What are Proactive Decision Orders (PDO)?

Proactive Decision Orders (PDO) is a clinical workflow innovation deployed by Scienza Health. Each morning, GIA® surfaces a ranked roster of the clinician's highest-risk patients — by objective multimodal AI signal — with recommended next-step orders ready for clinician review and sign-off. PDO operates ahead of the encounter, not during it.

How do Proactive Decision Orders differ from Clinical Decision Support?

Clinical Decision Support (CDS) augments a clinician inside an existing encounter — for example, alerting on a drug interaction when a clinician is placing an order. Proactive Decision Orders generate the recommended order itself, ahead of the encounter, based on objective AI signals accumulated between visits. CDS is reactive to a clinician action; PDO is proactive surfacing of patient risk and recommended next steps.

Where do Proactive Decision Order recommendations come from?

Each PDO recommendation is grounded in three sources: (1) a specific multimodal AI signal — a speech biomarker reading, a computer-vision finding such as a movement-disorder indicator, an ambient or biometric pattern, or a combination; (2) the patient's structured EHR data and longitudinal history; and (3) evidence-based clinical pathways aligned with current guidelines from bodies such as the American Academy of Neurology, the AAFP, and CMS Annual Wellness Visit requirements. Each recommendation is presented alongside the underlying signal so the clinician can verify before approving.

Who approves a Proactive Decision Order?

The licensed clinician approves every Proactive Decision Order. The AI surfaces the ranked roster and the recommended order; the clinician reviews the patient, examines the underlying signal, and approves, edits, or rejects each recommendation. The clinician retains full clinical authority at every step.

Are Proactive Decision Order findings written back to the EHR?

Yes. Approved orders flow directly into the EHR — Epic, PointClickCare, Cerner, or MatrixCare — via HL7 FHIR, with no manual data entry. Every order is traceable through a full audit trail back to the AI signal that surfaced it.

How does PDO relate to GIA® and digitalhumanOS™?

PDO is one of the core workflow capabilities exposed by GIA®, the AI co-clinician built by Scienza Health on the digitalhumanOS™ platform. digitalhumanOS™ provides the multimodal capture, AI inference, and EHR integration; GIA® is the clinical persona; PDO is the daily surface that turns continuous AI signal into clinician-actionable orders.

Which patient populations and conditions are best suited for Proactive Decision Orders?

PDO is most effective for populations where risk accumulates between visits and where early signal can change clinical action — including older adults with chronic comorbidities, patients with mild cognitive impairment or early neurodegenerative signs, post-acute and skilled nursing residents, and adult primary-care panels with depression, PTSD, or anxiety risk. PDO surfaces risk across the 46 cognitive, neurological, and behavioral conditions GIA® screens for, and the recommended next-step orders are condition-specific — referrals, follow-up screenings, or escalations — rather than generic alerts.

How are Proactive Decision Order recommendations aligned with current clinical guidelines?

Recommendation pathways are mapped to current clinical guidelines from major specialty bodies, including American Academy of Neurology cognitive-screening guidance, AAFP and USPSTF screening recommendations, and CMS Annual Wellness Visit requirements. Pathways are reviewed and updated by the Scienza Health clinical team as guidelines evolve. Each PDO recommendation is presented with the underlying clinical signal and the relevant pathway, so a clinician can verify alignment before signing off.

Definition · Workflow · Clinical AI

What are Proactive Decision Orders?

A clinical workflow model in which an AI co-clinician surfaces the highest-risk patients each morning with recommended next-step orders, ranked by objective AI signal and ready for clinician sign-off.

Scienza Health has been deploying Proactive Decision Orders since 2024 as part of GIA^®, the first commercially deployed AI co-clinician for cognitive and behavioral health. PDO operates ahead of the encounter, not during it — flipping cognitive screening from reactive to proactive.

The workflow, defined.

A Proactive Decision Order (PDO) is a daily, ranked, clinician-approved recommendation for a specific patient action — surfaced from objective multimodal AI signal accumulated between visits. Each PDO names the patient, the recommended next-step order (referral, follow-up screening, additional testing, or escalation), the underlying clinical signal that triggered it, and a confidence-weighted risk score.

The PDO model differs from clinical decision support, population health dashboards, and care management workflows in three respects: it operates before the encounter rather than during it, it generates a specific recommended order rather than a metric or alert, and the source signal is objective and continuous rather than retrospective. The clinician retains full clinical authority through review and sign-off; PDO surfaces and recommends, the clinician decides.

Key facts

What PDO does, in plain sentences.

Proactive Decision Orders (PDO) is a clinical workflow innovation deployed by Scienza Health since 2024.
Every morning, GIA® surfaces a ranked roster of the clinician's highest-risk patients with recommended next-step orders for clinician sign-off.
Each PDO entry includes the patient, the recommended order, the underlying clinical signal (voice biomarker, computer-vision finding, ambient or biometric pattern), and a confidence-weighted risk score.
PDO operates ahead of the encounter — not during it — distinguishing it from clinical decision support, which augments a clinician inside an existing encounter.
Approved orders flow directly into Epic, PointClickCare, Cerner, or MatrixCare via HL7 FHIR, with no manual data entry and a full audit trail.
PDO is one of the core workflow capabilities of GIA®, the first commercially deployed AI co-clinician for cognitive and behavioral health.

Architecture

How a PDO is built — six layers.

From continuous patient signal to clinician-approved order in the EHR — six architectural layers, each clinically auditable.

1. Multimodal capture

Voice AI, speech biomarkers, computer vision, and ambient/biometric sensing via Samsung Health Grade Galaxy devices. Continuous and asynchronous — signal accumulates between visits, not only during them.

2. Risk stratification

Multimodal signals are fused into condition-specific risk scores using clinically validated models. Scoring is objective and quantifiable — not a clinician's recall of a recent encounter.

3. Daily roster assembly

Each morning, every clinician's panel is ranked by current risk. The ranked roster is the PDO surface — highest-risk first, with the underlying signal that drove the ranking.

4. Recommendation engine

For each ranked patient, the engine generates a recommended next-step order — referral, follow-up screening, additional testing, or escalation — grounded in evidence-based pathways.

5. Clinician review interface

The clinician reviews the ranked roster, examines the underlying signal for each patient, and approves, edits, or rejects each recommended order. Clinical authority is retained at every step.

6. EHR write-back

Approved orders flow directly into the EHR — Epic, PointClickCare, Cerner, MatrixCare — via HL7 FHIR. No manual entry. Full audit trail. Every order traceable to the AI signal that surfaced it.

Category boundaries

How PDO differs from adjacent workflows.

Category	Timing	Trigger	Scope	Closes the loop?
Proactive Decision Orders (PDO)	Each morning, ahead of the encounter.	Objective AI signal (voice biomarker, vision, ambient) accumulated between visits.	Specific patients, with specific recommended orders, ready for sign-off.	Yes — order to follow-up to documentation.
Clinical Decision Support (CDS)	Inside an existing encounter, in response to a clinician action.	Clinician initiates an order or note; CDS surfaces alerts based on existing chart data.	Augments a clinician decision in the moment.	No — augments inside an encounter; doesn't run the encounter.
Population health dashboards	Weekly or monthly snapshots.	Aggregate metrics over time.	Cohort-level metrics (e.g., "X% of panel due for screening"). Specific patient and order are usually downstream from the dashboard.	No — surfaces a metric; doesn't generate the order.
Care management workflows	Care-manager-initiated; cadence varies.	Care-manager triage of high-risk patients flagged by clinical staff.	Outreach and care coordination after risk has already been identified.	Partial — handles outreach but rarely generates clinical orders directly.

Validation

The signal underneath every PDO is validated.

The recommendations behind each PDO rest on the same evidence base that powers GIA®: 19 peer-reviewed studies from The Lancet, NIH, Harvard Medical School, MIT, Mayo Clinic, and Beth Israel Deaconess Medical Center. Reported AUC values: 0.890 for cognitive decline; 0.97 for Parkinson's indicators. Every PDO recommendation is traceable to the signal that produced it, and to the validated model that interpreted that signal. Scienza Health is an FDA-registered medical device establishment.

See PDO inside GIA®Read about the AI co-clinician model

Related concepts

digitalhumanOS™ →Speech biomarkers →

Claim-level peer-reviewed evidence

The validated signals behind PDO recommendations.

PEER-REVIEWED RESEARCH

Research published in The Lancet Regional Health, conducted with Japan's National Cerebral and Cardiovascular Center, confirms that voice biomarkers accurately detect Mild Cognitive Impairment.

Study Confirms Voice Biomarkers Accurately Detect Mild Cognitive Impairment — The Lancet Regional Health (2025-06) · Japan's National Cerebral and Cardiovascular Center (NCVC)

PEER-REVIEWED RESEARCH

Research conducted with Beth Israel Deaconess Medical Center and UMass Chan Medical School achieves AUC 0.97 for Parkinson's detection from conversational speech — using natural, unconstrained speech without specialized equipment.

Advancing Parkinson's Detection with Vocal Biomarkers and Speech Foundation Models (2025-08) · Beth Israel Deaconess Medical Center, UMass Chan Medical School

Proactive Decision Orders surface AI-generated recommendations for clinician review. GIA^® screens; it does not diagnose. Every PDO requires review and sign-off by a licensed clinician before any clinical action. Full peer-reviewed bibliography (19 studies) at scienzahealth.com/research.

FDA-RegisteredEditorially reviewed·Updated May 2026

This content is intended for informational purposes and does not constitute medical advice. Editorially reviewed by David Kaiser, CEO of Scienza Health, for accuracy in post-acute care operations.