From consultation to bill: designing atomic clinical workflows

The most reassuring features in clinical software are the ones nobody notices: the prescription that's always there when the bill is, the bed that can't be double-booked, the total that never drifts by a rupee. None of that is luck — it's atomicity, applied deliberately.

Completing a consultation should do everything, or nothing

In MedOps, a doctor starts a consultation from an appointment, records clinical notes and a follow-up date, and marks it complete. On completion, two things must happen: a prescription is generated and a consultation-fee bill is created. The important word is “and.”

If those were two independent writes, a failure between them would leave a prescription with no bill, or a bill with no prescription — exactly the kind of inconsistency that erodes trust in a record. MedOps wraps them in a single database transaction, so completing a consultation either produces both artefacts or rolls back cleanly and produces neither.

“Both or neither” is a small phrase that prevents a large class of support tickets.

Admitting a patient is a race you have to win

Inpatient care has its own atomicity problem: two operators trying to admit different patients to the same vacant bed at the same moment. Without protection, both succeed and you have a conflict on the ward.

MedOps treats admission as a transactional operation with explicit guards:

• The bed's vacancy is checked and claimed inside the same transaction that creates the admission.
• A duplicate-admission guard prevents admitting a patient who is already admitted.
• Discharge and bed transfer follow the same discipline, so the bed map always reflects reality.

Money lives in paise

Floating-point currency is a classic source of drift: add enough 0.1 + 0.2 and your totals stop matching. MedOps stores every monetary amount as an integer number of paise, and only formats to rupees at the edges for display. A ₹773.00 bill is 77,300 paise in the database — exact, addable, and immune to rounding surprises.

// Store and compute in the smallest unit…
const totalPaise = items.reduce((sum, i) => sum + i.amountPaise, 0);

// …format only when showing it to a human.
const display = `₹${(totalPaise / 100).toFixed(2)}`;

Status, not deletion

Bills carry a type (consultation or prescription) and a status (unpaid, paid or cancelled). Cancelling is a status change, not a delete — the record stays, the history stays, and the tenant-scoped audit log captures who changed it. Operations can update status inline from a filtered, paginated list, and nothing of value is ever quietly erased.

The principle underneath

Every example here is the same idea wearing different clothes: group the writes that must agree into one unit of work, and make money exact. Get those two right and the system feels trustworthy, because the data simply never contradicts itself.

Want to walk a real consultation-to-bill flow on seeded data? Book a demo — we'll spin one up for your roles.