The four-hour clock¶

A patient arrives at an accident and emergency department. A clock starts. If the patient is admitted, transferred or discharged within four hours the attendance counts towards the standard, and if not it counts against it. Since 2004 the NHS in England has undertaken that patients should spend no more than four hours in A&E, and since 2010/11 the level of the standard has been 95 per cent rather than the 98 per cent it began with.

The clock is a good instrument in most of the ways an instrument can be good. It is objective, it is recorded automatically, it means the same thing in Truro as in Newcastle, and nobody has to agree on what counts as a serious case for it to work.

The shape of the curve¶

In February 2019 the Strategy Unit of NHS Midlands and Lancashire Commissioning Support Unit published an analysis of attendance durations, and the interesting part is not the average. It is the shape.

“The formulation of the 4-hour target tends to focus attention on the timeliness of treatment of a small subset of patients attending A&E, causing a distortion in the distribution of attendance durations.”

The curve does what a curve of that kind should do until it approaches the boundary. Few patients leave within half an hour, most leave after about two hours, and the frequency then falls away. Then, in the report’s words, “a significant spike in the frequency curve occurs in the 30 minutes leading up the 4-hour mark. This is followed by a commensurate trough in attendance durations after 4 hours.”

A histogram of English A&E attendances has a tower in it, at 211 to 240 minutes, and a hole immediately after.

The patients at three hours and forty minutes¶

The report is careful about the cause and offers the explanation most consistently cited: “that A&E staff identify a subset of patients who are likely to imminently breach the 4-hour target and expedite care and treatment to avoid such a breach. Similar efforts are not made for attendances that are expected to be substantially shorter or longer.”

Nothing in that sentence is falsification. Nobody alters a record. Clinicians look at who is about to cross a line and get on with those people first, which is a defensible thing to do with a queue and an entirely rational thing to do with a queue that is being scored.

What it changes is the order. Attention moves towards patients at three hours and forty minutes and away from patients at ninety minutes and at six hours, and proximity to a deadline is not the same quantity as clinical need. The target did not make the department faster. It made it faster at the boundary.

Ten per cent, and steady¶

The analysis then does the thing that makes it evidence rather than an anecdote. It estimates the size of the distortion as the difference between attendances ending in the half hour before the target and the average of the half hours either side, and tracks it monthly from April 2010 to March 2016.

“Following some initial increases in 2010/11 when the target level was reset from 98% to 95%, there has been little change in the size of the distorting effect. This implies that the ability of the current A&E system to distort the distribution of attendance durations is marginal (c. 10%) and stable. This is plausible because distortion requires the use of limited management resources.”

Around a tenth of attendances, unchanging across six years, and bounded by how much management attention a hospital has to spend on the boundary. The distortion is real, measurable and modest, and it did not grow as the pressure did.

The deterioration the clock did not cause¶

The same report is emphatic that the collapse in measured performance is not an artefact. No single factor explains the rise in four-hour breaches since 2010: patients arriving are more complex and more acutely ill, inpatient occupancy has risen and with it the wait between a decision to admit and an admission, more tests are being done, and departments have become slower to recover from periods of pressure. Taken together, the analysis says, those factors explain most of the deterioration.

The National Audit Office records that July 2015 was the last month the standard was met, that 711,881 patients waited over four hours in December 2022, an all-time high, and that general and acute beds ran at 92.3 per cent occupancy in the last quarter of 2022-23.

A hospital at 92 per cent occupancy is not failing a clock because of the clock.

The clerk’s brief¶

June 2023: The audit office counts the queue¶

The National Audit Office reported on access to unplanned or urgent care, recording 711,881 A&E patients waiting over four hours in December 2022, an all-time high; 90,998 ambulance handovers taking longer than 30 minutes in March 2023, 25.9 per cent of all handovers; general and acute bed occupancy at 92.3 per cent in the fourth quarter of 2022-23; and July 2015 as the last time the NHS met the standard. The clerks note that the standard has now been missed for longer than it was ever met.

February 2019: Somebody plots the distribution¶

The Strategy Unit of NHS Midlands and Lancashire Commissioning Support Unit published an analysis of attendance durations, finding a significant spike in the frequency curve in the thirty minutes before four hours and a commensurate trough after it, attributing this to staff expediting patients likely to breach imminently, and estimating the effect at around ten per cent of attendances and stable from 2010 to 2016. The clerks note that the report declines to call this gaming, and that the behaviour it describes is what a clinician would do if told that one line among many is the one that is counted.

2010: The level is lowered¶

The level of the standard was reset from 98 per cent to 95 per cent in 2010/11, and the Strategy Unit records an increase in the distorting effect at that point, after which it settled. The clerks observe that moving a target changed the shape of the distribution before it changed anything about a patient.

2004: The clock starts¶

The NHS in England undertook that patients should spend no more than four hours in accident and emergency. The clerks record that the undertaking was expressed in a unit nobody has to interpret, that this is why it worked as a measure, and that it is also why it can only ever have described the crossing of one line.

The last thirty minutes¶

Patients were counted, clocks were read, and the percentages published each month were what the records contained. What the percentage could not show was its own shape: a tower of attendances in the half hour before the boundary and a hole immediately after, worth about a tenth of everybody who came through the door. The clerks’ standing assessment is that a target expressed as a deadline buys a reordering of attention around the deadline, that the reordering is bought at the expense of patients who are nowhere near it, and that this cost appears in no published figure because the published figure is a count of one side of a line.