Diagnostic accuracy in the medical intensive care unit
© The Author(s) 2001
Published: 2 March 2001
To evaluate the frequency of diagnostic errors assessed by autopsies.
Retrospective review of medical and pathological records.
An 11-bed medical ICU at a 860-bed General Hospital.
Patients who died in ICU between January 1998 and December 1999.
Pre mortem diagnosis, length of stay and presence of chronic disease were determined from medical records. Medical diagnosis were rated into three levels of clinical diagnostic certainty as complete ('golden standard') certainty (Group L1), minor diagnostic uncertainty (Group L2), and major diagnostic uncertainty (Group L3). Autopsy results were obtained from final pathology report. A panel of three intensivists reviewed the findings and distributed patients into three error groups: Group A – the autopsy confirmed the clinical diagnosis (fully correct diagnosis), Group B – the autopsy demonstrated new active diagnosis, which would have probably not influenced the therapy (non-fatal diagnostic error), Group C – the autopsy demonstrated new active diagnosis, which would have probably changed the therapy (fatal, but potentially treatable error).
The overall mortality in the treated population was 20.3% (270/1331 patients). Autopsies were performed in 126 patients (46.9% of deaths), more often in younger patients (66.6 ± 13.9 years vs 72.7 ± 12.0 years, P < 0.001), in patients staying less then 24 h in ICU stay (4.7 ± 5.6 days vs 6.7 ± 8.7, P = 0.0549) and in patients in Group L3 without chronic diseases (15 vs 1, P < 0.001). According to pre mortem clinical level of diagnostic certainty 46.0%, 24.6% and 29.4% of patients were in group L1, L2 and L3 respectively (ns between groups). After analysis of pathological findings, fully correct diagnosis (group A, 60 patients [47.6%]) was found in 60.3%, 40.0%, 34.2% of patients in group L1, L2 and L3 respectively (ns between groups). Non-fatal diagnostic errors (group B, 54 patients [42.9%]) were found in 31.0%, 50%, 55.3% of patients in group L1, L2 and L3 respectively (ns between groups). Fatal, but potentially treatable errors (group C, 12 patients [9.5%]) were found in 8.7%, 10.0% and 10.5% of patients in Group L1, L2 and L3 respectively (ns between groups). ICU length of stay shorter than 24 h was not related with frequency of group C errors.
The autopsies are performed more often in younger patients without chronic disease, in patients with shorter ICU stay and in cases with low clinical diagnostic certainty. No level of clinical diagnostic certainty could predict the pathological findings. Autopsy remains the essential verification of clinical diagnostic certainty in critically ill patients.