Using A Bivariate Linear Regression Model To Predict Time To Maturation In Arteriovenous Fistula Patients
Christina M. Stuart, BS, Nicholas B. Safian, BA, Joseph W. Schaefer, BCHe, Paul J. Dimuzio, MD FACS, Dawn M. Salvatore, MD FACS.
Thomas Jefferson University Hospital, Philadelphia, PA, USA.
OBJECTIVES: To explore the potential influence of patient age, body mass index, hemoglobin A1C and lipid profile on the maturation of arteriovenous fistulas in patients with advanced chronic kidney disease. Previous studies have examined these factors in the context of a binary outcome (maturation success versus failure) but have not further defined their effect on time. Here we attempt to develop a model that can predict time to fistula maturation, and thus fistula usability, using these simple parameters.
METHODS: A retrospective, single-center, single-surgeon review of patients who underwent surgical creation of a fistula was performed. Fistula maturation was determined based on surgeon expertise and ultrasound findings. Pearson product-moment correlation coefficients were computed to assess the relationship between time to maturation (TTM) and various continuous variables, and ultimately a simple logistic regression model was developed to predict time to fistula maturation in days.
RESULTS: A total of 91 cases of mature fistulas were reviewed. The majority of patients were male (68%) and nearly half were African-American (47%). The average age was 58±14 years and the average BMI was 28.2±7.0. Approximately half (46%) of patients were diabetic with average A1C of 6.0±1.2, and the majority (64%) carried a diagnosis of hyperlipidemia with the average cholesterol of 150±59. Nearly all (96%) of patients carried a diagnosis of hypertension and 40% had hyperparathyroidism. The average time to maturation was 44 days. There was a significant positive correlation between time to maturation and age (p =0.000), BMI (p = 0.000), A1C (p = 0.002), and cholesterol (p = 0.000). Using linear regression plots, a simple equation that predicts time to maturation in days using age and BMI was developed (p = 0.000). Validation of the model using a paired t-test showed no significant difference between the observed TTM and the expected TTM based on the model (p = 0.000), as well as a Pearson product-moment, which showed strong correlation between the predictive model and observed TTM (r = 0.560, p = 0.000).
CONCLUSIONS: Age, body mass index, hemoglobin A1C and total cholesterol levels all have strong positive correlations with time to fistula maturation in adults. Patient age and BMI can be used to predict time to maturation (in days) using a simple linear model.
Back to 2020 ePosters