Monsoon fever-admission forecasts for hospital beds, staff and platelets
TL;DR Feverward tells a hospital chain or blood centre how many fever admissions to expect in each of its cities over the next two to six weeks of monsoon, and what that means for beds, nurses and platelet units. Hospitals pay a yearly licence per site. It works because fever season turns up every year but its size doesn't.
India recorded 289,235 dengue cases in 2023 and 233,519 in 2024[1], and national surveillance logged 774 dengue outbreaks between 2021 and 2025, clustered in a few states and districts[2]. Physicians describe monsoon as the season when dengue, scrub typhus and leptospirosis all peak together[3]. For a hospital that means a few weeks when fever wards overflow.
Dengue drives blood demand too. In a Kerala tertiary hospital, 92% of transfused dengue patients needed platelets[4], and platelets keep for only days. Most hospitals plan by last year's calendar and react once the emergency room fills up, which is a bit late in my opinion.
What it does, in regulatory terms: operational demand forecasting for hospital planning. It makes no prediction about any patient, so it is not medical device software. It uses aggregate admission counts, not patient records.
Each week Feverward takes the hospital's own fever admission counts, local rainfall and temperature, public outbreak reports and seasonal climate signals. It returns a forecast band per site for the next two to six weeks, translated into beds, nursing shifts and platelet units. After the season it scores itself against what actually happened.
The seed pitched AI that predicts outbreaks for public health systems. I reframed it because national surveillance belongs to the government and city-level vector-control forecasting already has players. Hospitals face the same wave from the other side, and they buy faster than municipalities and can measure the value in empty beds and wasted platelets.
Users are operations heads, nursing superintendents and blood-centre managers. Payers are multi-site private hospital chains and large standalone blood centres. I found no public figure for what Indian hospitals spend on demand forecasting. Price a yearly licence per site with a cheaper first-season pilot. Blood centres are a second buyer, because expired platelets are a direct, countable loss.
Climate research shows real lead time. A 2026 Kerala model found climate patterns give about three months of dengue predictability and projects incidence rising 16–20% by 2040[5]. In Odisha, rainfall tracks cases with a one-month lag[6]. Fever seasons aren't getting any smaller.
There's no public figure for this market. The narrow signal is a fever season that sends over 2 lakh confirmed dengue cases a year through the system in four of the last five years[1], concentrated in states with dense private hospital networks[2].
One Canadian outbreak-intelligence firm sells outbreak forecasts to governments and to pharma and life-science firms planning where products will be needed[7]. India's national platform collects outbreak reports for public health, not hospital planning[8]. City-focused dengue forecasting serves municipal spray crews. Hospital chains use bed-management software that tracks today's occupancy but doesn't forecast seasonal waves. The gap is a hospital-side seasonal forecast.
| Difficulty | Moderate (42/100) |
| Starting capital | ₹45–85 L |
| Time to launch | 6–9 months |
The modelling is modest. The real work is getting clean multi-year admission data from each hospital, which often sits in messy billing systems, and earning trust for a full season before the forecast changes rosters or blood-collection drives. Each new site needs its own back-test.
No medical device licence is needed, since forecasts concern hospital demand and not any patient's care.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't use these forecasts for any decision about an individual patient. And a site with thin history should get a wide band, not false precision.
Want to work out the AI layer for Feverward? Talk to Yash →
I'd call it a fair bet if you have hospital operations or health-analytics experience and access to two or three chains. If you're hoping to sell to governments then it's not for you. First, back-test five seasons for one chain and show the forecast beats last year's calendar by a clear margin on peak timing and size.
Plan for roughly ₹45–85 L, mostly on data work for each hospital, modelling and a full-season pilot.
No medical device licence, since it predicts demand, not patient outcomes. You need DPDP compliance and should take only aggregate counts.
Yes, wherever seasonal fevers or flu fill hospital wards. Check local data protection law and the reuse terms of public health data.
The free Feverward validation kit has the line-by-line capital breakdown, the effort scoring, the data and privacy checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for hospital operations heads, and 5 assumptions to test. Download the kit →
Thinking about building Feverward or something close to it? The make-or-break work is getting clean admission history and proving one season's forecast, and you only get one monsoon a year to do it. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.