Quantifying the variability of optimal cutpoints and reference values for diagnostic measures
- Subject Areas
- Clinical Trials, Epidemiology, Psychiatry and Psychology, Public Health, Statistics
- optimal cutpoints, reference values, normal ranges, boostrapping, diagnostic research, variability
- © 2014 Hirschfeld et al.
- This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ PrePrints) and either DOI or URL of the article must be cited.
- Cite this article
- 2014. Quantifying the variability of optimal cutpoints and reference values for diagnostic measures. PeerJ PrePrints 2:e635v1 https://doi.org/10.7287/peerj.preprints.635v1
Aim: Empirical studies in medicine – and most other fields of study - yield results that are uncertain to a certain degree. Medical research on interventions has made tremendous leaps forward by quantifying and reporting this uncertainty using p-values and confidence intervals. In contrast to this, most diagnostic studies that determine optimal cutpoints or reference values for diagnostic measures ignore that their outcomes, i.e. the specific cutpoints or normal ranges they recommend, are subject to chance variability. Methods: In this paper we use a simple simulation approach to quantify the variability of optimal cutpoints for two published studies. The first determined an optimal cutpoint for Becks Depression Inventory (BDI) in adults. The second determined reference values for Quantitative Sensory Testing (QST) in children. Results: We find that frequently employed cutpoints to interpret BDI scores and QST results are highly variable. For the BDI we find that replication of this study may identify values between 14 and 21 as optimal cutpoints. The lower cutpoint results in a misclassification of 15% of the healthy adults as depressed, the upper cutpoint results in a misclassification rate of 2%. For the QST we find that the upper end of the normal range HPT varies between 46.9 and 50.2 degrees Celsius. Conclusions: Based on our results we argue that researchers should be required to estimate and report the variability of reference values and optimal cutpoints for diagnostic tools. This may improve the harmonization of findings across studies and provides a rationale for planning future studies.
This is a submission to PeerJ for review.