Big data study using health insurance claims to predict multidisciplinary low vision service uptake

Abstract

SIGNIFICANCE There is a lack of research from high-income countries with various health care and funding systems regarding barriers and facilitators in low vision services (LVS) access. Furthermore, very few studies on LVS provision have used claims data. PURPOSE This study aimed to investigate which patient characteristics predict receiving multidisciplinary LVS (MLVS) in the Netherlands, a high-income country, based on health care claims data. METHODS Data from a Dutch national health insurance claims database (2015 to 2018) of patients with eye diseases causing potentially severe visual impairment were retrieved. Patients received MLVS (n = 8766) and/or ophthalmic treatment in 2018 (reference, n = 565,496). MLVS is provided by professionals from various clinical backgrounds, including nonprofit low vision optometry. Patient characteristics (sociodemographic, clinical, contextual, general health care utilization) were assessed as potential predictors using a multivariable logistic regression model, which was internally validated with bootstrapping. RESULTS Predictors for receiving MLVS included prescription of low vision aids (odds ratio [OR], 8.76; 95% confidence interval [CI], 7.99 to 9.61), having multiple ophthalmic diagnoses (OR, 3.49; 95% CI, 3.30 to 3.70), receiving occupational therapy (OR, 2.32; 95% CI, 2.15 to 2.51), mental comorbidity (OR, 1.17; 95% CI, 1.10 to 1.23), comorbid hearing disorder (OR, 1.98; 95% CI, 1.86 to 2.11), and receiving treatment in both a general hospital and a specialized ophthalmic center (OR, 1.23; 95% CI, 1.10 to 1.37), or by a general practitioner (OR, 1.23; 95% CI, 1.18 to 1.29). Characteristics associated with lower odds included older age (OR, 0.30; 95% CI, 0.28 to 0.32), having a low social economic status (OR, 0.91; 95% CI, 0.86 to 0.97), physical comorbidity (OR, 0.87; 95% CI, 0.82 to 0.92), and greater distance to an MLVS (OR, 0.95; 95% CI, 0.92 to 0.98). The area under the curve of the model was 0.75 (95% CI, 0.75 to 0.76; optimism = 0.0008). CONCLUSIONS Various sociodemographic, clinical, and contextual patient characteristics, as well as factors related to patients' general health care utilization, were found to influence MLVS receipt as barriers or facilitators. Eye care practitioners should have attention for socioeconomically disadvantaged older patients when considering MLVS referral.