Big data analysis to uncover determinants of patient appointment compliance at Kresge Eye Institute

Abstract

Purpose : Patient appointment compliance (AC) remains a major obstacle to the physician-patient relationship and in health care delivery. This study aims to uncover appointment and patient characteristics functioning as determinants to AC at Kresge Eye Institute (KEI).

Methods : Retrospective analysis of electronic medical records (EMR) was performed across all appointments scheduled 01/2014 to 12/31/2018 at KEI’s 23 out-patient Michigan locations. Patient arrival to appointment was classified as compliant (CO) and patient cancellation/no-show was classified as non-compliant (NC). Driving distance (DD) was generated from patient zip code. Appointment rank (appt rank) describes relative chronology in patient appointments. The following statistical analyses were used: chi-square (categorical and binary variables), Mann-Whitney U (continuous variables), and logistic regression (to control for covariates).

Results : A total of 835,207 appointments across 13 ophthalmic specialties were studied. Our population was predominantly African American (59.65%), female (60.5%), and mean age at appointment was 57.6 yrs (IQR: 47.4-71.4; SD 20). Medicare was the most frequent primary insurance at 258,414 appointments. Mean DD was 17.4 miles (range 0.3-578; SD 9.3). Overall, 60.3% of appointments were classified as CO.
Relative to NC, CO appointments were associated with patients who were older (62.8 vs. 58.8 years; p<0.00001), male (40.1% vs. 38%; p<0.0001), higher appointment rank (7 vs. 6; p<0.00001), and Medicare insurance (32.2% vs. 29.1%; p<0.0001). Retina and Glaucoma specialties had significantly more CO than NC patients (p<0.0001). DD differed significantly across CO and NC (CO IQR:13.4 to 20 miles; NC IQR: 16.8 to 20 miles). When covariates were controlled for (DD, age, race, sex, specialty, scheduling location/provider), then DD was found to be a significant predictor of AC (p<0.0001, OR: 0.997, 95 CI: 0.996 to 0.997)- though the most powerful predictor was sex (p<0.0001, OR: 1.087, 95 CI: 1.05 to 1.12).

Conclusions : Demographic and administrative characteristics were found to be associated with AC. It is notable that driving distance was found to be associated with AC even when all other confounders were controlled for. Interestingly, patient sex was found to be the most powerful predictor of AC. In the future, patient accessibility and satisfaction can be considered with AC to better the quality of care.

This is a 2020 ARVO Annual Meeting abstract.