Using as an experimental model the results obtained in assaying thyroid hormones and tumor markers, two peculiar aspects of the Bayesian analysis are considered: (a) the imprecision associated with the assessment of diagnostic sensitivity, specificity and predictive value, and (b) the implications of the choice of the cutoff point (ie, the medical decision limit). As for the former aspect, still largely disregarded in the literature, no significant difference results when applying binomial statistics or the "bootstrap" resampling technique for uncertainty evaluations; the critical effects on the imprecision extent related to the size of available statistical samples and of the cutoff value are confirmed, no effect being instead related to the extent of distribution skewness. The choice of decision limit (eg, the limit corresponding to the maximum efficiency, to the maximum Youden's index or to a given specificity) obviously proves crucial in establishing the diagnostic value of test results; furthermore, the use of maximum-efficiency cutoff points looks questionable when low prevalences are concerned. Thus, rather than from some rigid criteria, this choice should proceed from a comparative analysis of operational curves obtained for given cutoff values, such as the functions prevalence vs predictive value or prevalence vs misclassification rate.

Critical aspects of bayesian analysis: The imprecision in performance assessing and the choice of decision limits

FERDEGHINI, Marco;
1995-01-01

Abstract

Using as an experimental model the results obtained in assaying thyroid hormones and tumor markers, two peculiar aspects of the Bayesian analysis are considered: (a) the imprecision associated with the assessment of diagnostic sensitivity, specificity and predictive value, and (b) the implications of the choice of the cutoff point (ie, the medical decision limit). As for the former aspect, still largely disregarded in the literature, no significant difference results when applying binomial statistics or the "bootstrap" resampling technique for uncertainty evaluations; the critical effects on the imprecision extent related to the size of available statistical samples and of the cutoff value are confirmed, no effect being instead related to the extent of distribution skewness. The choice of decision limit (eg, the limit corresponding to the maximum efficiency, to the maximum Youden's index or to a given specificity) obviously proves crucial in establishing the diagnostic value of test results; furthermore, the use of maximum-efficiency cutoff points looks questionable when low prevalences are concerned. Thus, rather than from some rigid criteria, this choice should proceed from a comparative analysis of operational curves obtained for given cutoff values, such as the functions prevalence vs predictive value or prevalence vs misclassification rate.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11562/7408
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact