Validated · March 2026 · Tertiary care institution, coastal Karnataka

AI-assisted triage for frontline health workers

Kinila AI Health helps ASHA workers, ANMs, and community health volunteers assess patient symptoms accurately and flag emergencies — even in low-resource settings across Karnataka.

Start an assessment See validation results ↓
97.1%Emergency detection sensitivity
99%Safe output rate
81%Overall triage accuracy
0.81Weighted Kappa agreement
90Paired AI + clinical sessions
How it works

Structured guidance at the point of care

A two-step workflow designed for busy community health workers — no medical degree required.

Structured patient intake

Guides the health worker through patient age, gender, current medications, and known conditions before symptom entry — so nothing clinically important is missed.

Four-level triage output

Returns a clear Green / Yellow / Orange / Red triage level with plain-language guidance — monitor, refer, or escalate urgently. Designed to support, not replace, the CHW's judgement.

Emergency detection

Validated at 97.1% sensitivity across 8 critical emergency types including febrile seizure, acute MI, neonatal sepsis, pre-eclampsia, and hypertensive crisis. No emergency was missed.

Session tracking

Links each assessment to a site, institution, and session reference — making it easy to audit cases, follow up, and report to programme supervisors.

Safety-first by design

When uncertain, the tool errs toward caution — over-referring rather than under-triaging. 99% of all outputs in the validation study were rated safe or borderline-safe.

Built for CHW workflows

Designed with and for ASHA workers, ANMs, and community health volunteers — not hospital clinicians. No prior clinical training required to use the tool effectively.

Who uses it

Designed for frontline health workers

Kinila AI Health is a decision-support tool for workers providing care in communities where specialist access is limited or delayed.

Primary user

ASHA workers

Accredited Social Health Activists conducting household-level health assessments and referrals in rural and peri-urban communities.

Primary user

ANMs & CHOs

Auxiliary Nurse Midwives and Community Health Officers at sub-centre and PHC level managing a wide range of clinical presentations.

Evaluator

Health officials & NGOs

District health officers, programme managers, and NGO partners looking to support CHW decision-making with a validated AI tool.

Partner

Funders & researchers

Institutions seeking evidence-based AI tools for community health in South Asian primary care settings — with a clear validation pathway and funded evaluation in progress.

Evidence

Independently validated in Karnataka

A structured validation study conducted in March 2026 using standardised clinical case scenarios reflecting PHC-level South Indian presentations across fever, GI, respiratory, obstetric, NCD, and paediatric categories.

97.1%
Emergency sensitivity
33 of 34 RED cases correctly identified
99%
Safe output rate
73 safe · 16 borderline · 1 unsafe
81%
Overall triage accuracy
Meets pre-specified target of >80%
0.81
Weighted Kappa
Substantial agreement (target >0.60 ✓)

Safety profile — 90 paired sessions

73 safe (81.1%)
16 borderline
1
■ Safe outputs ■ Borderline (reviewed — not dangerous) ■ Unsafe (1 case — adaptation target)

Study design

20 standardised clinical vignettes · 90 paired AI and student assessments · 4–5 participant groups · Conducted at a tertiary care institution in coastal Karnataka as part of the EVAH Pathway A application.

Critical case performance

All 8 RED emergency types — febrile seizure, acute MI (known and unknown history), severe dehydration, pre-eclampsia, neonatal sepsis, severe diarrhoea, and hypertensive crisis — correctly identified across every participant group.

Error pattern is clinically safe

All errors are over-cautious (upward triage), never dangerous under-triage of emergencies. The 4 ORANGE under-triage cases are specific, addressable adaptation targets: TB screen, hypoglycaemia, PID, and diabetic foot.

→ Next phase

The funded evaluation will deploy across 39 PHCs at partner institutions in coastal Karnataka, with actual CHWs as primary users and native language support integrated.

Research partnership

Conducted in partnership with a medical institution

Tertiary care institution, coastal Karnataka

The validation study was conducted at a medical institution in coastal Karnataka — a pre-deployment evaluation using standardised clinical case scenarios reflecting primary health centre presentations in South India. The evaluation will extend to more PHCs across partner institutions in the region.

March 2026  ·  Details available on request
Behind the tool

Built by Kinila Ventures

Kinila AI Health is a product of Kinila Ventures, a partnership firm based in Udupi, Karnataka, working at the intersection of community health, life sciences, and AI.

Kinila Ventures

Partnership firm · Udupi, Karnataka

Building tools and solutions at the intersection of community health, life sciences consultation, and applied AI for underserved populations in India.

kinilaventures.com →

Kinila Life Sciences

Biotech & life sciences consultation

Expert consultation for biotech and pharmaceutical organisations navigating research, regulatory, and clinical development challenges in India and globally.

Kinila Commons

Community library initiative

A community-facing knowledge and library initiative connecting people in the Udupi region with learning resources, civic knowledge, and public information.

Ready to support your health workers?

Whether you are a community health worker ready to use the tool, a health official evaluating it for your programme, or a funder interested in the evidence — we would like to hear from you.

Start using the tool

For ASHA workers, ANMs, and community health volunteers

Get in touch

For health officials, NGOs, funders, and research partners

About Kinila Ventures

Learn more about the organisation behind this tool