Standard setting Determining the pass mark - OSCEs
The Old Way (1) …..I think that the pass mark in this is exam is probably about here….. there is a natural break I can see dividing the bottom from the rest
The Old Way (2) …..The pass mark is 60%.
Assessments
Match Outcome Objectives
Assessments Match Outcome Objectives Integrated
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout Progressive testing
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout Progressive testing House style
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout Progressive testing House style From approved list
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout Progressive testing House style From approved list Student information
Assessments Match Outcome Objectives Integrated Clinical Competencies tested throughout Progressive testing House style From approved list Student information External Examiners
Standard setting KnowledgeSkills
Angoff Hofstee Borderline method KnowledgeSkills Standard setting
Borderline procedure Used for clinical (OSCE) examinations
Borderline procedure Used for clinical (OSCE) examinations Examiners score the students’s performance at the station e.g 17/20 Examiners judge the overall performance clear pass / borderline / clear fail Mark sheets rated borderline identified and the scores of borderline students averaged Process repeated for each station Calculate the median borderline score for all stations
Borderline Method
A 1 A 2 A 3 Pass/Fail Station 15 (static) Oral Lesions Score Count AAAA A A A AAAAAAAAAA A A A A A
Borderline group pass fail Median Borderline score 58.41
Borderline group pass fail Median Borderline score How wide should this band be?
Borderline group pass fail Median Borderline score How wide should this band be? +/- 1 - standard deviation - standard error - or what?
The Standard Error of Measurement depends on the reliability of the test (R) depends on the standard deviation of the test (SD) SEM = SD 1 - R acts as a confidence interval in high stakes situations
The Standard Error of Measurement depends on the reliability of the test (R) depends on the standard deviation of the test (SD) SEM = SD 1 - R acts as a confidence interval in high stakes situations R = Variance Variance x Error
Borderline group Pass Fail Median Borderline score Pass Score } - 1 SEM Fail score } + 1 SEM
Number of students Mean Score (%) Overall Borderline Score (%) Standard deviation Standard Error of Measurement Pass mark (%) Reliability
Internal Reliability of the Exam Cronbach’s Alpha Cronbach’s alpha Shows whether randomly split halves of the exam by item vary together Split half Correlations (SH) Item / Stations Population (students)
Borderline Grading System Borderline Fail Pass Good Pass Top Mark Borderline Mark Distinction Pass Mark Pass Fail
Reducing measurement error 1. Increase reliability of exam 2. Compromise with feasibility/cost * Content specificity * More items in the exam * Blueprinting * Quality assurance of item writing * Training examiners/standardised patients * Feedback from exam performance