Pathology that arrives on its own

We built the HL7 ingestion as a standalone microservice rather than as a feature inside the application. It receives files securely from HealthLink, parses HL7 ORU messages, extracts the biomarker values and their units, maps them to the correct category, matches the result to a patient on name and date of birth, flags the new result to the clinical team, and logs every error and mismatch it encounters.

Keeping it separate matters. Pathology feeds are the least predictable input in the system, and isolating them means a malformed message from one provider is a logged ingestion failure rather than an application outage.

Biomarker mapping, the part nobody plans for

Rather than hard coding a fixed list of analyte codes and hoping every lab agreed with it, we built a biomarker mapping screen into the admin portal. When a result arrives under a code or description the platform does not recognise, it is surfaced to the admin team as unmapped rather than discarded. They map it once, against the canonical biomarker it belongs to, and every subsequent result from that provider under that name flows straight through into the right trend, the right graph and the right traffic light.

The practical effect is that a new referring lab, or a lab that renames a panel, does not require a code change or a developer. The clinic handles it themselves in a few minutes.

Reference data sits behind it. Peaklife supplied the minimum, maximum and median values defining below average, average and optimal for each tracked biomarker, and the platform scores every incoming value against those bands. Some biomarkers traffic light ranges are actually defined via the HL7 data itself, and in that case we take that range rather than the client’s defined range, as sometimes it may also vary by lab.

Unmatched results are surfaced, not lost

Where a result cannot be matched to a patient automatically, it goes to a dedicated review screen in the admin portal. The team sees the incoming record and assigns it to the correct patient manually. Nothing is auto assigned on a partial match, and nothing sits in a failure log that nobody reads.

OCR for everything that does not arrive as HL7

Not every result comes through the feed. Imaging reports, external records and documents a patient brings with them arrive as PDFs. We added OCR document parsing so clinical data can be read out of those documents and captured against the patient rather than retyped.

The biological age engine, their science implemented into a web system

The biological age algorithm and the scoring model behind it are Peaklife’s intellectual property. Our job was to implement it faithfully and keep it adjustable, so the formula can be refined as the clinic’s own evidence base grows without a rebuild.

It computes from multi system data across blood work, fitness and imaging and sits at the top of the patient dashboard, alongside risk scoring across metabolic, cardiovascular and inflammation domains, the biomarker traffic light, fitness and imaging summaries, clinician recommendations and the patient’s next appointment.

The commercial side runs in the platform too

Membership packages, subscription tier status and renewal dates are managed inside the platform, and the admin team can raise ad hoc invoices and track them through their status.

Referrals generated, not retyped

The referral form generator produces output ready for pathology and imaging partners, prefilled with the patient’s name, date of birth, address, Medicare number, contact details and so on. It removes a repetitive piece of reception / clinician work to help optimise business processes.