Quantifying diplopia with a questionnaire.

PURPOSE: To report a diplopia questionnaire (DQ) with a data-driven scoring algorithm.

DESIGN

Cross-sectional study.

PARTICIPANTS

To optimize questionnaire scoring, 147 adults with diplopic strabismus completed both the DQ and the Adult Strabismus-20 (AS-20) health-related quality-of-life (HRQOL) questionnaire. To assess test-retest reliability, 117 adults with diplopic strabismus. To assess responsiveness to surgery, 42 adults (46 surgeries).

METHODS

The 10-item AS-20 function subscale score (scored 0-100) was defined as the gold standard for severity. A range of weights was assigned to the responses and the gaze positions (from equal weighting to greater weighting of primary and reading). Combining all response option weights with all gaze position weights yielded 382848 scoring algorithms. We then calculated 382848 Spearman rank correlation coefficients comparing each algorithm with the AS-20 function subscale score.

MAIN OUTCOME MEASURES: To optimize scoring, Spearman rank correlation coefficients (measuring agreement) between DQ scores and AS-20 function subscale scores. For test-retest reliability, 95% limits of agreement and intraclass correlation coefficient (ICC). For responsiveness, change in DQ score.

RESULTS

For the 382 848 possible scoring algorithms, correlations with AS-20 function subscale score ranged from -0.64 (best correlated) to -0.55. The best-correlated algorithm had response option weights of 5 for rarely, 50 for sometimes, and 75 for often, and gaze position weights of 40 for straight ahead in the distance, 40 for reading, 1 for up, 8 for down, 4 for right, 4 for left, and 3 for other, totaling 100. There was excellent test-retest reliability with an ICC of 0.89 (95% confidence interval, 0.84-0.92), and 95% limits of agreement were 30.9 points. The DQ score was responsive to surgery with a mean change of 51 ± 34 (P<0.001).

CONCLUSIONS

We have developed a data-driven scoring algorithm for the DQ, rating diplopia symptoms from 0 to 100. On the basis of correlations with HRQOL, straight-ahead and reading positions should be highly weighted. The DQ has excellent test-retest reliability and responsiveness, and may be useful in both clinical and research settings.