More than twenty years of working on complex health and workforce problems taught me that the hardest part is rarely finding an answer. It is working out what problem we are actually trying to solve.
Most of the problems I work on are messy. They cross professional boundaries, organisations, funding systems and jurisdictions. Different groups experience different parts of the problem, and the people most affected by it often have little control over the conditions creating it. The people with the authority to change those conditions, meanwhile, may experience the problem very differently — or barely see it at all.
In those environments, there is rarely a shortage of ideas: we should develop a framework; run a survey; create a new role; build a training programme; change the curriculum; set up a network; rite a strategy (I have, at some stage done all of these, many simultaneously). The temptation is always to start there, with the activity, and I have spent most of my career learning why that is often the wrong place to begin.
The more useful first question is what, exactly, are we trying to solve — and then, whose problem is it? Because the answer can look very different depending on whether you ask the person receiving a service, the practitioner delivering it, the organisation employing them, or the system funding and regulating it. That sounds obvious. In practice, it is surprisingly difficult. It is also where PEASI came from.
The simple intervention that wasn’t simple at all
I have worked on a lot of very complex projects. One of the more methodologically demanding was a large UK evidence synthesis examining a seemingly straightforward intervention: appointment reminders. Surely the logic was simple. People forget appointments. Send them a reminder. Attendance improves.
Except that wasn’t really the problem.
Commissioned by the UK’s National Institute for Health Research, the project — TURNUP, Targeting the Use of Reminders and Notifications for Uptake by Populations — combined a conventional systematic review with theory development and a review informed by realist principles. Rather than simply asking whether reminders worked, we were trying to understand how they worked, for whom and under what circumstances. The project ultimately drew on 466 papers covering 463 separate studies, including 31 randomised controlled trials and 11 systematic reviews, and developed a conceptual model that examined the interaction between the reminder, the patient, the healthcare setting, wider social conditions and the systems through which appointments were cancelled and rebooked.
At that stage, the existing evidence was strikingly theory-light. Much of it treated non-attendance as though it were primarily a problem of forgetfulness (or at least the fault of the service user). But people miss appointments for all sorts of reasons: transport, work, caring responsibilities, health literacy, competing priorities, difficulty contacting the service, the way the appointment was communicated, or simply because the system made changing the appointment extraordinarily difficult. The intervention, in other words, could not be understood independently of its context.
The project forced us to move beyond asking does this intervention work, towards asking what is happening here, what mechanisms are involved, and what conditions need to be present for a particular intervention to produce the outcome we want. That distinction has stayed with me ever since.
A few years later, the same problem appeared in a completely different form
Between 2014 and 2015, I evaluated a demonstration trial across eleven general practices in regional New South Wales that placed public-sector nurse practitioners and specialist nurses into general practice settings — a model generally known as co-location. The project had originally been designed around allied health professionals, but by the time it reached the ground, workforce availability meant the clinicians who actually took up the co-located roles were nurse practitioners and specialist nurses instead. That gap, between the intervention as planned and the intervention as delivered, turned out to be its own small lesson: problem definition has to include a hard look at who is actually available to do the work, not just who you imagined doing it.
Again, the activity looked straightforward – co-locate clinicians with a view to create better integrated care. But putting two professionals in the same building does not create integration.
The evaluation used realist principles and a logic model examining drivers, contexts, mechanisms and outcomes, and it quickly became apparent that the same nominal intervention behaved very differently depending on the environment around it. For co-location to work, practices had to identify the right patients. Clinicians had to trust one another. GPs had to understand and use the service. Appointment and referral systems had to function. Information had to move between organisations. Suitable rooms had to be available. Financial incentives could not actively work against the model. And somebody had to broker the relationships and keep the machinery moving.
Across the trial, four co-located clinicians delivered close to 300 consultations to more than 200 patients over eleven months. The numbers were modest; what mattered was what they revealed. The analysis operated across clinical, professional, organisational and system levels, and the project improved patient experience and professional relationships in many settings — but the wider system still contained structural barriers, including different funding and accountability arrangements across general practice and public services.
The important lesson wasn’t simply whether co-location “worked”. It was that co-location wasn’t really the intervention. It was one activity inside a much larger causal system, and that is a very different way to think about service redesign.
The recurring problem: we mistake an activity for a solution
I began seeing the same pattern everywhere. A workforce has vacancies, so we develop a recruitment campaign. A profession lacks data, so we run a survey. People are working in silos, so we create an interprofessional workshop. Staff lack career progression, so we build a career framework. A policy isn’t gaining traction, so we develop a communication strategy.
Sometimes those activities are exactly what is required. But often they aren’t — the activity has been selected before anyone has adequately defined the problem. And once an activity has acquired momentum, especially if it has funding, executive support or political visibility, something curious happens: the planning process becomes how do we successfully deliver this activity, rather than is this activity actually addressing the problem. That is the trap I became increasingly interested in.
Different people may be describing different problems
There is another complication. Complex systems rarely contain a single authoritative perspective on the problem. A workforce planner may see a supply shortage. A clinician may see an impossible workload. A manager may see vacancies. A professional association may see poor recognition. A consumer may see difficulty accessing care. Treasury may see rising expenditure. A regulator may see risk. None of those perspectives is necessarily wrong, but neither is any one of them the whole problem.
This is why I have increasingly used structured co-design as part of diagnosis rather than simply as stakeholder consultation after the important decisions have been made. The point is not to collect everyone’s preferences and average them. It is to understand how the problem looks from different positions in the system — and where the capacity and authority to change it actually sits. That has become particularly important in my recent work.
From a curriculum project to a system-change problem
A recent university project began with what could easily have been framed as an education problem: how do we introduce interprofessional education across a large health faculty? The faculty included many health disciplines, each with its own distinct professional culture, curriculum, accreditation environment and history. Some had mature approaches to interprofessional learning; others did not. There was no realistic prospect of imposing one uniform educational model and expecting it to work equally well everywhere.
So we began with diagnosis. We used structured co-design to ask what successful interprofessional education actually needed to achieve, what was preventing that from happening now, and where those barriers sat across individual, organisational and system levels. Those perspectives were then tested and extended through the evidence base.
The result was not a list of teaching activities. It was an eight-domain Theory of Change addressing professional identity, collaborative capability, curriculum architecture, assessment, faculty capability, practice education, governance and future models of care. The work explicitly treated interprofessional capability as a system issue rather than a discrete educational intervention.
And that diagnosis changed the governance of the project. If the barriers include institutional priorities, assessment, accreditation, workload, curriculum architecture and relationships with health-service partners, this cannot remain a small teaching-and-learning project operating quietly at the edge of the organisation. The analysis helped make visible where authority for change had to sit, and the project gained executive sponsorship, strategic visibility and a clearer institutional mandate.
That is an important part of problem definition that is often missed. Sometimes the most useful finding is not what needs to happen. It is who has to be able to make it happen.
The same architecture can scale
I have recently used the same underlying reasoning in work on inclusive postsecondary education for people with intellectual and developmental disability. Again, the presenting problem could have been described simply as insufficient access or insufficient awareness. Instead, co-design with people with lived experience, universities and other partners was used alongside evidence synthesis to understand the barriers operating at individual, organisational and system levels. Those insights were organised into eight interacting domains and ultimately an overarching Theory of Change and implementation pathway.
What emerged was a much more useful diagnosis. The challenge was not simply persuading people that inclusion was desirable — the system contained interacting questions of pathways, institutional capability, funding, accountability, policy and coordination. A communication campaign alone could never solve that.
From complicated methodology to a practical diagnostic
For years, I approached these problems using different combinations of programme logic, Theory of Change, realist evaluation, evidence synthesis, co-design and systems analysis. Some of that work was methodologically elaborate. TURNUP is one example. Another is a large-scale evaluation of Queensland Health’s allied health workforce redesign program that I undertook in 2013. It was a retrospective evaluation of 55 pilot projects of workforce change, involving more than 500 staff, across thirteen health disciplines. We built a method for it at the time and called it Inductive Logic Reasoning: a logic model of drivers, contexts, mechanisms, outputs and outcomes, tested against 122 project documents, 67 completed surveys and a round of interviews. Reading that paper again now, it is essentially PEASI in an earlier accent — the same discipline of naming the problem, its enablers and its intended outcomes before reaching for an activity, years before I had a simpler name for it. Both projects were rigorous, important and, if I am candid, laborious.
But I kept finding myself asking the same questions. What is the problem? What change are we actually seeking? What conditions would allow that change to happen? What should we therefore do? How will we know whether we are moving in the right direction? And at what level of the system do those things sit?
In 2021, I started codifying those recurring questions into a simpler practical method. I called it PEASI. I now use it in some form in almost every complex problem I work on — sometimes as a rapid diagnostic around a table, sometimes to structure a co-design process, sometimes as the organising framework for a substantial evidence review, and sometimes expanded into a complex Theory of Change involving multiple organisations, professions and levels of government. The depth changes, but the discipline of starting with the problem does not.
It builds on, and complements, the range of existing models available by simplifying the engagement process by exploring five questions.
Complexity doesn’t mean confusion
Complex problems are not solved by pretending they are simple. But complexity also shouldn’t become an excuse for endless analysis. What I have been trying to develop is a disciplined way of sorting through the mess sufficiently to act intelligently.
That means resisting the urge to start with the activity. It means understanding the problem from the perspectives of the people experiencing it. It means examining the organisational and system conditions around them. It means working out where the leverage — and the authority — for change actually sits. Only then does it make sense to ask what we should do.
That is the idea behind PEASI. And after more than twenty years of working with complex workforce and health-system problems, I increasingly think that getting that first question right is where most of the work is.
PEASI™ is a practical method for making sense of complex problems and developing workable pathways for change. You can explore the method and download the free PEASI Canvas at healthworxfutures.com/peasi.
