Checking a Virtual-Ward Press Release
A hospital communications officer has drafted a press release: 'Our virtual ward reduced readmissions by 32% and saved 1,200 bed-days. This proves that home monitoring works better than hospital care.' The underlying audit shows 640 patients admitted to the virtual ward over 12 months; 96 were readmitted within 30 days. In the previous year, 610 similar patients received standard care and 120 were readmitted. During the virtual-ward year, the hospital also changed its discharge criteria, moving less sick patients out earlier, and 18% of eligible patients declined the virtual ward. Readmission data were missing for 12% of the virtual-ward group and 4% of the standard-care group. The officer asks you to check the release before publication and to suggest one rewritten sentence that is defensible.
1. Calculate the readmission proportions in both groups. How do the discharge-criteria change and the missing data affect the 32% claim?
Read the worked answer
Virtual ward: 96/640 = 15.0%. Standard care: 120/610 = 19.7%. The relative reduction is (19.7 − 15.0)/19.7 = 23.9%, not 32%. If the missing virtual-ward readmissions were all counted as readmissions, the proportion would be (96 + 0.12 × 640)/640 = (96 + 76.8)/640 = 27.0%. For standard care, (120 + 0.04 × 610)/610 = (120 + 24.4)/610 = 23.7%. Under that worst-case assumption, the virtual ward looks worse, not better.
The discharge-criteria change is a confounding variable: patients in the virtual-ward year were less sick at discharge, so lower readmissions may reflect case mix rather than home monitoring. The 18% who declined may also differ systematically from those who accepted. The 32% figure is not reproducible from the provided numerators and denominators, so it should not be published without the underlying calculation.
2. Identify the unstated assumption in 'This proves that home monitoring works better than hospital care' and explain the ambiguity in 'works better'. Rewrite the sentence defensibly.
Read the worked answer
The unstated assumption is that the two groups were otherwise comparable and that the only relevant change was the virtual ward. The audit shows that discharge criteria changed, that a substantial minority declined the virtual ward, and that missing data were unequal. Without adjustment, the comparison cannot isolate the effect of home monitoring. 'Proves' is too strong for an uncontrolled before-and-after audit; even a well-conducted trial would usually support 'was associated with' rather than 'proves'.
'Works better' is ambiguous because it could mean fewer readmissions, fewer bed-days, lower cost, better patient experience or lower mortality. The release cites readmissions and bed-days but not the other outcomes. A defensible rewrite is: 'In a before-and-after audit, patients managed on the virtual ward had a lower unadjusted 30-day readmission proportion than patients treated the previous year (15.0% vs 19.7%); these findings are observational and may be affected by changes in discharge criteria and missing data.'
3. What additional analysis or wording would make the release acceptable to a clinical governance committee? State two priorities.
Read the worked answer
First, the analysis should be adjusted or stratified for case mix, especially the discharge-criteria change. A propensity-matched comparison, a difference-in-differences design, or at minimum a table of age, comorbidity, acuity and social factors would show whether the groups were comparable. The committee should also see a sensitivity analysis that treats missing readmissions in different ways, because the direction of the effect changes under worst-case assumptions.
Second, the release should report absolute numbers with denominators and avoid causal language. It should state that the design is observational, give the unadjusted proportions, acknowledge the missing data and the 18% decline rate, and describe the virtual ward as 'associated with' the observed readmission proportion. If bed-day savings are claimed, the method of calculation and the uncertainty around it should be reported. With those changes, the release would be honest about what the audit can and cannot show.