A multivariate analysis of dermatology missed appointment predictors.

Dermatology appointment nonattendance rates range from 17% to 31%, 1-3 and patients who miss appointments without prior notification (no-shows and same-day cancellations) disrupt schedules , decrease access for others, resulting in underutilization of resources and thereby increasing cost, and interrupt continuity of medical care. 1,4 Therefore, we set out to determine if easily attainable variables from scheduling data could be used to predict patients likely to miss dermatology appointments. Methods | This study was conducted in the department of medical dermatology at the Massachusetts General Hospital under the approval of the Partners institutional review board. The department employs 30 physicians, rendering approximately 59 000 visits annually. It uses the IDX scheduling system and Televox for automated reminder calls 3 days prior to visit. Patients, at that time, were not charged for missing appointments. The outcome measure, missed appointments, was the failure to arrive without notifying the practice before the appointment date. Data were collected for 47 348 medical dermatology appointments from August 2010 through July 2011, representing 80% of appointments. The number of previously missed medical dermatology appointments was collected from Sep-tember 2008 through July 2011. New and follow-up appointments (routine and urgent) for weekday appointments were included. Appointments on days with heavy snowfall or outside the standard session (ie, 8:00 AM–12:00 PM and 1:00-5:00 PM Monday-Friday) were excluded. The following variables were identified through literature searches and input from department leadership: appointment type, weekday, appointment hour, wait days (number of days between scheduling and appointment date), language, age, insurance type, sex, and number of previously missed medical dermatology appointments. Insurance type was divided into commercial insurance with a copy, plans with co-insurance, Medicare, Medicaid, free care, and self-insured. Data were extracted using Standard Query Language (SQL) and were analyzed using SAS software (version 9.2). Univari-ate regression tests were performed on all variables, and a mul-tivariate logistic regression was performed on statistically significant variables (P < .05). Results | The final cohort had 41 893 records with 7812 missed appointments (18.6%). Forty-one percent of patients who missed appointments did arrive at a future appointment within 1 year. Through univariate analyses (Table), all variables were statistically significant (P < .05). All variables were included in a multivariate logistic regression, and weekday, wait days (Figure), language, age group, insurance type, and number of previously missed appointments remained statistically sig