Sepsis Bundle at 54%: Training Problem or System Problem?

Not mostly, and the audit says so.

• The steps fail at different rates. Cultures 62%, fluids 48%, repeat lactate 39%. A knowledge deficit would depress all steps similarly. Divergence within the same bundle points to different barriers per step.
• The nurses describe an authority barrier, not an information gap: they cannot get a lactate before a physician agrees to order fluids. A 45-minute lecture cannot fix an order set or a protocol scope.
• Unit variation is the decisive signal. Two of eight units sit at 80% under the same lecture. Same education, different results means the difference is local workflow, leadership, or process design.
• Repeat lactate at 39% is the classic "nobody owns the second draw" problem - a process design issue, not a memory issue.

Rough estimate: 20-30% knowledge and role clarity, 70% system. Before designing anything, do a rapid process map on one high-performing and one low-performing unit, observe five cases, and check four things: the order set, lactate turnaround time, whether a nurse-driven protocol exists, and fluid availability.

If you launch training first, you will get a small initial bump, the system barrier will reassert, and education will be blamed for a failure it could not prevent.

Intervention (contingent on the diagnosis above):

• Week 1: process mapping; revise the order set; clarify nurse-driven lactate draw and fluid initiation under protocol; build a repeat-lactate trigger (a scheduled second draw at 2 hours).
• Weeks 2-3: 15-minute case-based microlearning in the existing staff meeting slot, focused only on the discordant steps, using two local cases and an explicit escalation path. One-page job aid at every workstation.
• Weeks 3-6: physician champion rounds with the hospitalists; nurse-led huddle prompts; weekly unit-level feedback of each unit's own compliance data.

Evaluation, mapped to Kirkpatrick:

• Level 1, reaction: two questions after each session.
• Level 2, learning: five case vignettes covering who can order, when to repeat, and how to escalate.
• Level 3, behavior: chart audit of the four steps at 30 and 60 days, plus direct observation of the first three sepsis patients per nurse.
• Level 4, results: bundle compliance, time-to-antibiotics, and mortality or length of stay as a trend only.

State the expected effect up front: 54% to roughly 70-75% from education plus order-set change - not 85% from a 30-minute talk. Promising 85% would guarantee that a successful project is judged a failure.

Start from their concern rather than from the guideline. "Cookbook medicine" usually means: I have been burned by a protocol applied to a patient it did not fit. So ask which patients worry them. The answer is almost always heart failure, end-stage renal disease, and the 45 kg elder.

Then bring data, not authority: our sepsis mortality, our fluid-overload cases, and the unit-level variation. Often both harms are real, and the conversation shifts from whether to follow the bundle to how to document clinical judgement within it.

Practical moves:

• Co-present with the physician champion; peers persuade peers.
• Show the pathway for documented modification of fluids, so the protocol is not mindless.
• Present the data unit by unit; physicians respond to their own numbers.
• Ask them to own one driver - early antibiotics or the repeat lactate - so they are co-designers.
• If resistance persists, take it to the sepsis committee as a quality issue with the data attached. Do not let it become an educator-versus-physician dispute, and do not accept the framing that education alone will close a 31-point gap.