Is integrated patient experience intelligence finally possible in the NHS? Part 1: Fragmentation

Summary

Is integrated patient experience intelligence finally possible in the NHS? Or will infrastructure without design leave us with expensive technology and limited value?

For years, patient feedback has existed in frustrating fragmentation. Complaints sit in one system, surveys in another, Friends and Family Test (FFT) scores somewhere else entirely. All sit in a different system. None connected. No one recognises the patterns until harm has occurred.

The infrastructure pieces needed to solve this are forming. The NHS Federated Data Platform (FDP) is operational. The NHS App is expanding. Patient Reported Experience Measures (PROMS) are being designed and validated. Large language models (LLMs) are proving capable of analysing unstructured feedback at scale.

However, having these pieces available does not deliver integrated patient experience intelligence. That requires deliberate strategic design: decisions about what capabilities to build, data architecture purpose-fit for those outcomes, governance frameworks enabling appropriate uses, sophisticated analytical infrastructure, organisational change and sustained commitment. Ben Kenyon examines it all in this three-part series:

• Part 1 examines where we are today: the fragmentation challenge, why feedback sources don't connect and the safety signals we're losing as a result.

• Part 2 explores the infrastructure pieces now coming together: the FDP, NHS App and LLMs, and why deployment alone won't deliver integration.

• Part 3 shows what becomes possible if designed properly, and what actually needs to be true beyond the technology: strategic intent, governance, analytical capability and the organisational change that infrastructure cannot substitute for.

Content

Fragmentation

NHS trusts currently collect patient feedback through, at my last count, around seven separate channels: formal complaints, FFT, national surveys, Patient Advice and Liaison Service (PALS), local surveys, online reviews and PROMs. Each exists in different systems, managed by different teams, on different timescales. When a patient experiences the same failure across multiple touchpoints, they tell the NHS many times and the NHS hears it zero.

My work, applying LLMs to complaint data from the past two years, makes the analytical cost of this concrete. Complaint narratives typically contain three to five distinct issues per complaint. Manual coding processes capture, on average, 1.25*. We are systematically discarding roughly two-thirds of the intelligence sitting in our own records.

When one patient tells the same story four times

An example is Mrs Smith, who has been discharged from hospital after heart treatment. Over the following weeks, her experience generates four feedback signals:

• Day 3: FFT text message. Mrs Smith gives a score on the care she received as 2/10. "Discharge chaotic. No medication information."

• Day 8: Formal complaint. "Nobody explained how to take my new heart medication or side effects to watch for. Discharge letter didn't reach my GP for two weeks. Called ward three times, kept getting transferred." Categorised: Communications, Admissions Discharge.

• Day 12: Readmitted with medication-related complications.

• Week 6: Selected for National Inpatient Survey. Answers negatively to every question in the discharge and medication section. Data arrives at the Trust 10 months later.

What the Trust sees: one complaint, one poor FFT score in the monthly aggregate, one survey response benchmarked a year later, one readmission flagged in a clinical dashboard. What it cannot see: the same systemic failure described four times by the same patient. The complaints team doesn't know she flagged this in FFT. The clinical team reviewing her readmission doesn't know she complained about medication instructions. Nobody knows this is the third Ward 4B patient this month with an identical pattern.

Seven separate worlds

This fragmentation is structural, not accidental. Complaints teams process statutory returns under regulatory requirements. Patient experience teams manage FFT through third-party platforms. Quality teams receive national survey data from external contractors on their own schedule. Local teams run pulse surveys. Patients leave feedback through online reviews that may or may not get picked up. Clinical teams collect PROMs. PALS teams can be entirely separate from complaints teams and report nothing anywhere.

Multiple channels. Different systems. Different governance. Different teams. Rarely connecting, and nobody whose job it is to connect them.

Why integration is hard

Three structural barriers compound each other. Statutory status differs: complaints are mandated by  the Parliamentary and Health Service Ombudsman (PHSO) oversight, FFT is contracted but not statutory, surveys are centrally commissioned, local feedback is discretionary. Timescales are entirely misaligned: FFT is near real-time, complaints land weeks post-discharge, some surveys are annual. Ownership is fragmented, with separate departments and no single accountability for integrated intelligence.

There's growing recognition that this cannot continue. The infrastructure needed to move beyond it is forming. In part 2, I will examine what it could enable and why deployment alone won't deliver it.

*1.25 manual coding figure derived from analysis of NHS England KO41a complaints data (113,780 complaints, 279 trusts). 3-5 themes per complaint from my own LLM analysis across NHS organisations.

Other blogs in the series:

• Is integrated patient experience intelligence finally possible in the NHS? Part 2: the infrastructure blueprint

• Is integrated patient experience intelligence finally possible in the NHS? Part 3: what integration unlocks

About the author

Ben Kenyon has spent around 20 years working on healthcare's harder problems with data, analytics and, more recently, AI. He began on the NHS Graduate Informatics Training Scheme and went on to lead the Business Intelligence function at Manchester University NHS Foundation Trust, the UK's largest hospital trust, before moving into consultancy and health technology.

During his time at Quantium Health, he led the team responsible for taking Quail from concept through to commercialisation and deployment across the NHS. Quail became the first third-party product to go live on the NHS Federated Data Platform and one of the earliest production applications of large language models to unstructured NHS patient feedback—work cited in the NHS 10 Year Health Plan and recognised with an HSJ Award.

His interest lies in the design decisions that turn healthcare data into meaningful intelligence, and, ultimately, intelligence into better decisions and outcomes for patients.