The Rapid Response Team's Significant Quarter
A hospital quality-improvement team reports that its new rapid response team reduced inpatient mortality. Before the team launched, 4,000 consecutive admissions had 128 deaths (3.2%). During the first full quarter after launch, 4,200 admissions had 109 deaths (2.6%). A chi-square test gives p = 0.04, and the team concludes that the rapid response team saved lives. During the same quarter, the hospital opened a 12-bed step-down unit, introduced a sepsis alert in the electronic record, expanded palliative care consultations, and changed the code-status documentation form. The case mix also shifted: the post-launch group had more elective surgical admissions and fewer admissions from nursing homes. No risk adjustment was performed. The team wants to submit the result to the quality committee as proof of effectiveness and use it to justify hiring more rapid-response nurses.
1. Calculate the absolute risk reduction and number needed to treat. Interpret both in plain language.
Read the worked answer
Before-launch mortality was 128/4,000 = 0.032, or 3.2%. After-launch mortality was 109/4,200 = 0.026, or 2.6%. The absolute risk reduction is 3.2% - 2.6% = 0.6 percentage points, or 0.006 as a proportion. The number needed to treat is 1 / 0.006 = about 167 admissions. In plain language, the hospital would need to expose roughly 167 admissions to the rapid response team, or to the entire package of changes happening in that quarter, to prevent one death. That estimate is fragile because the confidence interval around the difference is wide. Using the normal approximation, the standard error of the difference is about 0.0037, so the 95% CI for the absolute difference is roughly -0.1 to 1.3 percentage points. The lower bound includes no benefit, and possibly harm. A p-value of 0.04 is borderline, and the NNT of 167 is not automatically clinically impressive without context about cost, harms, and whether the mortality difference is real.
2. Explain why p = 0.04 does not establish that the rapid response team caused the mortality difference. Identify at least three threats to validity.
Read the worked answer
The p-value addresses only whether the observed difference is unlikely under a null hypothesis of no difference, assuming no bias and no confounding. It does not establish causality. First, secular trend is a major threat: mortality may have been falling for reasons unrelated to the rapid response team. Second, co-interventions occurred at the same time: a step-down unit, a sepsis alert, expanded palliative care, and a new code-status form could each affect mortality. Third, case-mix changed: more elective surgical admissions and fewer nursing-home admissions would lower expected mortality. Fourth, ascertainment and documentation changed: a new code-status form and palliative care expansion may have increased withdrawal of life-sustaining treatment or changed how deaths were recorded. Fifth, regression to the mean is possible if the pre-launch period was unusually bad. Sixth, multiple comparisons and selective reporting may inflate the significance of one favorable result. The team should not present the before-after comparison as proof of effectiveness; it is a signal that requires a stronger design.
3. What analysis or design would provide stronger evidence for the quality committee?
Read the worked answer
A stronger design would be an interrupted time series with at least 12 months of pre-implementation and 12 months of post-implementation data, modeling the pre-existing trend and looking for a change in level or slope after the rapid response team launched. If another comparable hospital or ward did not implement the team, a difference-in-differences analysis would help control for secular trends. The analysis should risk-adjust for age, comorbidity, admission source, surgical versus medical status, and do-not-resuscitate status, and it should pre-specify the primary outcome. The team should also measure process outcomes: activation rates, time to activation, transfers to higher care, and unplanned ICU admissions. It should track potential harms such as inappropriate activations or resource strain. If the team wants a causal claim, a stepped-wedge cluster design across wards could be considered, but it must be planned in advance. For the current data, the committee should treat the result as preliminary and require a more rigorous evaluation before expanding staffing on the basis of this quarter alone.