Why the Pharmacy Fill-Error Gain Slipped Away

The process did not revert on its own; four specific control mechanisms degraded, and they interact.

1. The audit - the primary detection mechanism - collapsed Completion fell from 95% to 40%. Whatever the checklist would have caught, it was not being looked at for most of the week. This is the first failure because it removed your ability to see the other three while they were small.

2. Standard work was not transferred to new people Five of 22 technicians - nearly a quarter of the workforce - were never trained on the post-kaizen standard. Errors from untrained staff are not a process regression; they are a predictable consequence of a control plan that had no onboarding step. The control plan addressed the people in the room at the time of the kaizen and nobody else.

3. The hard-stop override became routine Override use went from 3% to 18%, a six-fold increase. The barcode scan was the strongest error-proofing device in the improved process. At 18%, it is no longer a hard stop; it is a suggestion. This alone can account for a substantial part of the drift, and it deserves its own Pareto before anything else is changed.

4. The reaction plan depended on a signal that stopped arriving The plan required action after two consecutive audit failures. With audits at 40% and irregular, two consecutive failures become unlikely and the trigger is effectively dead.

5. Ownership The lead pharmacist was the plan's single point of execution with no deputy and no escalation when completion dropped. A control plan with one owner and no cover is a plan with a single point of failure.

Order of investigation: override rate first, then untrained staff, then audit design.

What is wrong with the current chart

1. Wrong chart type. The metric is errors per 10,000 prescriptions - a count of defectives or defects relative to a varying opportunity. An individuals chart ignores the changing denominator. If prescription volume varies by day, the limits are wrong and the chart will signal on busy days. Use a p-chart if each prescription is judged pass/fail, or a u-chart if multiple errors per prescription are possible, with limits computed per subgroup size.
2. Limits computed from the improvement period. Twelve points, and they are the points collected while a special cause (the kaizen and the attention around it) was active. Those limits encode the project, not the process. They are almost certainly too tight, so ordinary variation will look like a signal, or the reverse - the chart generates false alarms and gets ignored, which is the fastest route to a chart nobody uses.
3. Twelve points is too few. A baseline needs roughly 20–25 subgroups, and you need to establish stability before computing limits.

Replacement

• Chart: p-chart (or u-chart) on errors per prescriptions filled, with daily or per-shift subgroups sized so the expected count per subgroup is at least 5. If subgroup size varies, plot variable limits or use a Laney p′ chart if overdispersion is present.
• Limits: recompute from a stable historical baseline of at least 20–25 subgroups that genuinely represent routine operation - including the period when the kaizen support was withdrawn, since that is the state you must sustain.
• Phases: show the pre-kaizen, pilot and current periods as separate phases with their own limits, so the chart tells the story rather than hiding it.
• Supplementary chart: plot override rate on its own chart. It is a leading indicator and will move before the error rate does.

Revised reaction plan - trigger, action, owner, escalation

Trigger	Action	Owner	Escalation
One point outside the p-chart control limits, or a run of 7 on one side	Log the event, review the last 20 prescriptions for common cause, report at the weekly huddle	Shift lead	Pharmacy manager if the next point is also out
Audit completion below 90% in any week	Named deputy completes the outstanding audits within 48 hours	Lead pharmacist and deputy	Manager if below 90% two weeks running
Override rate above 6% in any week	Review every override reason code; require a documented reason per override	Lead pharmacist	Manager and IT if above 10%
Any new technician before independent shift work	Sign-off on standard work plus 20 observed prescriptions	Designated preceptor	Manager
Two consecutive failed audit items	Re-train and re-audit within 5 working days; verify with 20 observed prescriptions	Preceptor	Manager

Next 30 days

1. Week 1: publish the weekly dashboard (error rate p-chart, audit completion, override rate) and confirm a named deputy for the lead pharmacist. Root-cause the override increase by reason code - is it scanning failures, missing labels, or workaround behaviour under time pressure?
2. Week 2: train the five new technicians and add standard work to onboarding so it cannot recur. Retrain all staff on when an override is permitted and what it must be documented against.
3. Weeks 2–4: rebuild the audit so it survives a busy week - short daily sample rather than a long three-times-weekly exercise, and rotate auditors so it is not one person's burden.
4. Week 4: recompute control limits from the stable baseline, review against the 1.4 target, and report honestly whether the gain is recovered or partly lost.
5. Close-out lesson: add a control-plan review at 30 and 90 days after every project close, checking that the plan is functioning, not merely written.