Common questions about CareEZ

No. CareEZ is a triage support tool, not a diagnostic medical device. It is not cleared or registered as a medical device in Hong Kong, the EU, or the United States. Every output includes a disclaimer directing users to consult a qualified Speech-Language Pathologist (SLP) before acting on any recommendation.

CareEZ is designed to extend, not replace, professional clinical judgment. The distinction matters: a medical device undergoes regulatory review for diagnostic efficacy; CareEZ operates as a decision-support layer in the hands of carers, subject to SLP oversight.

On our baseline test set, Snap-to-IDDSI achieves 4 of 8 IDDSI levels correctly (50%) under Claude Vision. All misses are ±1 level — the model errs toward a safer, more restrictive texture rather than a dangerously permissive one. This clinical-safety bias is intentional: over-referral is always preferable to under-referral in dysphagia care.

Our Year 1 Q3 target is 75% exact-level accuracy. Independent validation results are published in the changelog as they become available.

CareEZ is engineered with a clinical-safety bias: when uncertain, the system over-refers rather than under-refers. Institutional users operating under a Tier 2+ agreement are required to follow our adverse-event reporting protocol; all incidents are reviewed by the clinical advisory team.

All outputs include explicit "consult qualified SLP" disclaimers. Pilot sites receive onboarding training covering escalation pathways. We maintain a public changelog of accuracy improvements and known limitations so users always know the current confidence envelope.

No. Voice biomarker analysis for Cantonese aspiration risk is currently research-staged, with a planned release in Year 1–2. The /api/voice-biomarker-cantonese endpoint stub is live today for workflow integration and validation testing — partners can build against its interface now and be ready when the model goes live.

Follow the changelog and research page for milestone updates. The research is being conducted in collaboration with the HKU Swallowing Research Laboratory.

No single entity owns CareEZ's data. The platform operates under a co-stewardship model: the HKU Swallowing Research Laboratory holds clinical oversight, and Full Linkage Community Care Association (a Hong Kong s.88 registered charity) holds the public-interest data mandate.

Carewells Limited operates the platform as the commercial operator under these co-governance principals. This structure ensures that clinical data cannot be monetised unilaterally by the operator. Full details are at governance.

No PHI is stored by default. All public-tier API calls are stateless — inputs are processed and discarded; no conversation history or identifiable data is retained on CareEZ servers.

The honest exception: institutional operators (Tier 2+) may enable server-side logging within their own infrastructure at their discretion, subject to their own data protection obligations under the Hong Kong Personal Data (Privacy) Ordinance. Carewells does not receive that log data unless explicitly contracted.

Carewells Limited is the operator of the CareEZ public asset — not the owner. Carewells funds operations through its commercial layer (Tier 2+ enterprise subscriptions and the SeniorDeli food-modification product line).

The public-good layer — free carer access, multilingual tools, and the 照護食 Adoption Index — is funded by grants and social-enterprise surplus, not by monetising carer data. This separation is structural, not aspirational. See governance for the legal and contractual details.

CareEZ is built on the HKCSS Care Food Standard (護食標準), not merely aligned with it. The texture classification logic maps directly to the HKCSS Seven Key Elements and the companion IDDSI framework.

CareEZ (via Carewells / SeniorDeli) is a co-drafting institution of T/SATA 094-2025 (the Greater Bay Area group standard on General Requirements for Dysphagia Foods for Elderly Care) and is named in companion standards T/SATA 084-2025 and T/SATA 085-2025. A full alignment table is at standards.

No. CareEZ extends the reach of SLPs; it does not replace them. Clinical diagnosis of dysphagia, the formal prescription of texture levels, and swallowing rehabilitation remain within the exclusive scope of qualified SLP practice.

CareEZ handles the triage and point-of-care guidance layer — the 640,000 carers making meal decisions daily without a clinician present. SLP supervision remains mandatory for institutional users, and every CareEZ output includes an explicit referral prompt when risk indicators are detected.

Apply at /trial/. Pilot slots for Year 1 are conditional on the outcome of the HKEX IFS 2026 award, which funds the institutional rollout. Applications are reviewed on a rolling basis; priority is given to RCHEs, NGO care operators, and FDH employment agencies.

You will receive an acknowledgement within five business days of submitting your application. Early applicants are added to the pre-launch cohort regardless of award timing.

A typical pilot site commitment is 8–12 weeks, covering: an onboarding session (2 hours, in your staff's language), access to the full Tier 2 feature set at no cost during the pilot, a structured feedback protocol (weekly 5-minute check-in plus exit interview), and optional participation in the 照護食 Adoption Index baseline assessment.

Pilot sites are named in our published research outputs (with consent). Full details — including the governance and data-sharing terms — are at trial and governance.