrument thresholds before any scenario was run. Reporting follows ASME V&V 40-2018, STARD 2015, and Consort-pilot 2016.

Suicide risk -- Columbia C-SSRS: 100-per-cent sensitivity, 100-per-cent specificity. Every patient with active ideation identified before housing assignment. Zero false alarms across 37 non-SI patients.

Withdrawal -- CIWA-Ar and COWS: 100-per-cent sensitivity, 100-per-cent specificity. Every presentation including benzodiazepine withdrawal -- a condition with no validated scale and no systematic flag in a paper-only intake process.

PREA-R victim risk: 100-per-cent sensitivity, 100-per-cent specificity. Every high-risk placement identified before general population assignment, consistent with 28 CFR 115.41.

Camera fitness clearance: 100-per-cent sensitivity, 100-per-cent specificity. Including two methamphetamine presentations standard PERCLOS screening would have cleared.

MAT readiness: 89-per-cent sensitivity, 100-per-cent specificity.

Postrelease overdose risk: 88-per-cent sensitivity, 97-per-cent specificity. Seven Critical cases per cohort, each generating naloxone protocol and MAT referral before release.

72-hour deterioration: 48 per cent at elevated tier, 100-per-cent specificity. No high-acuity patient misrouted.

Camera physiological pilot -- 25 scenarios: 92-per-cent clearance accuracy, 100-per-cent Unfit sensitivity, 96-per-cent discrepancy detection.

Three gaps found and corrected

A validation that finds nothing is not a validation. Before finalizing results, the team identified and corrected three gaps. The most common poly-drug intake response was that multiple substances were not triggering CIWA-Ar or COWS scoring; expanding the keyword matching raised withdrawal sensitivity from 75 per cent to 100 per cent. Benzodiazepine withdrawal was receiving no clinical flag because no validated scale exists for it; a seizure-precaution rule was built and added. A calibration error on one passive suicidal ideation case was identified and corrected against published C-SSRS thresholds, which moved suicide sensitivity from 67 per cent to 100 per cent. All three corrections are documented in full and available on request.

Commercial context

Every correctional facility booking carries documented liability exposure when no structured clinical assessment is conducted. Suicide in custody is the leading cause of jail death in the United States. Withdrawal deaths are preventable and have been litigated. Postrelease overdose mortality carries a societal cost that exceeds $1-million per incident and is documented in the New England Journal of Medicine. These are not hypothetical risks, and they show up on legal invoices.

This validation ties sensitivity and specificity data to published clinical standards across every one of those risk categories. A procurement officer, health director or risk counsel reviewing it has a documented evidentiary basis for a deployment decision. The platform needs no hardware, no IT engagement and no capital budget. Deployment follows upon facility on-boarding and agreement execution.

The 60-day live pilot is structured, prespecified and ready. Two 30-day phases, parallel operation alongside standard paper intake and then independent operation with full nurse override authority throughout, converted to a 12-month software-as-a-service agreement on meeting five predefined performance criteria. The platform is built, validated and commercially structured. Facility discussions are active, and the company expects to provide an update in the near term.

Dr. Rahul Kushwah, chief operating officer of QScreen AI, stated: "This is not a capabilities update. It is a performance report with published references behind every number. We ran the same cohort through a paper-only intake model, and the detection gap was material across every clinical category. The facilities we talk to already know they have a liability problem. What this gives them is the clinical evidence to act on it, with no hardware requirement, no IT project and a 60-day pilot that converts to contracted revenue on meeting prespecified criteria."

About QScreen AI Inc.

QScreen AI is a health technology company building a proprietary artificial intelligence engine with quantum inspired computing and advanced physiological sensing to clinical and occupational health assessments across correctional facilities, addiction medicine rehabilitation and industrial work force screening in multiple jurisdictions.

We seek Safe Harbor.