ICU readmission as a turnover signal

System failure identification

Patients discharged from intensive care under capacity pressure return within days. The bed was freed and immediately reconsumed, so apparent turnover improved while real capacity did not. Pressure to free beds produces a metric that improves as care worsens.
Data pipeline diagnostics

Turnover and readmission are owned by different reports. Nothing joins a discharge to its own subsequent readmission, so a premature discharge is recorded as two separate successes. Attribution stops at the unit boundary, so the causal link is never drawn.
Predictive computational model

Compute net turnover: discharges minus readmissions within a defined window, attributed back to the discharging shift. The metric then penalises the practice it is meant to detect. Netting the two aligns the measure with the outcome it is a proxy for.
Workflow integration

Report net turnover at the same cadence and in the same view as gross turnover. Presented separately, the more favourable number is the one that gets quoted. Metrics competing for the same decision are resolved by whichever is quoted first.

System failure identification