Process measures, length of stay, and escalation of care showed no special cause variation after the intervention

Description #

Aside from mortality, the intervention produced no detectable special cause variation. Process measures did not shift: average monthly critical response nurse consults went 5.67/month pre- to 5.45/month post-intervention, and average monthly RR system activations went 11.3/month pre- to 15.7/month post-intervention, neither showing special cause variation. For the main secondary outcomes, the average rate of escalation of care after RR activation actually rose (43.75% or 59/135 pre- vs 45.73% or 102/222 post-intervention) and average post-RR length of hospitalization fell (20.45 days/patient pre- vs 19.52 days/patient post-intervention), but the authors "found no special cause variation that linked these changes to intervention implementation."

"Between the pre- and post-intervention periods, neither the average monthly number of critical response nurse consults for LEP patients (5.67/month pre- to 5.45/month post-intervention) nor the average monthly number of RR system activations (11.3/month pre- to 15.7/month post-intervention) showed special cause variation." (Raff, 2024, p. 1108)

"While LEP patients experienced an average increased rate of escalation of care after RR system activation (43.75% or 59/135 escalated pre- versus 45.73% or 102/222 post-intervention), and decreased length of hospitalization post-RR system activation (an average of 20.45 days per patient pre- versus 19.52 days per patient post-intervention), we found no special cause variation that linked these changes to intervention implementation." (Raff, 2024, p. 1109)

Methods Context #

What? #ⓘ

The observables: monthly critical response nurse consult count, monthly RR activation count (process measures), and monthly average post-RR escalation of care and length of hospitalization (outcome measures), each plotted on pre/post X bar and R control charts.

"Process Measures. Measures to assess intervention activities and fidelity included the monthly number of RR critical response nurse consults; the monthly number of RR system activations; and monthly average DI score at the time of RR system activation." (Raff, 2024, p. 1106)

How? #ⓘ

Quasi-experimental pre-post QI design using statistical process control (X bar and R control charts, Nelson Rules) on monthly cross-sectional aggregates, comparing the period before vs after the May 2022 intervention go-live.

"We also report pre-versus post-average values across all months for process and outcome variables to provide context for analytic findings." (Raff, 2024, p. 1107)

Who? #ⓘ

All adult non-ICU hospitalized LEP patients experiencing an RR system activation at a 950-bed quaternary academic medical center (UNCMC), May 2021–March 2023; 222 individual patients across 302 activations (135 activations / 68 critical response consults pre; 167 activations / 60 consults post).

"In the pre-intervention period (May 1, 2021, to April 30, 2022), there were 135 RR system activations and 68 critical response nurse consultations." (Raff, 2024, p. 1108)

Other Notes #

The directional changes in length of stay and escalation are not statistically supported here; only mortality met special-cause criteria. The authors speculate that stable length of stay despite higher acuity at activation "may suggest positive intervention effects on recovery of LEP RR activation survivors" (Raff, 2024, p. 1110), but emphasize this is unconfirmed.

Caveats #

• Single-center pre-post QI pilot with a small sample and bundled intervention limits causal inference about the mortality effect This is a single-center, quasi-experimental pre-post pilot at one quaternary academic medical center, so findings may not generalize. The population was modest (222 patients; mortality counts of 8 pre vs 7 post) and the project period was constrained. The design supports only limited causal inference: the multi-component "bundle" (EWS dashboard, RR-nurse monitoring, EWS-triggered activation, and interpreter integration) was implemented together, so the mortality decrease cannot be attributed to interpreter integration specifically, and secular trends cannot be ruled out. The authors also could not comprehensively assess implementation fidelity or user perceptions of the intervention.