Subjective assessment of frequency distribution histograms and consequences on reference interval accuracy for small...

Area of Science:

• Statistics
• Biostatistics
• Clinical Laboratory Science

Background:

• Estimating reference intervals (RIs) accurately is challenging, particularly with limited sample sizes.
• Inaccurate RIs can lead to misdiagnosis and inappropriate patient management.

Purpose of the Study:

• To identify optimal statistical methods for estimating RIs based on sample size and population distribution.
• To evaluate the accuracy of sample frequency distribution histograms in representing population distributions.
• To compare RI estimation strategies using histograms versus goodness-of-fit tests.

Main Methods:

• Statistical methods (parametric, nonparametric, robust) were evaluated for accuracy across various sample sizes (n=20-60) and distributions (Gaussian, log-normal, left-skewed).
• Frequency distribution histograms were generated from simulated populations, and reviewers classified distributions based on visual assessment.
• RI accuracy was compared between histogram-based strategies and goodness-of-fit tests.

Main Results:

• Parametric, nonparametric, and robust methods showed improved accuracy for lower and upper reference limits depending on distribution type.
• Sample histograms accurately classified population distributions between 71% and 93.9% of the time.
• Histogram-based strategies, particularly when assessed by a statistician, were significantly more accurate and precise than goodness-of-fit tests (P < 0.001).

Conclusions:

• A strategy utilizing histograms can enhance the accuracy of reference interval estimations.
• Inter-reviewer variability in histogram interpretation was observed and requires further investigation.
• Future research should explore factors influencing histogram interpretation to standardize its application.