From Requires Improvement to Good: the action plan that actually moves the rating

On a return inspection following Requires Improvement, the inspector arrives with the previous report and the action plan the provider submitted in response. We read the plan against the findings. We read the findings against the practice we observe on the day. We read the practice against what the team can tell us about how it changed.

The reading is not symmetrical. We are looking at the plan to find specific things: actions that target the actual cause of the original finding, owners who can answer for the action without referring up, due dates that the team treated as commitments rather than forecasts, closure paragraphs that describe what is now different rather than what was completed.

We are looking at the practice to test whether the changes the plan claims actually happened. Sampling five to ten records against the relevant action lines. Talking to staff about the new protocols. Reading the assurance calendar to see if the audit cycle actually re-audited the gap. Asking the registered manager specific questions about the underlying clinical patterns the original report flagged.

By the time we left, every return inspection fell into one of three patterns. Each one shaped the report.

Pattern one: the plan moved the practice. The actions targeted the actual causes; the closures evidenced specific changes; the audit re-audits showed the gaps closing; the team could describe what was now different and why. The findings on the new report were different from the previous report's findings. The rating moved from Requires Improvement to Good, or to Good in most key questions and an in-progress finding on one. This pattern was less common than I had expected when I started doing the work. The teams that achieved it tended to have done something specific in month two: they took the action plan apart and re-wrote each line until the underlying change was clear and the closure test was unambiguous.

Pattern two: the plan was completed; the practice did not move. Every action ticked green; the audit cycle re-audited but the percentages were unchanged; the team could describe the actions but not the changes they were supposed to cause. The rating stayed at Requires Improvement and the findings recurred. This was the most common pattern. The teams had treated the plan as the work rather than as commitments to a specific practice change. Some teams in this pattern were demoralised after the return; the energy had gone into the plan and the underlying problem remained.

Pattern three: the plan touched the practice partially. Two or three of the original findings showed genuine change; one or two showed paperwork-only completion. The rating shifted in mixed directions: improved on some key questions, unchanged on others. This pattern was less stable; what distinguished pattern three from pattern one was usually that the team had not yet figured out the difference for the trickier findings. Pattern three often became pattern one on the inspection after that if the team had a clear-eyed read of which lines had actually moved practice.

The teams that achieved pattern one tended to do four things in the six to eight months between inspections.

One: they re-wrote the plan in month two. The first version of the plan, submitted to CQC within the published window, was the version the regulator saw. Two or three months in, the registered manager took the plan apart and re-wrote each action until the change-to-practice was unambiguous. The re-write was internal. The CQC submission did not change. The internal version became the working document.

Two: they distinguished completion from change. Every action carried two closure criteria: the completion criterion (what done looks like) and the change criterion (what is now different in the practice the action was supposed to affect). Completion without change did not close the action. This is the discipline the platform exists to enforce by default; without it, the temptation to mark the action complete on the completion criterion alone is considerable.

Three: they re-audited fast. The re-audit on the assurance calendar was scheduled three to four months after the change was implemented, not twelve months later. Three months is long enough that the change had time to embed and short enough that the team remembered why they made it. The re-audit either showed the gap closing (action complete with evidence) or showed the gap unchanged (action incomplete, root cause not addressed, needs a different intervention). Either way the team knew sooner.

Four: they accepted that some findings needed a different intervention. The first action the team tried was an educated guess at what would work. If the re-audit at month four showed the gap was unchanged, the team did not double down on the same intervention. They escalated: a different control, a different owner, a structural change rather than a policy change. Re-audit at month seven tested the new intervention. By the return inspection, the audit cycle showed two attempts, both documented, and the second one working. That is a service learning. That is what we wanted to read.

The single best signal in the action plan is the closure paragraph. A plan I read once carried this closure on a finding about consent quality: “Training delivered to all clinical staff on 14 February. Policy updated to v3.1 same date. Sign-off completed.” Three sentences of completion. The return inspection sampled five elective surgical records and three of them had consent gaps similar to the original finding.

A plan I read on a similar finding three months later carried this closure: “Training delivered to all clinical staff on 14 February. Audit on 18 May sampled twelve elective consent forms and showed 96 per cent compliance with the revised standard, up from the 78 per cent in the original finding. Two of the twelve non-compliant lines were the same consultant; one-to- one supervision arranged for that consultant 22 May with re-audit scheduled for August.” That is a closure paragraph that signals a service learning. The return rating on that finding moved from Requires Improvement to Good.

The pattern is small enough that it is easy to dismiss. The pattern is also consistent enough across the return inspections I did that I could often read the action plan in advance of the visit and predict the rating outcome. Completion-language plans returned to Requires Improvement. Change-language plans moved to Good. The discipline of writing the change rather than the completion is the discipline of running the improvement as actual improvement rather than performative compliance.

The platform we built makes the change-language discipline easier to maintain. Every improvement action carries a named change-to-practice field at closure; closure without it is a guarded step the platform makes deliberately visible. Audit-cycle items spawn re-audits on the assurance calendar at the cadence the team commits to. The aggregate dashboard shows the open actions, the closure rhythm, and the audit cycle in one place so the team can see the rhythm of improvement without a separate reporting project.

None of this replaces the work. The discipline is the work. The platform reduces the friction of doing the discipline well by removing the scattered-systems problem that makes completion-language tempting. Closure paragraphs that describe change rather than completion are still written by humans; the platform makes the change paragraph the path of least resistance.

On the wider Well-led pattern the rating changes sit inside: Why providers fail Well-led. On the first thirty minutes of the inspection that leads to a rating in the first place: Unannounced CQC inspection: the first thirty minutes. On the action lifecycle itself: Improvement action plans that actually move the rating.