Methods to facilitate the interpretation of pooled diagnostic test accuracy estimates by means of selecting a representative pre-test probability

ID: 

110

Session: 

Poster session 1

Date: 

Sunday 16 September 2018 - 12:30 to 14:00

All authors in correct order:

Oerbekke M1, van Enst A1, Jenniskens K2, Scholten R3, Hooft L3
1 Knowlegde Institute for Medical Specialists, Utrecht, The Netherlands
2 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, The Netherlands
3 Cochrane Netherlands, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, The Netherlands
Presenting author and contact person

Presenting author:

Michiel Oerbekke

Contact person:

Michiel Oerbekke
Abstract text
Background:
Test results alone are not informative for clinicians unless they can estimate the prevalence in their setting. Diagnostic test accuracy (DTA) reviews facilitate the interpretation of the pooled test performance in a hypothetical cohort using a pre-test probability. However, the methods used in DTA reviews to select the target condition's pre-test probability are unknown.

Objectives:
To assess the methods used in Cochrane DTA reviews for selecting a pre-test probability to demonstrate a test's performance using summary sensitivity and/or sensitivity.

Methods:
We selected DTA reviews from the Cochrane Library on 2 February 2018. Reviews were eligible when a pooled or summarised accuracy measure was provided. Data were extracted by one author and checked by a second author.

Preliminary results:
From 81 DTA reviews 59 were eligible (307 meta-analyses). We observed the following methods: using a point estimate (median (62 analyses), mean (11 analyses)), using a point estimate and a measure of dispersion (median and range (1 analysis), mean and range (1 analysis), median and lower/upper quartile (3 analyses), median and lower/upper quartile and range (5 analyses)), using literature (studies reporting prevalence (15 analyses), WHO suggestion (10 analyses), guideline (4 analyses)), using an assumption (27 analyses), or using an unclear or partially unclear method (12 analyses).

Preliminary conclusions:
This is still an ongoing study and updated results and conclusions will be presented during the conference. No consensus currently exists on what method should be used to select a representative pre-test probability. However, it is probably more informative to use multiple pre-test probabilities from data included for analyses (e.g. a point estimate and a measure of dispersion). Multiple pre-test probabilities could facilitate the test's performance interpretation for clinicians in their own practice.

Patient implications:
With the presentation of only one pre-test risk in a hypothetical cohort clinicians may be incompletely informed, which might lead to a higher chance of patients being misdiagnosed. Multiple pre-test risks in hypothetical cohorts may provide clinicians with more information to judge a test's applicability in their practice, potentially safeguarding patients from misdiagnoses.

Relevance to patients and consumers: 

With the presentation of one pre-test risks in a hypothetical cohort, clinicians may be incompletely informed which might lead to a higher chance for patients being misdiagnosed (i.e. incorrect diagnosis [false positives] and/or missed correct diagnosis [false negatives]). It may underestimate or overestimate the test’s diagnostic capability, depending on the disease prevalence in the setting where the test is utilized in practice. Multiple pre-test risks in hypothetical cohorts may provide clinicians more information to judge the test’s applicability in their practice, potentially safeguarding patients for misdiagnoses.