ASHASetu

About the Project Inspiration

Healthcare does not always fail because treatment is unavailable. Sometimes it fails because the referral itself gets lost.

When a primary-care provider refers a patient to a specialist, there can be little visibility into what happens afterward—whether the appointment was booked, whether the patient reached the facility, whether the consultation happened, or whether the report returned to the referring provider. Research cited in our work shows substantial referral drop-off and communication gaps across care settings.

This problem is particularly important in India, where rural patients may face long travel distances, specialist shortages, fragmented referral pathways, and connectivity limitations.

These challenges inspired us to propose ASHASetu, a predictive care-continuity infrastructure designed around the health workers who can actively follow up with patients.

Our Idea

ASHASetu aims to predict which referrals are likely to fail before they fail and recommend an appropriate intervention.

At the time a referral is created, the proposed system would evaluate explainable operational signals such as:

Distance to the referred facility Referral urgency Historical completion rate for that specialty at the facility

The system would then generate a risk score and explain the major contributing factors. For example:

87% High Risk 55 km distance + urgent referral + low neurology completion rate

Rather than simply marking the referral as overdue, ASHASetu would propose an action such as recommending a closer facility, triggering ASHA follow-up, or escalating an overdue referral.

The concept also includes a Facility Reliability Index, which would use completed referral data to identify facility- and specialty-level bottlenecks such as poor completion rates and long delays.

Proposed Technical Approach

Our proposed architecture consists of:

Frontend: React / Next.js Backend: Node.js / Express Database: PostgreSQL with Drizzle ORM Scheduling: Redis Low-connectivity channel: SMS / IVR

The proposed workflow is:

$$ \text{Referral Created} \rightarrow \text{Risk Scoring} \rightarrow \text{Explain Risk} \rightarrow \text{Recommend Intervention} \rightarrow \text{Follow-up} \rightarrow \text{Outcome} $$

For the initial concept, we propose an explainable weighted scoring model rather than black-box diagnostic AI. The system is intended to predict operational referral failure, not diagnose disease or recommend medical treatment.

What We Learned

While developing the concept, we learned that healthcare technology should be designed around the actual workflow and constraints of the healthcare system.

A patient-facing app may not be the best solution when smartphone access or digital literacy is limited. This led us toward an ASHA/health-worker-centered approach, where the person responsible for follow-up receives a prioritized worklist and actionable recommendations.

We also learned that transparency is important in healthcare. Instead of simply producing a percentage, the proposed system explains why a referral is considered high-risk.

Challenges

The main challenges we identified are:

Data availability: Real referral-outcome data would be required to properly validate and calibrate the risk model.

Adoption: Health workers may already rely on paper, verbal, or informal referral processes, creating potential adoption friction.

Connectivity: Rural deployment cannot assume reliable broadband access, motivating the proposed SMS/IVR fallback.

Validation: Any future predictive model would need to be evaluated against real-world referral outcomes before being used operationally.

Future Vision

ASHASetu is currently a proposed solution for the Hack2Heal 2.0 idea round, not a deployed or validated system.

Our vision is to eventually develop and pilot the concept with PHC/CHC networks, collect real referral outcomes, validate the risk-scoring methodology, and integrate real notification and healthcare systems.

ASHASetu aims to turn a referral from a handoff into a continuously monitored care journey—predicting where it may fail and enabling intervention before the patient falls through the gap.