Sample Size for Clinical Audit
For a clinical audit targeting 95% confidence with ±5% margin of error, you should audit 100-250 observations depending on the expected complication or non-compliance rate, with 150 observations being optimal for most scenarios with moderate prevalence (around 20%).
Evidence-Based Sample Size Determination
Core Principles for Audit Sample Sizing
The precision of your audit estimate improves with larger sample sizes and is influenced by the underlying prevalence of the outcome you're measuring 1. The relationship between sample size and precision follows these patterns:
- For low prevalence outcomes (10%): Audit sizes of 100-150 observations provide adequate precision 1
- For moderate prevalence outcomes (20%): Precision is not meaningfully improved beyond 150 observations (95% CI: 13.9%-27.3%) 1
- For high prevalence outcomes (50%): Precision plateaus after 200 observations (95% CI: 42.9%-57.1%) 1
Practical Application
A study of 2,274 peripheral intravenous catheter audits over 5 years with complication rates of 21% demonstrated that audit sizes of 100-250 observations per round achieved the desired 95% confidence level with ±5% margin of error 1. Individual audit rounds ranged from 69-220 observations, with complication prevalence varying from 7.8% to 39% 1.
Critical Consideration: Cluster Effects
When auditing across multiple clinical areas or wards, focus on sampling a larger number of wards with fewer observations per ward, rather than many observations from few wards 2. This approach accounts for the cluster effect—infection control practices and clinical behaviors tend to be similar within the same clinical area, which can bias your results if you oversample from limited locations 2.
Determining Your Specific Sample Size
Your sample size calculation should be based on three key factors 2, 3:
- Desired precision level (your ±5% margin of error)
- Expected compliance or complication rate (estimate from previous audits or literature)
- Acceptable confidence level (your 95% confidence)
The conventional approach uses Type I error (α) of 5% or below and power of 80% or above when calculating sample sizes 4. However, for clinical audits focused on estimating prevalence rather than testing hypotheses, the focus shifts to precision of the estimate 5.
Frequency and Timing Considerations
The reference period and audit frequency should match your clinical context 6. Most successful audit systems use:
- Monthly audits for high-volume, rapidly changing practices 6
- Quarterly audits for stable processes with moderate patient volumes 6
- Real-time or near real-time data when available, as this enhances credibility and actionability 7
Common Pitfalls to Avoid
- Insufficient sample size: Audits with fewer than 100 observations often lack precision for meaningful conclusions, particularly at moderate prevalence rates 1
- Ignoring cluster effects: Sampling heavily from one or two wards creates bias and reduces generalizability 2
- Annual audits: Systems providing only annual feedback showed no significant improvements in patient care 7
- Oversampling: Beyond 250 observations, you gain minimal additional precision while substantially increasing workload 1