Multi Health Specialists: 1,000+ clinical hours recovered from document review

Multi Health Specialists is a London-based multi-disciplinary health practice with seven clinicians across physiotherapy, nutrition, sports medicine, and occupational health. The practice sees approximately 120 patients per week and manages a significant volume of referral correspondence and clinical documentation.

The problem

Clinicians were spending an average of four hours per day on non-clinical document work: reviewing referral letters, summarising patient histories from GP correspondence, drafting routine letters, and processing discharge summaries. For a practice of this size, that represented over 1,000 hours of clinical expertise per year applied to tasks that did not require it.

The volume of incoming referral documents varied significantly in format and quality. GP letters ranged from structured forms to unstructured prose. Summarising the relevant information for a clinician before an appointment required reading the entire document and identifying the key clinical points. That work was being done by clinicians themselves, at clinical rates, because there was no alternative.

"We had seven highly trained clinicians spending a third of their day doing work that required literacy, not clinical expertise. That is an expensive way to read letters."

The solution

The sprint built a document processing pipeline that handled inbound referral letters, GP correspondence, and clinical admin documents. The system extracted the clinically relevant information from each document, structured it into a standard format, and surfaced it to the clinician before the appointment, flagging any items requiring specific attention.

The approach used deterministic extraction for structured fields (patient name, referral date, GP details, ICD codes where present) and a bounded language model for unstructured clinical narrative. Every output was presented to the clinician as a structured summary with the source document attached. Clinicians retained full oversight and could override or expand any summary.

Routine outgoing letters, including discharge summaries, referral acknowledgements, and appointment confirmations, were templated and populated automatically from the patient record. Clinicians reviewed and approved before sending, rather than drafting from scratch.

The result

Document review time dropped from four hours per day to under one hour. Across seven clinicians over a full year, that represents over 1,000 hours returned to direct patient care. The practice has used that capacity to add 15 new patient appointments per week without increasing headcount.

The system has been operating for six months without a clinically significant error in the summary output. Two summaries in the first month required corrections, which were captured in the review process before reaching the clinician. Both corrections were used to improve the extraction logic during the post-handover support period.

"The summaries are more reliable than what we were producing ourselves, because we were rushed. The system is never rushed."

Compliance note

All processing is conducted on infrastructure that meets UK GDPR requirements for health data. No patient data is used to train external models. The pipeline uses a private, contract-governed API arrangement. Full data processing documentation was provided at handover for the practice's DSPT (Data Security and Protection Toolkit) records.