Just in time

The Toyota Production System stands on two pillars. Just in time is one. The other is jidoka, usually translated as automation with a human touch, and what it means in practice is that any machine or any worker who detects a defect stops the process immediately.

Those two are not independent ideas that happen to sit together. Just in time removes the inventory that used to hide problems, and jidoka is what catches the problems once they are no longer hidden. Take the buffer away without the stopping mechanism and you have not implemented the system. You have removed its shock absorber and kept the road.

Almost every healthcare programme that describes itself as just in time has taken the first pillar and left the second, usually because the first saves money this year and the second costs authority. That is not a partial implementation. It is the dangerous half on its own.

On a Toyota line any worker can pull the andon cord and halt production. That is an extraordinary thing to give a junior employee and it is the safety valve that makes the rest of the system possible. When the buffer is thin and something goes wrong, the line stops, the problem is fixed, and nothing defective moves downstream.

A hospital has no equivalent. You cannot tell the ambulance to come back on Thursday. You cannot pause the emergency department while somebody investigates why the last three handovers went badly. Demand arrives whether or not you are ready for it, and refusing it is not a pause, it is a harm.

Follow that through and the conclusion is uncomfortable for anybody selling efficiency. The single mechanism that makes thin buffers safe in manufacturing is unavailable in an emergency pathway. So the buffer in a hospital must be larger than the equivalent buffer in a factory, not smaller, and an improvement programme that reasons from Toyota to a leaner ward has inverted the argument it is citing.

There is a version of the andon cord available, and it is worth naming because it is the thing worth building: the authority of a clinician or a nurse to stop admitting, to divert, to cancel a list, or to escalate and be listened to. Every hospital claims to have it. Very few have one that can be pulled by somebody junior at three in the morning without a career consequence, and the ones that do are measurably safer.

The third thing people forget is that Toyota levels its schedule before it removes inventory. Heijunka means smoothing production so that the same mix of work happens every day, which is what makes small buffers survivable, because tomorrow looks like today.

Hospitals have two populations and only one of them can be levelled. Elective work is scheduled by the hospital, so its variability is largely self-inflicted: lists that run Monday to Thursday, clinics that batch, discharges that cluster at five in the afternoon and not at all on Sundays. All of that is levellable and most of it is never levelled, which means there is a great deal of genuine just in time improvement available in the elective pathway that nobody is taking.

Emergency work cannot be levelled at all. The arrival process is outside the hospital, it is stochastic, and it has a mean but no manners. Applying a method that depends on levelled demand to a stream whose defining property is that it is not levelled is not rigour. It is a category error with clinical consequences.

So the practical rule is simple enough to be remembered in a meeting. Level what you schedule, and buffer what arrives.

None of this is an argument that the method does not apply. It applies powerfully in the parts of a hospital that meet its conditions, and those parts are large.

Consumables and supply, where lead times can be made reliable and a short stop is survivable, is the natural home of it, and two-bin kanban systems in a stockroom are among the most reliably successful improvements in healthcare operations anywhere. Sterile services, pharmacy resupply and theatre set preparation are the same problem.

Elective flow, once levelled, is the second. Pull rather than push is the third and the most underused: work that starts only when the next step is ready to receive it, so that a patient is not moved to a ward that cannot take them and a discharge is not prepared without the transport that completes it. Most hospital flow is push, which is how patients end up finished and stationary.

What those three have in common is that the process is inside the hospital, the variability is at least partly self-inflicted, and something can be stopped without anybody being harmed. Where all three hold, use the method hard. Where none do, do not.

There is one more condition, and it is the one that turns this page into part of the siting decision rather than a methodology essay.

Just in time requires suppliers who deliver when they said they would. Not quickly: reliably. The safety stock a hospital must hold depends on the variability of both demand and lead time, and the lead time term dominates surprisingly fast, because it is multiplied by the whole daily demand rather than by its standard deviation.

Which means the question of how much stock to hold is not a question about the stockroom or about the discipline of the people in it. It is a question about the supply chain outside the building, and the answer changes by nearly an order of magnitude between a dependable supplier and an undependable one.

Meridian is considering a market where lead times are longer and materially less predictable than in London, with import dependence, customs, currency and a supplier base that is thinner for specialist consumables.

The instinctive move for a group with a Lean reputation is to arrive and run the playbook: cut inventory, free the working capital, demonstrate discipline. The arithmetic above says the opposite. In a market with unreliable lead times the correct inventory position is substantially higher, and a group that cuts it will discover the cost as cancelled lists and unavailable implants, which will be recorded as clinical or operational failures rather than as the supply chain decision that produced them.

The honest version of the method says something more useful and much less impressive. Hold the stock the variability requires today, and spend the effort on reducing the variability, which for a new site means supplier development, local qualification, dual sourcing, and buffer stock held deliberately with the reason written down. Then, and only then, take the inventory out.

That is the same instruction as everywhere else in this study: remove the rocks and then lower the water. It is slower than the alternative, it is the version that works, and it is the version that survives being examined by a buyer who knows what questions to ask.