Health, Care, Trust and AI
- Mar 25
- 9 min read
Updated: Mar 26
Why the most undervalued metric in health systems is the hidden precondition for every technology you're about to deploy
In my last post, I wrote about Austria's road safety transformation. Between 2006 and 2024, road fatalities fell by 52% — even as the number of licensed drivers kept rising. The key was not technology. It was sequence. Austria redesigned the system before utilisation increased. Before deploying more, it asked a question that most systems never bother with: what is the environment into which we are deploying this?
Today, the overwhelming majority of health systems are asking how fast they can deploy AI.
They are not asking whether the people they serve still trust the system enough for any of it to matter.
THE MISSING METRIC
Health Systems Have a Performance Checklist. Trust Is Not On It.
We measure everything in healthcare. Bed occupancy. Waiting times. Readmission rates. Mortality ratios. Patient satisfaction. We have built dashboards of extraordinary sophistication to track what our systems produce.
We do not measure trust.
We measure satisfaction instead — and we have confused the two for decades. Satisfaction tells you how someone felt after an encounter. Trust tells you whether they will come back at all. Whether they will adhere to the medical advice they were given. Whether they will engage early enough before things become a crisis. Satisfaction is a rearview mirror. Trust is the road ahead.
1 in 4
Global Confidence
People who believe their health system works well with only minor changes needed
16 / 19
EU Trend
OECD countries where satisfaction with the healthcare system has declined since 2021
−31%
US Trust Collapse
Fall in trust in physicians and hospitals, April 2020 to January 2024 — from 71.5% to 40.1%
JAMA Network Open · Perlis et al. · 443,455 respondents · 2024
These are not satisfaction scores. They are measurements of the substrate into which every digital health tool, every AI algorithm, every telehealth platform is now being deployed. That substrate — the accumulated willingness of citizens to rely on their health systems — is thinning.
Into this, we are deploying AI.
EUROPEAN CASE STUDY
What Trust Erosion Actually Looks Like
Greece is not a story of governance failure. It is a story of what happens to a health system when a decade of fiscal compression — imposed by external constraint, not domestic choice — erodes the structural conditions that make people willing to rely on public services.
The headline indicators are, on the surface, reassuring. Life expectancy sits above the EU average. Universal coverage exists in law. The public system is staffed and operating. By the metrics most health systems report, Greece functions.
Now, follow the signs.
Greece is not chosen here to be singled out or held up as a cautionary tale. It is chosen because its trust indicators are unusually well documented — and because what they reveal is not exceptional. It is directional. The same pressures that have compressed the trust environment in Greece — fiscal constraint, workforce strain, widening gaps between what the system promises and what people experience — are present, in varying degrees, across most European health systems. Greece is further along the curve. But the curve is a shared one.
Indicator | Greece | EU / OECD Avg | Source |
Satisfied with healthcare availability | 27% | 64% | |
Unmet medical needs (2024) | 12.1% among highest in OECD | 3.4% | |
Unmet needs — lowest income quintile | 18.7% | — | |
Unmet needs — highest income quintile | 2.9% | — | |
Out-of-pocket spending (% of total health expenditure) | 34% | 16% | |
Public share of health expenditure | 61% 2nd lowest in EU | 80% | |
Chronic patients who trusted their last provider | 57% lowest in OECD | 78% | |
Unmet dental care needs — at risk of poverty | 61.2% | 13.6% |
Two in three Greeks are not satisfied with the availability of healthcare. The poorest Greeks report unmet medical needs at a rate more than six times higher than the wealthiest. Out-of-pocket spending is more than double the EU average — not because Greeks prefer private care, but because they have made a rational calculation, at household level, that the public system may not be there when they need it.
And at the furthest edge of that calculation: 61% of Greeks at risk of poverty report unmet dental needs. They have not merely disengaged from one service. They have quietly, and entirely rationally, stopped expecting.
⚠ The Trust Inequality
The income gap in unmet needs — 18.7% in the lowest quintile, 2.9% in the highest — is the largest recorded anywhere in the EU. This is the reflection of a trust inequality in its deepest sense. Those with the fewest resources have the least ability to compensate when the system fails them.
Greece is not unique in having arrived at this point. It is simply further along a trajectory that other European health systems — under their own workforce pressures, underfunding cycles, and post-pandemic credibility deficits — can already recognise in their own data.
A Single Striking Fact
Greece invested in health ICT at a 132% higher rate in 2023 than in 2015. Thirty active digital health transformation projects are currently under way. The technology investment is real. It is accelerating
And it is landing into precisely the trust environment described above.
THE SPEED MISMATCH
AI Is Entering the Space Where System Trust Has Already Receded
Trust in a healthcare system is built slowly — through repeated, reliable experience. It is lost quickly — when those expectations are not met. Technology moves on a different clock entirely. Investment cycles, procurement cycles, deployment timelines — measured in months, not years. These timelines are not aligned. Yet they are now overlapping.
Read in context
With very few exceptions, the US healthcare system is provider, payor, and pharma-centric by design. Trust, in that architecture, was never the primary design objective. It is, in many respects, the last thing to fix — because the system was not built around it in the first place.
European health systems are a different story entirely. They are not broken. They are stretched. The trust deficit visible in countries like Greece is not a structural design flaw — it is the accumulated consequence of fiscal compression, workforce strain, and decades of underinvestment in primary care and proper prevention. The purpose of these systems — universal, equitable, publicly funded care — remains intact. What has eroded is confidence that they can still deliver on that purpose.
This distinction matters enormously. Deploying AI into a broken system is one problem. Deploying AI into a stretched system whose citizens have started to lose faith is a different problem — and a more recoverable one. But only if trust is treated as the precondition it actually is, and not as an afterthought to be addressed once the technology is live.
You cannot borrow trust from a system that does not have it to lend.
THE ARCHITECTURE OF TRUST
AI Cannot Substitute for the Trust It Requires
Trust in healthcare is not one relationship. It never was. It operates on three distinct layers simultaneously — and all three must hold.
Patient → Clinician
Does the patient believe the clinician is acting in their interest?
The most visible layer. Measured, studied, and still declining in most European systems.
Clinician → System / Data
Does the clinician believe the tools they are given are safe, purpose-driven, and explainable?
The most overlooked layer. The hinge on which everything else turns. Rarely measured. Quietly eroding.
Citizen → Institution
Does the public believe the health system exists to serve them?
The foundational layer. The one that determines whether either of the other two can hold.
AI touches all three layers at once. Most digital health deployments address none of them explicitly. They are evaluated on efficiency, on clinical outcomes, on cost per interaction. Almost never on whether they deposit into the trust account or withdraw from it. Every deployment does one or the other. Most executives deploying them do not know which.
The second layer is the one most consistently overlooked. We speak of patient trust. We rarely speak of clinician trust — and yet it is the hinge on which everything else turns.
A clinician who does not trust the system they work within cannot be a credible advocate for that system to their patients. A clinician who feels excluded from decisions about the tools they are expected to use will not adopt those tools with conviction — and without conviction, adoption is performative at best and actively harmful at worst. In stretched European health systems, where workforce morale is under sustained pressure and the gap between what clinicians were trained to do and what the system now allows them to do has widened considerably, this layer of trust has quietly eroded alongside patient trust. Largely unremarked. Never measured.
⚠ This is not a soft problem. It is an operational one.
AI tools deployed without the genuine buy-in of the clinical workforce will be worked around, selectively used, or silently abandoned. The technology will be live on paper. The trust required to make it effective will not be there.
WHERE TO BEGIN
Rebuilding Trust in the Health System
Frameworks do not build trust. Experiences do. The work of rebuilding trust is not technical. It is structural, sequential, and — in systems that are stretched rather than broken — achievable. But only if it is treated as the foundation.
1 START WITH FRICTION, NOT 'INNOVATION'
Every system knows where trust breaks: the delayed appointment, the lost referral, the unexpected cost, the consultation reduced to minutes. These are not edge cases. They are the lived experience of the system. Treat them as a trust emergency. Measure them. Elevate them. Make them visible at board level alongside the metrics that already live there.
2 EQUIP THE WORKFORCE
Clinicians cannot restore trust they do not have the tools to build. A clinician who cannot explain a digital tool to a patient, who feels unsupported by the systems around them, who was never involved in designing what they are expected to use — that clinician cannot be a credible advocate for the system to the patient in front of them. The evidence is clear: physicians are more inclined to trust and implement new approaches when they perceive them as explainable and when they have been involved in shaping them. Co-design enables health systems to embed the specific needs and values of clinicians early in the process, while pre-emptively identifying barriers to adoption. Workforce education in digital health is not a training event. It is a structural intervention. And in most European systems, it has not been built.
3 GIVE PATIENT AND CLINICIANS DECISION RIGHTS, NOT ADVISORY ROLES
Advisory panels are consulted. Co-designers make decisions. The difference between the two is the difference between a system that listens and a system that changes. Patient and clinical representation in procurement, governance, and evaluation processes is not idealism. It is the operational condition for sustained adoption.
4 MEASURE TRUST EXPLICITELY
Until trust is measured — alongside waiting times, readmission rates, and bed occupancy — it will not be managed. Satisfaction scores describe what patients felt. Trust scores describe what they will do. They are not the same instrument. Start treating them differently.
5 SEQUENCE BEFORE YOU DEPLOY
Austria did not ask whether more drivers could handle faster roads. It asked whether the system around those drivers was designed to protect them when things went wrong. The sequencing shaped everything. Technology deployed into a depleted trust environment does not restore trust. It inherits the deficit and amplifies it.
AI is being pushed into health systems at investment speed. Trust — the condition patients and clinicians need in order to rely on those systems — is not being preserved or built at the same pace.
One is supply-driven. The other is demand-driven. And right now, they are moving in opposite directions.
AI does not create demand. It assumes it.
But patients do not engage with what is pushed. They rely on a system they trust. Clinicians do not adopt what is imposed. They use what they understand.
In the current context, the impact of technology on health indicators will be heavily defined by the trust environment within which it is deployed.
That environment is the work most of us haven't started yet.
SOURCES
Comments