Clinical Research Oriented Workshop (CROW) Meeting: September 3, 2015

Start Up: End of summer stories…

1.                  Discussion: Kairn Kelley on Dichotic Listening Statistics

a.       A foundational understanding in auditory processing includes:

                                                  i.      Low left ear score: integration problem (corpus callosum) or right hemisphere problem; this is where the problems usually lie

                                                ii.      Low left and right ear score: auditory cortex problem

                                              iii.      Low right ear score: undefined; may be that language is localized in the left hemisphere instead of the right (unusual) or something else is happening

                                              iv.      So the score of each ear is important, but for different reasons

1.      The difference between scores is also important

b.      The confidence interval of joint distribution is described here as the Joint CI

                                                  i.      Presented on a Cartesian plane, the Joint CI can be diagrammed as a diamond shaped octagon

1.      94% of scores should fall within this octagon; Kairn’s data demonstrate this

                                                ii.      Recommendation: plot 6 diagrams for each child (6 * 60 = 360 diagrams), color coded for how close or beyond the threshold for each test/kid

c.       Purpose: how to answer the questions

                                                  i.      How reliable are the raw scores (not just the left ear scores)

1.      No less reliable than the binomial distribution says it should be

2.      Straight-up reliability

                                                ii.      What is the impact of this randomness on clinical decision making (Note: the approved topic is: how well are tests in agreement with each other – Kairn now knows they are different)

1.      Does “ear advantage” (dominance) show up the same on each of the three tests?  (Not true, based on Kairn’s data)

a.       Dominance within a test was always the same

b.      Dominance between tests flipped, usually to “indeterminate” but sometimes to opposite side

                                              iii.      This analysis can lend itself to adaptive testing, to increase “n” tests in the ear that appears to be underperforming in order to be clear about the clinical threshold

d.      Summary:

                                                  i.      Capture reliability with CI

                                                ii.      CI, in retest, overlap substantially

                                              iii.      CI, across different tests, may have different or little overlap

1.      Each test is converted to a % of correct answers (so all have the same scale)

2.      Center points should be the same, size of zone will change by number of items in test

                                              iv.      Develop the CI for each possible “indeterminate” score and plot to create ranges of scores that will fall into clinically distinct zones; repeat for “test size,” i.e. number of items on a test

                                                v.      Future: measure children with known auditory processing disorders (using existing data)

2.                  Next Workshop Meeting(s): Fridays, 12:00 p.m. – 1:00 p.m., at Given Courtyard South Level 4.   

a.       September 10: Marianne – abstract submission for presentation on methodology

b.      September 17: (no Kairn)

c.       September 24: Kairn – update

d.      October 1:

e.       Future: Connie’s Pfizer application and reviewer comments

Recorder: Connie van Eeghen