Custom health software projects often begin as a point solution — a portal, a worklist, a tariff engine — and end as a integration programme with no owner. The clinic wanted a faster calendar; the estate inherited six contracts, four message formats and a patient index reconciled every Sunday night.
Point solutions multiply edges
Every new module introduces edges: authentication, identifiers, coding systems, error queues. Edges are where budget goes after the demo. A platform strategy collapses edges inside a single longitudinal record while exposing only the boundaries regulators and partners require. That is not rip-and-replace dogma; it is arithmetic.
UK providers connecting to national services need interoperable exchange — typically FHIR and HL7 where mandated — without making external schemas the system of record. Internal models should stay stable; facades should translate, not fork patient truth.
Module map as architecture artifact
Draw the module map before writing user stories: appointments, encounters, orders, results, medications, charges. If two boxes maintain separate patient keys, you have already chosen middleware as a lifestyle. Blackflow engagements that succeed treat the map as contract: one timeline, bounded services, explicit anti-corruption layers at the perimeter.
Promed HIS ships that map as product — visits, Promed EHR, Prolab LIS, pharmacy and billing on one tenant — so buyers inherit engineering decisions clinics used to fund twice. Readers comparing platforms: medical office software; UK clinic depth: medical practice management software uk.
Integration tax line items
- Identity reconciliation after each vendor upgrade.
- Transform maintenance when a partner changes segment usage.
- Dual training surfaces for clinical and operational staff.
- Audit gaps when logs span incompatible products.
- Release coordination across vendors with conflicting roadmaps.
Total cost of ownership spreadsheets rarely include edge tax explicitly — which is why diaries look cheap until finance measures rework.
Platform acceptance tests
Acceptance is an engineering proof, not a screenshot. Spin a sandbox tenant; create a patient once; book; document; order lab and imaging; ingest validated results; post charges — same identifier throughout. Measure events end-to-end with tracing, not operator testimony.
Release trains should not require a integration war room. When lab, imaging and prescribing share a tenant, regression scope is bounded — a property product owners feel as faster fixes and fewer weekend bridges.
When custom still makes sense
Bespoke UI for a specialty workflow is valid; bespoke patient indexes are not. Custom layers should sit on stable platform primitives — encounters, orders, observations — rather than redefining them per project. Blackflow deliveries that respect that line ship faster and audit cleaner.
Buyer workshops should produce a module map and an edge count. If the edge count exceeds the module count, you are not buying software — you are buying a permanent integration programme with a clinical logo.
Handover to operations
Procurement should score vendors on edge count, not feature slides. Two products with identical checklists on paper diverge wildly once identity, orders and results must align under load. Architecture reviews belong in the buying process — before contracts encode the wrong shape.
Runbooks should document event flows, not only servers. On-call engineers need a single trace across modules when a result stalls — not a contact list of three vendors blaming message formats. Platforms that ship with unified observability lower mean-time-to-recover and keep clinics seeing patients instead of bridges.
SLOs for result propagation should be defined in minutes, with alerts when queues age beyond clinical tolerance — the same discipline SaaS vendors use for payments, applied to observations that change prescribing. Capacity planning should model peak Monday mornings, not average Tuesday afternoons, when sizing workers and databases.
Custom software should reduce edges, not celebrate them. Medical office platforms that need a systems integrator per module are renting complexity — the most expensive subscription in healthcare IT.
