We keep treating downstream demand as a hospital problem when much of it begins somewhere else
Hospitals are very good at making problems visible. A person arrives in an emergency department, a clock starts measuring waiting time, a bed may be occupied, or a discharge is delayed, and a cost is coded in an account.
What is much harder to see is everything that happened before the person got there: the older person whose mobility gradually deteriorated until living at home was no longer safe; the person whose housing became unstable; the family that could no longer sustain an informal caring arrangement; the person with a drug or alcohol problem for whom crisis care became the most reliable point of access to the system; the person with chronic disease whose capacity to manage at home progressively unravelled; the person with disability whose support arrangement broke down; the person experiencing mental distress who needed sustained support long before they needed an emergency department. None of these is primarily a hospital problem. But eventually, if enough other parts of the system fail to hold, the hospital is where the failure shows up.
That distinction is important, because once a problem appears inside a hospital, we treat it as hospital demand. We count beds, presentations, length of stay, delayed discharge. We ask whether we need more emergency capacity, more wards, more clinicians, more discharge coordinators. Sometimes we do — hospitals are essential, and acute medicine and nursing do work that cannot simply be shifted somewhere else. But the hospital is also increasingly the default holding infrastructure for problems generated elsewhere in society. And if we mistake where a problem becomes visible for where it originates, we will keep building capacity at the wrong end of the system.
Need arises in people’s lives. Institutions divide it into categories. Hospital is where many of the unresolved pieces eventually reconverge.
Hospital demand is partly socially produced
This is not simply an argument about moving more healthcare into the community. Some of what eventually appears as hospital demand is produced well outside healthcare.
A 2025 review of three decades of admission-avoidance research identified deprivation, social isolation, living alone, multimorbidity and weaknesses in service access and coordination as interacting contributors to emergency admission.¹ The important point is not that any one of these predicts hospital use on its own. It is that clinical need, social circumstances and service capacity interact. A person living alone with heart failure, deteriorating mobility, insecure housing and limited access to community support does not have four separate problems. They have one life. But the systems responding to that life divide the problem between healthcare, housing, social care and community services.
By the time that complexity reaches an emergency department, we have converted a clinical, social and system problem into a hospital event.
We have known the policy direction for decades
This is not a new insight in health policy. For decades, health systems across wealthy countries have been attempting to move care away from institutions and towards people’s everyday lives: mental health care was deinstitutionalised, chronic disease management moved towards primary and community care, disability support was progressively separated from hospital-based health services, ageing policy increasingly centred on remaining independent at home, rehabilitation moved into communities, and primary care reform embraced multidisciplinary teams. The language changed too — person-centred care, care closer to home, prevention, early intervention, integrated care, hospital avoidance, restorative care.
The policy moved, as did much of the care. But we did not redesign the architecture underneath it at anything like the same pace. That is the problem I explored in We Moved the Care. We Kept Building the Same Architecture.
The result is a peculiar health system where we increasingly expect the community to absorb complexity, but when it cannot, the hospital remains the place with the strongest institutional capacity to catch it.
The hospital is measurable in a way that the upstream system isn’t
There is an important reason this keeps happening: hospitals generate unusually good metrics: bed occupancy, patient numbers, admission and discharge dates, diagnostic categories, procedures performed, and money attached to the episode. That makes the hospital extraordinarily legible to policymakers.
Compare that with the things that may have prevented the admission: someone noticing that an older person had become progressively less steady on their feet; a physiotherapist improving strength; an occupational therapist identifying why the shower had become unsafe; a pharmacist simplifying medicines; a support worker helping someone practise the tasks needed to remain independent; a community nurse noticing deterioration; a social worker resolving a housing or carer crisis; a peer worker helping someone remain connected to treatment; reliable transport; a meal; a functioning family support network; a home modification; a timely drug and alcohol service.
Some of those activities generate health-system data. Many do not. Some belong administratively to healthcare; others belong to aged care, disability, housing, social services, local government or the community sector. Some are formal work, some are unpaid, and some are not even recognised as an intervention. Yet together they can determine whether someone remains safely in their own life or becomes a patient in a hospital. Once the person enters hospital, all that complexity collapses into one beautifully measurable event: an admission. That is a data problem. But it is also a governance problem.
We count where the system has a boundary
Nearly twenty years ago, my colleagues and I encountered a version of this problem while studying community rehabilitation and intermediate care for older people. These were services designed to work across traditional service and role boundaries with the specific aim of keeping people out of hospital — hospital and home, health and social care, professional and support work, treatment and rehabilitation — and before we could even compare them, we realised we could not even describe them consistently. A systematic analysis produced 334 different elements being used to characterise apparently similar services, which we eventually reduced to six broad domains: context, purpose, service users, access, service structure and organisation of care.²
The gaps were revealing. Population context was often poorly described. Workforce numbers, skills and grades were inconsistently reported. The intensity of care was often unclear. Even capacity meant different things in different studies — beds, referrals, admissions, throughput, places or length of stay. Services have a very sophisticated system for counting activity once someone reached an institution, and a much weaker system for describing the architecture keeping them out of one.
When we later went looking for that architecture more systematically — auditing 186 community rehabilitation and intermediate care services across the UK — we found something almost as telling as the 334 elements. There was no consistent relationship between what a service was structured to do and how it was staffed.³ Two services delivering the same nominal level of care could have entirely different skill mixes, entirely different staff-to-referral ratios, entirely different assumptions about what the work required. The variation wasn’t explained by patient need; it was explained by history — who happened to have set the service up, on what budget, reporting to whom.
That is what an unplanned architecture looks like from the inside. Not quite chaos, exactly, just accumulated, uncoordinated decisions that nobody was ever asked to reconcile.
A bed is not a unit of population need
A hospital bed is a useful thing to count, but it is not the same thing as health-system capacity. Nor is an emergency presentation, a GP consultation or an allied health appointment. These are units through which existing institutions make their work visible; population need is something else entirely.
Consider an older person becoming less able to manage independently at home. What do they need? Some combination, most likely, of medical assessment, medication review, rehabilitation, strength and balance work, personal care, nutrition, cognitive support, equipment, housing modification, social connection, transport, carer support and ongoing monitoring — a package that is different for every person, that no profession owns, that often no single organisation owns, and that there may be no single funding stream capable of assembling. That is the architectural problem: the person’s need is integrated, and the institutions responding to it are not.
This is why “integrated care” so often becomes a coordination project
When systems recognise this problem, the usual response is to improve integration: create a referral pathway, employ a navigator, introduce case conferencing, create a multidisciplinary team, share records, improve discharge planning. All of those things may help. But they mostly operate after the underlying divisions have already been created — the funding remains separate, the organisations remain separate, professional jurisdictions remain separate, eligibility rules remain separate, workforce data remain separate, industrial structures remain separate, accountability remains separate — and then we create another role or process to reconnect them.
That is not necessarily integration. Sometimes it is simply coordination across dis-integration. There is a difference.
The hospital then becomes the institution that cannot say no
Not literally, of course — hospitals triage, refer and discharge people constantly. But structurally, an acute hospital has obligations and capacities that many upstream services do not. A community service may have eligibility criteria, a waiting list, restricted hours, a defined scope, a capped budget, a particular catchment, a workforce vacancy, or no presence at all in a particular region. The person can fail to fit each of those categories, but their need does not disappear. Eventually something happens that turns it into an emergency, and the acute system receives the accumulated consequence.
This is one reason I am increasingly uncomfortable when debates about hospital pressure begin and end with the hospital workforce. The people inside hospitals are often dealing with the final manifestation of problems they did not create and cannot solve from inside the hospital. The question is not simply how we make hospitals more efficient. What clinical, functional and social need is becoming visible as hospital demand because the wider system could not resolve it earlier?
The pattern is not hypothetical
A workforce analysis I carried out for an Australian state’s public hospital system a few years ago is as clear an illustration of this dynamic as I have come across.⁴ The population it served had significantly higher need for occupational therapy than the national average — an older population, higher rates of disability support, lower socioeconomic status, higher rates of diabetes and stroke — while access to occupational therapy sat well below the national average. The population that needed the most had access to the least.
The consequence was not that occupational therapy shrank. Overall numbers actually grew, driven by disability and aged care reform. But almost all of that growth landed in the private sector, while the public hospital establishment barely moved. And within the public system, the scarce workforce that remained was steadily pulled toward the acute end of care: toward inpatient wards, toward discharge, toward the presenting emergency. Community-based, preventive, admission-avoiding work — the work occupational therapy is best evidenced to do — was the first thing rationed away.
One community-based occupational therapist put the mechanism more plainly than any policy document could: “Everything needs to be solved in the hospital due to lack of community services. We never have any capacity to fix the system stuff because [we’re] always pulled into patient care.”
This is the whole argument in miniature. Population need was highest exactly where system capacity was lowest. The workforce that existed was steadily reallocated toward the place where failure becomes visible, at the direct expense of the work that would have prevented the failure occurring. And every one of those reallocation decisions was locally rational — a manager doing the only thing available to keep patients safe today, at the cost of the capacity that would have kept them safe next year. Nobody designed that outcome. The architecture produced it anyway.
This isn’t an isolated case. A much larger workforce research program in Victoria found the same underlying mechanism at a state-wide scale: across most allied health professions, the constraint was rarely a shortage of workers. It was a shortage of funding — allied health was consistently treated as a “Cinderella service,” the first to be cut when budgets tightened, regardless of the evidence for what it delivered.⁵
Herein lies the paradox. Under funding pressure, the work that gets protected is the work directed at the presenting emergency — acute assessment, immediate risk, discharge. The work that gets cut is prevention, rehabilitation and community maintenance, because its benefits are more difficult to count, realised later, and elsewhere, outside the budget period in which the cut is made. So the system responds to scarcity by doing more of the expensive, visible work and less of the cheap, invisible work that would have reduced demand for it. Tightening the funding doesn’t just fail to fix the underlying problem — it makes the problem worse, in a cycle that reinforces itself: rising acute demand consumes the capacity that might have prevented it, which drives acute demand higher again.
That is the real question this pattern points to. Not how do we make hospitals more efficient, but what work needs to happen before the crisis — and who pays for it — so the cycle doesn’t simply repeat with next year’s budget?
This changes the workforce question
Once you ask that, health workforce planning looks different. You stop beginning with how many doctors, how many nurses, how many physiotherapists, how many care workers — and begin further upstream instead. What needs exist in this population? What work would prevent deterioration, restore capacity or support people to live well? What capabilities does that work require, and where should those capabilities sit so that people can reach them before crisis? What needs professional judgement, and what requires time, continuity, relationship or repeated support? What funding and governance would allow those capabilities to operate together? Only then does it make sense to ask which occupations should deliver the work.
That is not anti-professional. Deep professional expertise matters enormously, and some boundaries should remain because the consequences of error are serious and complex judgement requires extensive training. But professional categories should be an answer to the workforce-design problem. They should not automatically define the problem.
Our multidisciplinary health workforce
A great deal of my own research has involved allied health, rehabilitation and support work — not because I think health-system reform should revolve around allied health, but because these workforces frequently operate at the seams: between hospital and home, between treatment and function, between health and social care, between professional judgement and repeated hands-on support, between registered and unregistered work. Boundary workforces reveal the architecture particularly clearly.
If a person needs physiotherapy, occupational therapy, nursing, social support, personal care and housing assistance to remain safely at home, the interesting question is not which profession deserves more recognition. It is why the system makes it so difficult to assemble those capabilities around the person in the first place. That is the question I want Health Work 2.0 to ask.
The hospital is the warning light
There is a useful way to think about this. A warning light on a dashboard tells you where a problem has become detectable; it does not necessarily tell you which component created it. Hospital demand works the same way. An emergency department under pressure is real, a delayed discharge is real, an occupied bed is real — but each may also be signalling something about the wider system: insufficient community capacity, poor access, unmet rehabilitation need, housing instability, weak social support, fragmented funding, workforce shortages, workforce capability that exists but cannot be deployed, or institutions that simply do not connect around the person’s need.
The hospital is often where all of those failures become visible at once. If we respond only by enlarging the place where the warning light appears, we should not be surprised when it comes on again.
Start with the need, not where the failure landed
For most of my career I have worked around the edges of this problem — community rehabilitation, intermediate care, professional boundaries, support workforces, workforce planning, regulation, career structures, classification, integrated care. For a long time they looked like separate research questions. I increasingly think they were versions of the same one: we organise systems around institutional categories and then ask people to experience them as integrated care. The hospital is where the contradiction becomes hardest to ignore.
Perhaps the next generation of health workforce planning needs to start somewhere else entirely — not with the hospital, not with primary care, not with allied health, not with medicine, not even with the workforce.
Start with the population. Then the need. Then the work. Then the capability. Then build the architecture capable of putting those things together before the hospital has to catch what fell through.
That is where Health Work 2.0 begins.
Notes
- Eusebio, C. and Leary, A. (2025) ‘Admission avoidance in UK healthcare: what works and what doesn’t? A hermeneutic review’, medRxiv preprint. https://doi.org/10.1101/2025.03.06.25323302
- Nancarrow, S., Moran, A. and Parker, S. (2009) ‘Understanding service context: development of a service proforma to describe and measure older people’s community and intermediate care services’, Health and Social Care in the Community, 17(5), pp. 434–446.
- Nancarrow, S.A., Moran, A., Freeman, J., Enderby, P., Dixon, S., Parker, S. and Bradburn, M. (2009) ‘Looking inside the black box of community rehabilitation and intermediate care teams in the United Kingdom: an audit of service and staffing configuration’, Quality in Primary Care, 17, pp. 323–333.
- Based on an unpublished workforce analysis conducted by the author for an Australian state public hospital system (2022). Details have been de-identified, as the underlying report is not in the public domain; the pattern described is broadly representative of workforce dynamics documented across multiple Australian jurisdictions.
- Findings of the Victorian Allied Health Workforce Research Program, commissioned by the Victorian Department of Health and Human Services. Related work here.
