Patients were leaving with conditions undetected
Across 19 Texas-based primary care locations, Sona-2 passively analyzed voice from 374 routine patient appointments — no new workflow, no surveys, no additional clinical burden — and mapped the care gap the chart wasn't showing.
- Network
- 19 Texas primary care locations
- Volume
- 374 appointments · 194 analyzed patients
- Models run
- 20+ condition-specific models
- Analysis date
- February 2026
The care gap is structural, not clinical
Like most multi-location systems, a significant share of patients with diagnosable, treatable conditions were cycling through appointments without those conditions ever being identified — not through any failure of care.
Conditions were going undiagnosed
Sleep apnea, cardiovascular risk, metabolic dysfunction and allergic disease are highly prevalent yet severely underdiagnosed in primary care.
A 15-minute visit can't screen for everything
Clinicians cannot proactively screen every patient for every applicable condition in a standard appointment. That is a structural impossibility, not a quality problem.
The health signals were already present
Every patient interaction carries acoustic information correlated with sleep disorders, cardiovascular stress and metabolic strain. Without a system to interpret it, that signal passes through unheard.
The signal was already in the voice
Sona-2 analyzes only acoustic features of the voice and uses no personally identifiable information to perform its analysis.
Integrate and activate
Amplifier integrated with the practice's existing recording infrastructure, capturing patient voice from clinical interactions already taking place — no change to workflows or patient experience.
Analyze 20+ biomarkers
Hundreds of acoustic features — pitch, rhythm, speech rate, vocal quality, breathing patterns — run through 20+ condition-specific models simultaneously, producing a probability score per condition.
Prioritize patients for outreach
Detected conditions are weighted through a clinical scoring framework into tiered recommendations — Strong, Moderate, Consider — across four service lines, delivered at the clinic level.
Four programs. One consistent signal.
Across four clinical service lines, Sona-2 identified patients carrying risk profiles that warranted evaluation — patients who had not yet been referred. Percentages are of the 194 analyzed patients.
| Service line | Strong + moderate | Strong | Moderate | Consider |
|---|---|---|---|---|
| Sleep apnea program | 35.6% | 44 (22.7%) | 25 (12.9%) | 28 (14.4%) |
| Cardiovascular testing | 35.6% | 28 (14.4%) | 41 (21.1%) | 35 (18.0%) |
| Health & nutrition coaching | 32.5% | 21 (10.8%) | 42 (21.6%) | 31 (16.0%) |
| Allergy testing | 27.3% | 11 (5.7%) | 42 (21.6%) | — |
Sleep apnea and cardiovascular risk each surfaced in over a third of analyzed patients. These are not incidental findings — they represent a systematic gap between the clinical need present in the panel and the care those patients are receiving.
A population carrying more than the chart shows
Condition-level prevalence across the 194 analyzed patients. Most findings sat within the expected range for a male primary care population. One did not.
★ Anemia was flagged above the expected U.S. adult range of 3–8%.
Anemia flagged at 1.5–4× the expected rate
Sona-2 flagged anemia indicators in 11.9% of analyzed patients, well above the U.S. adult expected range of 3–8%. This was not a targeted objective of the pilot — it surfaced organically from the voice data. Anemia is associated with chronic fatigue, cognitive impairment and reduced treatment responsiveness, degrading outcomes across every service line. The finding warrants a dedicated clinical review and may indicate a systemic screening gap in the panel.
The same risk factors, multiple unmet needs
Clinical risk doesn't distribute randomly. The factors that drive sleep apnea also compound cardiovascular risk and nutritional need — when a patient carries one undetected condition, they typically carry several.
★ 32.5% of patients carry risk profiles warranting 3 or 4 clinical programs simultaneously — conditions that reactive identification will never surface at scale.
19 clinics. A network of 74.
The pattern held across diverse clinic locations and patient populations.
| Metric | Value | What it means |
|---|---|---|
| Eligible locations | 74+ | Texas primary care sites for a network-wide rollout |
| Conditions per patient | 2.7 | Average flagged — co-occurring metabolic and behavioral burden |
| Program opportunity | 55% | Analyzed patients flagged for at least one program |
The signal is already in your appointment audio
Every patient interaction carries acoustic intelligence that maps to programs you already run. Bring your own recordings and we'll run the same analysis against your panel.
Outputs are wellness and screening signals for clinical prioritization, not diagnoses. Sona-2 analyzes acoustic features only and uses no personally identifiable information.