TL;DR. Flow management helps identify dependencies between the tasks involved in preparing hospital discharge. Transferring it to healthcare means making waits understandable, preparing what can be prepared and revising forecasts as the patient’s situation changes. Individualised discharge planning has shown encouraging clinical results; the organisational proposals discussed here still need to be evaluated locally.

Lately, I’ve spent a lot of time studying flows: the movement of people, the circulation of information, digital exchanges and production flows. I wanted to understand what they have in common, what slows them down and what helps organise them better. In the spirit of the Transfert series, I wondered how these principles might shed light on certain care pathways. Preparing hospital discharge is a first setting in which to explore that question.

A hospital and a home connected by several converging paths.

A conceptual illustration of preparing the return home.

This article is part of Transfert, a project exploring how principles from other fields can be applied to care. It joins the catalogue of transfers, in French.

Contents

Imagine someone whose return home is becoming a possibility. The team still needs to confirm that their condition allows it. Support needs to be arranged, information required for the handover needs to be shared, and transport may be necessary. The patient asks when they can go home. The answer now depends on several people working in different places and to different schedules.

This fictional situation raises a practical question: how do we prepare a transition when all the necessary conditions are not yet in place?

This is what interests me about flow management. It helps us look at the time between actions and understand what makes the next action possible. Applied to hospital discharge, it could help prepare the handover while allowing for clinical uncertainty and the patient’s choices.

Discharge brings several conditions together

In a production system, an operation may depend on several elements prepared separately. Until they are all available, the next step remains on hold. The time required depends on the branch that finishes last and on any dependencies between branches.

Discharge preparation can be represented in this way, with an essential difference: some conditions change and need reassessment. Available transport means little if the patient’s clinical condition has changed. Promised support is insufficient if it does not meet the person’s actual needs at home.

For our example, let us distinguish four conditions: clinical readiness for discharge, a feasible arrangement for ongoing care, the necessary information being available, and transport being organised. This is a proposed framework for analysis, to be adapted with professionals and the person concerned.

Suppose, purely to understand the mechanism, that the three organisational tasks take four, two and one hours respectively. If they are independent and can start together, their minimum total duration is four hours. If they happen one after another, it becomes seven hours. These durations are invented for this calculation; they do not describe any hospital service.

The point of this reasoning is to identify which steps can actually proceed in parallel. Some requests can be prepared ahead of time; others require a prior decision. We also need to consider work that becomes unnecessary when circumstances change. Preparing early has a cost, which needs to be weighed against the delay it might prevent.

The question becomes specific: what information would allow this task to start earlier, and with what acceptable degree of uncertainty?

Making sense of “waiting”

Let us return to our hospitalised person. Support at home has not yet been confirmed. That sentence can describe very different situations.

The need may have been identified without a request being sent. A request may have been sent but be missing a document. The recipient may have received it and be assessing whether it is feasible. They may also have replied that no capacity is available.

These situations call for different actions. Following up with a partner who is waiting for information does not complete their file. Sending more requests to a provider without capacity does not create that capacity. Equally, adding resources elsewhere will not necessarily resolve an incomplete handover.

I would therefore try to describe the actual state of the task, with an identifiable next action. A statement such as “file sent, document X awaited, team Y to follow up” provides more direction than a broad category such as “social issue”.

This precision needs to remain proportionate. Each added field should support a decision or a handover. If teams have to maintain a separate dashboard alongside their daily work, we risk asking them to take on another activity whose usefulness remains to be demonstrated.

Locating today’s constraint

Goldratt distinguishes the performance of an operation from that of the entire flow. His discussion of production systems also reminds us that applications rest on assumptions specific to their environment. This distinction is useful when transferring an idea: a local improvement needs to be assessed through its effect on the rest of the care pathway. Goldratt’s original text.

In our example, imagine that transport is available, documents are ready and discharge is clinically possible. Support at home still needs confirmation. Preparing the documents even faster will do little to change the departure time that day. Understanding why support has not been confirmed, however, may identify a useful action or establish that an arrangement is genuinely unavailable.

The next day, for another person, the constraint may be different. The method needs to leave room for that possibility. A permanently designated “ward bottleneck” could obscure variations across needs, days or partners.

We also need to distinguish a temporary wait from a sustained shortfall. Complete requests that repeatedly wait because no professionals are available raise a capacity question. Files that circulate several times to obtain the same information also raise a question about how work is organised. Both can coexist.

This distinction gives measurement a purpose: documenting what depends on better coordination and what genuinely requires additional resources.

Planning ahead with a forecast that can change

Discharge cannot always be predicted at admission. Some needs can nevertheless be identified before the exact departure date is known. The challenge is to organise this preparation without turning an estimate into an irrevocable commitment.

A team could distinguish what is likely, what is confirmed and what has just changed. The expected date would be a working estimate accompanied by its conditions. It could be revised as the patient’s clinical course, needs or preferences require.

It would then be useful to identify which steps are reversible. Gathering information or checking whether support might be available does not involve the same commitment as a booking that is difficult to change. Gradual preparation would allow effort to be focused as the situation becomes clearer.

This approach raises another question: when do partners need the information in order to act? A document completed quickly after the decision may arrive too late for their own working schedule. The ward’s internal timeline represents only part of the timeline of the care pathway.

In queueing theory, variability in arrivals, service times and the number of resources influence waiting. Pooling and priority rules also change the results. None of this yields an ideal occupancy rate that applies indiscriminately to a hospital. Dai and He, 2011.

The practical translation would be to observe when requests arrive and when they can be processed. If they cluster within a short period, we would need to understand what causes that clustering before trying to change it.

What individualised discharge planning has shown

The Cochrane review by Gonçalves-Bradley and colleagues, published in 2022, includes 33 trials and 12,242 participants. For patients admitted with a medical condition, individualised discharge planning is associated with a mean reduction in length of stay of 0.73 days. The 95% confidence interval ranges from a reduction of 0.12 to 1.33 days; this result comes from 11 trials involving 2,113 participants, with moderate-certainty evidence.

Readmissions are also slightly reduced. Effects on costs remain uncertain. The populations studied were predominantly older people, and the interventions varied. Official Cochrane review summary.

This evidence supports preparation tailored to the person’s needs. It does not establish that the benefit comes from a particular tracking dashboard, or promise the same improvement in every service. The proposals developed here still need to be tested in context.

Measuring waits while preserving clinical judgement

The most delicate measurement would be the point at which a wait becomes organisational. A statement that “discharge is possible” may depend on the information available, the team’s assessment and the conditions of the return home. Its date should not be treated as a simple, definitive fact.

For local work, the team could define the criteria it uses and then document revisions. If the patient’s clinical condition changes, the situation needs to be reclassified without this being treated as a performance failure. Disagreement between professionals may itself point to information that needs clarification.

We could then track several moments: the need for ongoing support identified, a complete request received, a response obtained, the conditions for transition met, and actual departure. These markers allow distinct intervals to be examined and help avoid attributing the entire delay to the last visible participant.

Under its conditions, Little’s Law relates the average number of people present, throughput and average length of stay. It can help check the consistency of measurements. It does not determine the clinically necessary duration of a stay or demonstrate that earlier departure is desirable. Little, 2011.

Across care pathways, ongoing stays also need to remain in view. Looking only at people already discharged leaves those whose difficulties are still unresolved out of the calculation. The age of ongoing waits would add information to the average duration of completed waits.

A local trial whose results we could interpret

I would start with a recurring cause, within a setting where the people involved can work together. For example: a request for ongoing support that is regularly completed late. The hypothesis would be that identifying the necessary information earlier reduces that specific interval.

Before changing the organisation, observing a few consecutive care pathways would help establish what is missing, when and why. Both straightforward and difficult situations would need to be included. This initial observation would help formulate the intervention, rather than demonstrate its effectiveness.

The change could be limited to how the request is prepared and how its completeness is confirmed. We would then track the targeted interval, the work time required and requests that became unnecessary following a change in circumstances. The service could thus see whether preparing ahead brings a benefit or mainly creates additional work.

Interpreting the results requires attention to the period, workload and patient characteristics. A shorter delay during a quieter period is insufficient to attribute the improvement to the change. A structured comparison, suited to what is feasible locally, would strengthen the evaluation.

The scope would also need to extend beyond departure. A readmission, difficulty accessing support or a task transferred to a relative belongs in the assessment. We would need to ask what was actually made possible at home, and at what human cost.

This is how organising flow connects with care. In our example, the desired result would be someone returning home when their situation allows it, with feasible support and clearly identified contacts. Understanding the waits can help teams prepare for that moment and explain what remains necessary until then.

References and scope of the evidence

  • Gonçalves-Bradley DC et al. Discharge planning from hospital. Cochrane Database of Systematic Reviews, 2022, CD000313. DOI: 10.1002/14651858.CD000313.pub6. Detailed official summary consulted; literature search current to April 2021.
  • Goldratt EM. Standing on the Shoulders of Giants: Production Concepts versus Production Applications. Copy bearing a 2006 copyright notice, pages 4–8 consulted. Management essay, without clinical validation of this transfer.
  • Dai JG, He S. Queues in Service Systems: Customer Abandonment and Diffusion Approximations. Tutorial, 2011. Mathematical models under explicit assumptions.
  • Little JDC. Little’s Law as Viewed on Its 50th Anniversary. Operations Research, 2011, 59(3):536–549.

Links to the original sources appear throughout the text. The scenario and illustrative durations are invented. The proposed observation protocol has not been tested here; the article is not a discharge recommendation for an individual patient.