Real-world implementation — completion and positive-score rates
Completion and positive-score rate are implementation measures. They describe uptake and yield, not diagnostic accuracy or patient benefit; without a contemporaneous reference standard and case-mix information, they cannot establish how many cases were missed. This page summarises the available large-scale evidence.
The key review
Most of what is known comes from one systematic review: Penfold et al. (2024), which synthesised completion and positive-score rates for validated delirium detection tools used at scale (n ≥ 1,000) in general hospitals — 22 research studies and 4 audit reports covering six tools (CAM, 4AT, DOSS, bCAM, NuDESC, ICDSC). Professor MacLullich is a co-author. The search ended on 31 December 2022, so this is an evidence snapshot. PubMed · DOI
What the evidence shows
| Tool | Completion in routine use | Positive-score rate | Real-world signal |
|---|---|---|---|
| 4AT | High in most large studies | Admission 13–20%; post-op 21–28% | Interpret with case mix and timing |
| CAM | Up to 99% in dedicated programmes; often lower | Admission 8–51%; inpatient 2–20% | Some rates below published prevalence estimates; several explanations are possible |
| DOSS | Low alone; higher in combination | Inpatient 6–42% | Broadly aligned except in one model-development study |
| NuDESC | Reported without denominator in some studies | Inpatient 5–13% | Often below expected prevalence |
| bCAM | 12% (labour/delivery) to 98% (medical), phased | — | Setting and implementation stage both differed |
| ICDSC | Varies by setting/stage | — | General-hospital and ICU use |
Ranges are across different populations, care settings, timepoints and implementation stages, and are not directly comparable between tools. Overall completion rates across all studies ranged from 19% to 100%. Most included studies were at moderate–high risk of bias. Source: Penfold et al., J Am Geriatr Soc 2024;72(5):1508-1524.
Feasibility is distinct from accuracy
- A tool that is not completed cannot contribute to detection, but completion does not measure sensitivity. Validation accuracy and implementation uptake answer different questions.
- Compare local positive-score rates cautiously. Account for population, setting, timing and selective completion. A low rate is a reason to audit practice, not proof of under-detection.
- Interpret implementation reports in context. In the bCAM report, completion ranged from 12% in labour/delivery to 98% on medical services during phased implementation; setting and implementation stage both differed.
- Report your data. The review's central call is for health systems to analyse and publish their own completion and positivity data — the only way to understand and improve detection at scale.
Use this when choosing a tool
On the Choose a tool page and on each tool's card, weigh not only published accuracy but whether the tool can realistically be completed, every time, by the staff who will use it. Each tool's card shows a Real-world use note where implementation data exist.