Decision Trees for Indication of Cataract Surgery Based on Changes in Visual Acuity

2010 
Objective To develop decision trees based on prospectively collected data for determining the appropriateness of cataract extraction. Design Prospective observational cohort study. Participants Consecutive patients with a diagnosis of cataract who are on waiting lists to undergo cataract extraction by phacoemulsification. Methods Patients were randomly assigned to 1 of 2 independent cohorts: The derivation cohort included 3691 patients, and the validation cohort included 2416 patients. Sociodemographic and clinical data, including visual acuity (VA) and the Visual Function Index 14 (VF-14), were collected before and after cataract extraction. Univariate and multivariate linear regression, and regression trees analysis were performed in the derivation cohort. Decision trees obtained in the derivation cohort were validated in the validation cohort. Final results were divided into appropriate or inappropriate indications and compared with a previously established benchmark of desirable VA and VF-14 gain in relation to preintervention VA classes. Main Outcome Measures Preintervention VA and changes 6 weeks after the intervention. Results Among patients with simple cataract, predictors of significant improvement in VA after cataract extraction were preintervention VA and negative surgical complexity. Among patients with cataract and other ocular comorbidity, preintervention visual function and expected postintervention VA also predicted change in VA. When compared with a benchmark based on the minimal clinically important difference in VA after cataract extraction, sensitivity for the decision trees was 83% for both diagnostic groups and specificities ranged from 36.2 to 54.8. Conclusions A simple decision tree based on changes in VA can help identify appropriate patients for cataract extraction and be used to evaluate clinical practice or for quality control. Financial Disclosure(s) The author(s) have no proprietary or commercial interest in any materials discussed in this article.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    36
    References
    14
    Citations
    NaN
    KQI
    []