Chronic kidney disease · Payor AI · India
Kidney failure is visible in claims data years before anyone reads it.
vrucare scores the claims, pharmacy and diagnosis records Indian health insurers already hold, ranks members by risk of progressing to dialysis, and reaches the ones who matter through an AI voice agent in six Indian languages.
Typical eGFR decline in diabetic nephropathy, ~4–6 mL/min/year untreated. Staging per KDIGO. The gap between the two markers is the entire commercial opportunity: CKD cost rises roughly 7.5× from Stage 2 to Stage 4, and around 50× once a member starts dialysis.
The gap
The most expensive avoidable claim on an Indian payor's book
Indian insurers already hold the signal. A pre-policy creatinine, an annual health check, a discharge summary listing three antihypertensives and metformin — the footprint of failing kidneys accumulates in payor systems for years. Nobody reads it, because nothing has been built to read Indian claims data.
How it works
Rank the book, call the member, close the loop
Risk stratification on claims-only data
A gradient-boosted model ranks every covered member by probability of progressing to dialysis, with SHAP attribution giving the medical team a readable reason for each flag. Works without lab feeds — built for the data Indian payors actually have.
Voice outreach in six languages
An AI voice agent calls flagged members in English, Hindi, Tamil, Telugu, Kannada or Malayalam — explains the risk, and drives them to a creatinine test and a nephrology referral.
Conversion back to the payor
Who was called, who tested, who saw a nephrologist, and what moved. The payor sees intervention rates against the risk tiers, so the programme can be measured against claims cost rather than taken on faith.
Status
Where the product actually is
Built and demonstrable
End to end, today
- CKD risk model with SHAP explainability per feature
- Two-tier scoring: claims-only, or lab-enriched
- Case manager portal with risk tiers and call triggers
- Voice agent prototype across six Indian languages
- Full demo: claims file in, risk cohorts out, call placed
What the raise funds
Next twelve months
- Validation on real de-identified payor data
- Multi-horizon survival model with calibration gates
- DPDP-compliant pipeline, audit logging, access control
- Integration adapters for Indian TPA claim formats
- First two paid payor pilots
Business model
Priced per covered life, the unit payors already budget
| Line | Figure |
|---|---|
| Pricing, per member per month | ₹5–15 |
| Contract shape | Annual, per covered life |
| Infrastructure cost at scale, monthly | < $8,000 |
| Gross margin at scale | > 85% |
| Recurring revenue at 1,00,000 covered lives | ≈ ₹1.2 Cr ARR |
Revenue scales with lives under management rather than claims processed, which puts vrucare on the same side as the payor and the member: every prevented progression is worth more to the insurer than the fee, and the member keeps their kidneys longer. India is the launch market — TAM across all insured lives is approximately ₹5,600 Cr, of which commercially insured lives represent roughly ₹3,200 Cr.
Why now
The rails only just arrived
- Claims data is becoming machine-readable. NHCX under ABDM is standardising claims exchange between hospitals, insurers and TPAs on FHIR — the data stops being PDFs in a TPA folder.
- Repricing is constrained. IRDAI has closed the old levers — no exit at renewal, pre-existing cover mandated after waiting periods, a 60-month moratorium. Cost of care is what is left.
- Prevention is explicitly permitted. IRDAI's wellness and preventive guidelines let insurers fund exactly this kind of programme, including premium incentives for members who engage.
- Indian-language voice became affordable. The full stack runs for under $8,000 a month, which is what makes ₹8 per member per month a viable price rather than a rounding error.
Investors
The deck, the model, and a live walkthrough are available on request.
We are raising to fund validation on real payor data and the first two paid pilots. Happy to walk through the risk model, the voice stack, or the unit economics in detail.