Four artificial intelligence modules built into the CARE EMR platform — reducing documentation burden, empowering nurses, and supporting clinical decisions across 200+ ICUs in India.
CARE is the open-source Electronic Medical Record platform co-developed by the eGov Foundation and Open Healthcare Network (OHC), deployed across 10BedICU sites in 10 Indian states. Recognised as the 50th Digital Public Good, CARE is built on FHIR R5 standards and powered by state-of-the-art Multimodal LLMs and Sovereign Indic LLMs across the entire documentation and clinical intelligence stack.
The four AI modules below form the intelligence layer of the platform — reducing clinician workload, supporting nursing staff during critical overnight hours, and powering protocol-grounded diagnostic assistance.
Doctors and nurses consistently report spending too much time typing at a computer and not enough time with their patients. In a busy ICU, documentation competes directly with care delivery. CARE Scribe was built to eliminate this trade-off entirely.
Rather than manually typing into EMR form fields, clinicians simply speak naturally — either dictating notes solo or conversing normally with a patient during a consultation. Scribe listens to the dictation or conversation, intelligently extracts the clinically relevant data, and automatically fills the structured fields in the CARE EMR form.
Critically, the AI does not just transcribe free text — it maps the spoken content to the correct structured data types required by each EMR field, including SNOMED-CT codes for diagnoses and clinical findings, numeric values for vitals, investigation types, and medication details.
Once Scribe has processed the audio and populated the fields, the clinician is presented with the auto-filled form for review. They can accept, edit, or correct any field before saving — keeping the clinician firmly in control while eliminating the bulk of manual data entry.
CARE Scribe understands all major Indian languages. A doctor can consult with a patient entirely in Hindi, Tamil, Telugu, Kannada, Bengali — or any other major Indian language — and Scribe will follow the conversation. Critically, while the conversation happens in the local language, Scribe transcribes and fills the EMR fields in English, ensuring the clinical record remains in the standard medical documentation language, correctly structured, and interoperable across facilities nationwide.
In many 10BedICU spoke hospitals, there is no doctor physically present during early morning or late-night hours. A nurse managing a critically ill patient needs to act quickly and correctly — but may not have specialist guidance immediately at hand. The Nurse Assistant fills this gap.
The module works by ingesting the Standard ICU Protocol documents — the 20+ evidence-based treatment pathway flowcharts used across all 10BedICU sites, including drug dosing information, escalation criteria, monitoring parameters, and step-by-step clinical management pathways. These flowcharts are converted to structured text and fed into the AI as its primary knowledge base.
The Standard ICU Protocols were developed jointly by the eGovernments Foundation and the ISCCM (Indian Society of Critical Care Medicine) — India's foremost professional body for critical care specialists. These protocols are distributed to every 10BedICU hospital both as large wall posters displayed in the ICU for quick reference, and compiled into a printed protocol book for staff training and deeper reference.
A nurse can ask any clinical question in natural language — the AI identifies the relevant protocol from the 20+ defined pathways and returns the precise guidance. For example: "What is the correct dopamine dosing for a patient in septic shock?" The answer comes directly from the validated protocol, not from generic AI medical knowledge.
CARE Nurse Assistant supports all major Indian languages. A nurse in Tamil Nadu can ask a question in Tamil; a nurse in Manipur can ask in Meitei; a nurse in Rajasthan can ask in Hindi. The assistant understands the question and responds in the same language — making for a more natural and inclusive interaction. Critically, the clinical integrity of the protocol-based answer is fully maintained regardless of which language the conversation happens in.
Take all patient details in the CARE EMR → generate a complete, accurate discharge summary. The simplicity is the strength. In a busy ICU, clinicians should not need to manually compile what is already documented. The AI does it in seconds.
The CARE Discharge Summary module reads the patient's complete CARE EMR record and automatically generates an accurate, structured discharge document — no re-entry of data required, no manual compilation. It synthesises consultations, daily round entries, lab results, investigation reports, medication records, procedure notes, and nursing observations into a coherent clinical narrative.
Beyond discharge summaries, the module can generate other clinical summaries from the same EMR record — transfer summaries when a patient moves between facilities, interim summaries for review meetings, or condensed records for specialist referrals. This makes it a broader documentation utility within the CARE platform, not just a single-use discharge tool.
In critical care, a high-quality discharge summary is essential for care continuity — particularly when patients transfer from a 10BedICU spoke hospital to a higher-level facility. AI-generated summaries eliminate the documentation backlog that often delays discharge and contributes to bed unavailability.
The CARE CDSS is a new project currently being designed and developed. It represents the next frontier in CARE's AI capabilities — moving beyond documentation and Q&A support into active clinical diagnosis assistance grounded in the 10BedICU Standard ICU Protocols.
The CDSS is being built around the 20 Standard ICU Protocols that govern clinical management across all 10BedICU sites. Rather than attempting to cover all of medicine, the CDSS begins within these rigorously validated pathways — making it both clinically safe and immediately actionable.
The core function is differential diagnosis support via probability scoring. Given a set of clinical inputs, the system calculates a probability score indicating which of the 20 protocol-defined diagnoses the patient's presentation most likely corresponds to — giving the clinician a ranked list of probable conditions with confidence scores.
Patient symptoms · vital signs (HR, BP, SpO₂, temperature, RR) · laboratory test results (FBC, metabolic panels, ABG, cultures) · and potentially imaging findings in future iterations. The richer the inputs, the more precise the probability scoring.
The CARE CDSS features were showcased at the United Nations General Assembly 2024 — recognition of the global significance of deploying AI-powered clinical decision support in public health systems serving underserved populations at scale.
CARE AI is powered by state-of-the-art Multimodal LLMs and Sovereign Indic LLMs across the entire documentation and clinical intelligence stack — from multilingual speech understanding in Scribe, to protocol Q&A in the Nurse Assistant, document generation in Discharge Summary, and probability scoring in the CDSS. The use of Sovereign Indic LLMs ensures deep, accurate support for all major Indian languages, keeping clinical intelligence grounded in India's linguistic and healthcare context. The CARE AI capabilities were featured at the United Nations General Assembly 2024 — recognition of the global significance of deploying AI-powered clinical decision support within public health systems serving underserved populations at scale.
CARE is open-source. If you're a developer, clinician, or healthcare innovator, there are many ways to contribute to the platform and help extend these AI capabilities to more hospitals.