Healthcare operations from research to production

EstarMed did not simply digitize an existing spreadsheet. It changed how information was created in the field, how coordination understood the day in real time and how completed work became auditable billing.

Figures compare November 2025 — the first full month on the platform — with July 2026, measured directly against production records for visits, routes and patients. They are not perception survey results.

I observed the operation, defined requirements, designed the product, built the web and mobile experiences, integrated the data and real-time architecture, tested the system and delivered it to production.

Lovable and Cursor accelerated implementation and iteration; responsibility for product decisions, architecture, code adaptation, integrations, QA and production quality remained mine. Operations stakeholders supplied contractual rates, clinical constraints and business rules, then validated the system in daily use.

Healthcare professionals and drivers leave every morning to deliver vaccination, breastfeeding support and home-care services. Before EstarMed, a complete month of scheduling was assembled by hand. After vehicles departed, the state of the operation could only be reconstructed through calls, WhatsApp and end-of-day paperwork.

The challenge was not a more attractive dashboard. It was supporting a healthcare operation with little margin for error: if a visit disappears from the workflow, a patient can remain unattended.

I combined stakeholder interviews, full-route observation, spreadsheet audits and an in-product UX survey available to every role. The goal was to understand not only what people said they needed, but how information was actually lost, recreated or corrected during a working day.

“If the address is wrong, I lose half an hour driving in circles while the doctor waits in the car.” — Driver. “I don't know where each route is until someone calls me. If nobody calls, I find out the next day.” — Coordination. “One hand holds the equipment bag, the other the phone. I can't be hunting for buttons.” — Driver.

The product follows the service from monthly planning through daily assignment, field execution, live monitoring and billing. These are not isolated modules: each event created in one stage becomes evidence for the next.

All patient names, addresses and identifiable operational data shown here have been anonymized or replaced with demonstration data.

The most consequential design decision was removing simultaneous choices from the driver card. Version 1 showed “Arrived” and “Finished” together, and on real routes drivers frequently selected “Finished” without recording arrival. In version 2 the next action only appears after the previous event is recorded, so capturing arrival became part of the path rather than an act of discipline.

A second iteration replaced ordering by estimated time — which reshuffled the screen mid-route — with an explicit coordinator-set sequence. Time became supporting information rather than the structure of the screen. The best UX improvement in the project came from removing choices, not adding controls.

Full-shift observation with drivers plus an embedded in-product survey drove the changes that mattered. Complete field records — measured directly in the database — moved from 27% to 97% after the field workflow redesign.

Honest limitation: formal SUS and NPS studies have not yet been completed. Current validation is primarily behavioral and operational.

I designed and implemented the product, so interaction decisions moved directly into production components across React and React Native. Design engineering reduced translation loss between intent, interface and code.

Accessibility was built into the token layer: AA contrast, color always paired with label and icon, keyboard order and visible focus for coordination, 44×44 px field controls and domain language instead of generic software vocabulary.

EstarMed connected planning, field execution, live location and billing into one operational system. Monthly completed visits increased from 414 to 987 while the coordination team remained the same size, and complete field records improved from 27% to 97%.

Field products must be designed for sunlight, urgency, movement and unreliable connectivity — and reliable data begins with the interaction that creates it. Next: complete offline coverage across remaining field workflows, formal SUS validation with the Bogotá cohort and deeper predictive capacity analytics.