CLABSI SIR Rise After a Midline Initiative
The infection preventionist reports that the central line-associated bloodstream infection (CLABSI) standardized infection ratio (SIR) rose from 0.7 to 1.6 over two quarters. The raw CLABSI count went from 3 to 4. Central line days fell from 1,850 to 1,290 after a midline and peripheral IV initiative. Two of the four infections were in ICU patients with lines in place more than 14 days; one was in a patient transferred from a long-term care facility. Surveillance definitions were unchanged. The CMO wants to 'pause the midline program' immediately. The board quality committee asks whether care has gotten worse. The IP has line days by unit and month in a spreadsheet but no control chart.
1. Has care gotten worse? Reason through the count, the rate, and the SIR before anyone acts.
Read the worked answer
Not necessarily - and the three numbers tell different stories.
- Count: 3 to 4, a difference of one infection. That is fragile.
- Rate per 1,000 line days: 3/1,850 × 1,000 = 1.62; 4/1,290 × 1,000 = 3.10. The rate roughly doubled because the denominator fell about 30%, not because infections surged.
- SIR: SIR is observed divided by predicted. If predicted infections fell with line days, the same small observed count produces a larger ratio. An SIR of 0.7 to 1.6 can occur with one additional event when predicted counts are low.
- Small numbers: With 4 events, a 95% Poisson interval is roughly 1.1 to 10.2; with 3 events, roughly 0.6 to 8.8. The intervals overlap heavily, so the change is not statistically distinguishable from noise.
The honest answer to the board is: 'The rate is higher, but with four events we cannot yet tell whether this is a real change.' That is different from 'care is worse,' and it is different from 'nothing to see.'
2. What analysis and display should you produce before recommending any change to the midline program?
Read the worked answer
Build a signal-detection package, not a single ratio.
- u-chart of CLABSI per 1,000 line days with control limits, annotated with the midline initiative start. Apply standard special-cause rules; one point above a limit in a low-count process is weak evidence.
- Verify the denominator. Confirm line days are complete and that midlines are excluded from CLABSI denominators. A denominator shift that is really a data-capture change will mimic a rate increase.
- Stratify by unit, line type, and dwell time. Two long-dwell ICU infections and one long-term care transfer suggest a focus on maintenance and insertion source, not on midline use.
- Case review using a standardized tool for all four events: insertion bundle, dressing and hub care, dwell time, and whether the line was still indicated.
- Device utilization ratio to see whether central line use fell as intended.
Recommendation before the CMO acts: do not pause a program that may be reducing central line exposure. Instead, target long-dwell lines and review the three high-risk cases. Set an explicit trigger for escalation, such as two consecutive points above the control limit or any unit cluster.
3. How do you explain the apparent conflict to the CMO and the board without either minimizing or alarming?
Read the worked answer
Lead with the mechanism, then the uncertainty, then the decision.
- Mechanism: 'Our infection count went from three to four. Because we successfully reduced central line days, the rate per line day rose even though exposure fell. The SIR magnifies that effect when predicted counts are small.'
- Uncertainty: Show the confidence interval or the control chart. Say explicitly that four events cannot distinguish a real shift from random variation.
- Decision: 'We are not pausing the midline program today because the data do not show a special-cause signal and the program reduces central line exposure. We are reviewing every case and watching long-dwell lines. If we see a cluster or two consecutive points above the limit, we will escalate.'
Give the board a one-page visual with count, rate, and SIR side by side, and a plain-language caption. Offer a follow-up date. This approach preserves credibility: you neither defended a number nor overreacted to one.