The Fall Rate That Tripled Overnight
A 32-bed medical unit's fall rate jumped from 2.4 to 4.1 falls per 1,000 patient days between Q3 and Q4 against a benchmark of 3.0. Three things changed at once: a new mobile incident-reporting app let any staff member file in seconds, the quality department adopted a broader definition capturing "falls with any assistance," and the unit's case mix shifted toward observation patients (mean length of stay 4.8 to 3.1 days). The unit manager has already drafted a poster campaign and wants it launched Monday.
1. Before any intervention, how do you determine whether the increase is real?
Read the worked answer
I would hold the poster and interrogate the number, because three simultaneous changes map onto three known artifacts:
- Definition change - a broader case definition mechanically raises the numerator. I would re-extract Q4 events and re-classify them under the old definition. If the rate returns to ~2.4, the "increase" is a measurement change, not a safety change.
- Reporting-culture change - a faster reporting app increases capture of near-misses and minor events that were previously invisible. This is good; it means the old rate was falsely low. Rising rates after a reporting improvement are a sign the system started telling the truth.
- Denominator and case mix - shorter length of stay toward observation patients changes both patient-days and acuity mix. Falls per 1,000 patient days is sensitive to the denominator; a unit with high turnover has more admission-related fall risk per day.
The test I would actually run: stratified rate by fall type (injurious vs non-injurious). Falls with injury is the least artifact-prone measure, because injury is hard to over-report and clinically meaningful. If injurious falls are flat while total falls rose, the story is surveillance, not deterioration.
Conclusion to state: I would not launch a campaign against a number I cannot defend - and I would not dismiss it either. The honest statement to staff is: "We may be counting better. Here is what we can and cannot say."
2. Build a measurement family for a fall-reduction initiative: one outcome, two process, one balancing measure, each with its rationale.
Read the worked answer
Outcome - injurious falls per 1,000 patient days. Chosen over total falls because it is resistant to reporting-culture artifacts and is what patients actually care about. Reported monthly on a run chart with the denominator stated.
Process 1 - proportion of patients with a completed fall-risk assessment within 4 hours of admission and reassessed after any change in condition or new sedating medication. Rationale: it measures the reliability of the step that generates the plan. A fall-risk score that is documented once and never updated is a ritual, not a risk assessment.
Process 2 - proportion of high-risk patients with at least one documented tailored intervention matched to their specific risk factor (e.g., hourly rounding for delirium risk, early mobility for deconditioning, toileting schedule for urgency). Rationale: this measures whether the plan is individualized. "Fall precautions initiated" is a checkbox; matching an intervention to a mechanism is the actual work.
Balancing measure - restraint use and/or sedative administration rate. This is the critical one. Falls can be driven to zero by chemically or physically immobilizing patients, and that would look like success on the outcome measure while causing delirium, pressure injury, and functional decline. Naming the balancing measure in advance is a commitment not to buy the outcome with hidden harm.
Design note: I would state the target and timeframe before the first data point (e.g., 25% reduction in injurious falls over 6 months) so the team cannot move the goalpost after seeing the chart.
3. A staff nurse returns from a conference and asks to implement the "falls bundle" the speakers presented, which reduced falls 40% at a large academic center. How do you decide?
Read the worked answer
I would not say yes or no - I would ask four questions in order.
- What is the evidence quality? A 40% reduction at one academic center is a single-site result. Was it a controlled study or a before-after project? A before-after design cannot separate the bundle from the attention and monitoring that accompanied it (the Hawthorne effect), which is exactly what this unit just experienced with its reporting change.
- What exactly is in the bundle, and which components have independent support? Bundles often contain five elements where two carry the evidence and three carry the workload. Implementing all five dilutes reliability - you get five things done at 60% instead of two done at 95%. I would identify the components our own stratified data say address our fall mechanisms.
- Does it fit this context? Their unit's mean length of stay and observation-patient mix differ from ours. A bundle built around prolonged delirium prevention may not map onto a population that turns over in 3 days.
- What would it displace? New work consumes staff attention and time. If we add this, what documentation or rounding practice gets less reliable?
Decision rule: adopt the components with the strongest independent evidence, pilot them on one pod with a PDSA cycle, and compare against the process and balancing measures already defined. If the pilot shows process reliability ≥90% and no rise in restraint or sedative use, spread it. If it shows a chart improvement and a flat outcome, we learned something - and that is still a win, provided we were honest enough to look.