Plastic Surgery Research Council

Back to 2018 Program


Geospatial Econometric Exploration of Risk Factors for Loss to Followup in Cleft Care
Banafsheh Sharif-Askary, B.S., Peter G. Bittar, B.S., B.A., Alfredo Farjat, Ph.D., Beiyu Liu, Ph.D., Joao Vissoci, Ph.D., Alexander C. Allori, M.D., M.P.H.
Duke University, Durham, NC, USA.

PURPOSE: Cleft lip and/or palate (CL/P) is among the most common congenital conditions. Due to its numerous functional and aesthetic sequelae, long-term, multidisciplinary team care is essential. There are a limited number of approved cleft teams in the United States; consequently, some patients who live remote from the closest cleft team may never initiate care, and patients who begin care in infancy may be lost to followup. The reasons for loss to followup (LTFU) are varied and complex but may include low socioeconomic status (SES) and geographic isolation. This study aims to identify patient characteristics associated with LTFU for a mid-volume cleft team that serves a mixed urban/suburban/rural population in North Carolina.
METHODS: Medical records were retrospectively evaluated for 558 children with CL/P, aged 0-15 years, and treated at Duke Children's Hospital from 1998-2013. The primary outcome was LTFU, defined as three consecutive missed appointments and two years without seeing the team despite attempts at reestablishing followup. Patients who transferred care to other teams were not considered LTFU. The secondary outcome variable was age at last successful encounter. Patient demographics included cleft phenotype, sex, race/ethnicity, location (address, ZIP code, and county FIPS code), and rural/urban designation. SES index was assigned by linkage with the U.S. Census American Community Survey. Spatial dependency was evaluated using variograms, Moran's I test, Geary's CC test, and BB join count tests. The probability of LTFU was assessed using a Bayesian approach to hierarchical generalized linear geostatistical modeling. Risk maps were plotted to summarize at-risk populations.
RESULTS: 29% of patients seen in this time period were lost to followup. When ignoring spatial dependency, younger age at last encounter was a strong predictor of LTFU (p<0.0001), while SES and cleft phenotype were weakly associated with LTFU. When including spatial dependency in the model, both SES and phenotype became significant. Distance from the team and rural/urban designation were not statistically significant. Cartographic representation of predicted probability of LTFU was prepared using SES, age at last encounter, and cleft phenotype, revealing discrete pockets of at-risk communities (Figure 1A). Surprisingly, not all of these communities were far from our team; rather, some vulnerable communities were identified in the immediate vicinity (Figure 1B).
CONCLUSION: SES, age at last encounter, and cleft phenotype are associated with the risk of LTFU, and these factors are spatially dependent. Specific vulnerable communities were identified to be at high-risk for LTFU. In future work, we plan to expand the analysis to include patients from all approved cleft teams in NC and the Mid-Atlantic States region. This information will help cleft teams better allocate resources to high-risk areas so that deficiencies in care may be prevented or rectified.


Back to 2018 Program