Can we quantify harm in general practice records? An assessment of precision and power using computer simulation

Area of Science:

• Patient safety research
• Health services research
• Clinical informatics

Background:

• Improving patient safety requires accurate estimation of harm rates in specific populations and detection of changes over time.
• Clinical record review (CRR) is a suitable method for this, but optimal parameter settings are unknown.
• This study aimed to quantify CRR parameters, assess feasible CRR scenarios, and determine minimum requirements for precision and power.

Purpose of the Study:

• To determine and quantify parameters for clinical record review (CRR).
• To assess the precision and statistical power of feasible CRR scenarios.
• To quantify the minimum requirements for adequate precision and acceptable power in harm rate monitoring.

Main Methods:

• Monte Carlo simulation was used to explore CRR scenarios.
• 864 different scenarios were tested with 1000 random data sets each.
• Generalized linear models with a Poisson response were used to estimate harm rates and test for changes.

Main Results:

• CRR scenarios detecting at least 100 harm incidents yielded precise harm rate estimates.
• Detecting 20% or 50% harm reductions required CRR scenarios with at least 100 or 500 harm incidents, respectively.
• The number of detected harm incidents depends on baseline harm rate, review period, records per practice, number of practices, and review frequency.

Conclusions:

• A formula was developed to calculate minimum CRR parameter values for adequate precision and power in monitoring harm rates.
• Findings offer practical guidance for healthcare decision-makers, leaders, and researchers.
• This research supports efforts to measure and reduce harm at regional and national levels.