CUTOFF POINTS AND LIKELIHOOD RATIOS FOR TWO DIAGNOSTIC TESTS FOR PATIENTS WITH ASTHENOPIC SYMPTOMS

Title CUTOFF POINTS AND LIKELIHOOD RATIOS FOR TWO DIAGNOSTIC TESTS FOR PATIENTS WITH ASTHENOPIC SYMPTOMS
Author, Co-Author David Corliss
Topic
Year
1992
Day
Monday
Program Number
Poster 22
Room
Great Hall
Affiliation
Abstract Receiver Operating Characteristic (ROC) curves can be used to determine cutoff points in continuous clinical data by optimizing on the sensitivity (SENS), specificity (SPEC) or efficiency (EFF). Likelihood Ratios (LR+=SENS/(1-SPEC); LR-=(1-SENS)/SPEC) calculated at these cutoffs provide the clinician with quantitative estimates of how much a positive or negative test changes the odds that a patient has a condition. In the present analysis ROC curves were used to evaluate tests of binocular motor function that have been shown to have potential for the differential diagnosis of asthenopia (Daum KM, Rutstein RP, Houston G, et al. Optom Vis Sci 1989;66:218-28). Thirty-two of 100 pseudorandomly selected, paid volunteers were classified as symptomatic based on self-reported severity and frequency of headache, eyestrain, blur, slow focus, or double vision as ranked on a 5-point scale. A patient was considered symptomatic if the product of severity and frequency was >3 for any symptom. Of several findings that were compared on their optimized EFFs, a modified Sheard's amount (MSA), defined as 2x phoria + 6 - vergence to break opposite the phoria, had the best overall characteristics (SENS=0.53, SPEC=0.88, EFF=0.77, LR+=4.5, LR-=0.53, Chi-square=19.9). A MSA>=0 (a positive test) classifies a patient in the symptomatic group. The next best independent test was a measure of the near point of accommodation deficit (NPAD), defined as 100/(observed amplitude - age expected amplitude) (SENS=0.50, SPEC=0.87, EFF=0.75, LR+=3.78, LR-=0.58, Chi-square=15.69). A NPAD>=4 classifies a patient in the symptomatic group. Since the tests are independent, it is possible to multiply the prior odds that the patient's symptoms are asthenopic by the Likelihood Ratios for any combination of positive and negative results on the two tests to determine the posterior odds that a patient's symptoms can be classified as asthenopic. In this study sample for example, the odds of asthenopia incr
Affiliation of Co-Authors
Outline