STATISTICAL MODELS FOR PREDICTING A PATIENT’S RESPIRATORY DISTURBANCE INDEX

Abstract

SESSION TITLE: Clinical Prediction and Diagnosis of OSA SESSION TYPE: Original Investigations PRESENTED ON: 10/23/2019 10:45 AM - 11:45 AM PURPOSE: Respiratory disturbance index (RDI) is an important measure to diagnose the severity of obstructive sleep apnea (OSA). To get an accurate RDI value of a patient, an expensive sleep study must be undergone, and the patient needs to wear sleep and breathing monitoring sensors on the skin over a night. Although previous studies developed binary classification models for estimating the presence of OSA by using a patient’s anthropometric information [e.g., neck circumference, body mass index (BMI)], research is limited to directly predict a value of RDI. The present study developed statistical models to predict a patient’s RDI value by incorporating demographic information and anthropometric dimensions. METHODS: RDI data of 620 patients (254 females and 366 males) with a wide range of age and body size was collected at Torr Sleep Center in Corpus Christi, TX in 2007 and 2008. Prediction models for female and male patients were separately developed by employing a stepwise regression method (pin < 0.05 and pout > 0.05) using potential regressors including age, height (H; unit: inch), BMI (B; unit: kg/m2), neck circumference (NC; unit: inch), waist divided by hip circumference (WH), and two-way interactions among other variables (e.g., H × B, H × NC, and B × WH). Note that due to the correlation between waist and hip circumferences, the ratio between waist and hip circumferences was used as a potential regressor in the modeling process. RESULTS: The developed models for predicting RDI value for female and male patients are presented below. RDI female = -65.30 + 0.78 × BMI + 2.90 × NC + 0.29 × Age RDI male = -60.32 + 1.00 × BMI + 3.30 × NC The adj. R2 and root mean squared error (RMSE) of the female and male models are 0.17 and 22.5, and 0.16 and 24.7, respectively. RDI was significantly associated with age (p = 0.014), BMI (p < 0.001), and neck circumference (p = 0.010) for female and BMI (p < 0.001) and neck circumference (p = 0.005) for male. Although age does not have a significant effect on the RDI for the male patients, it was significant for the female patients. CONCLUSIONS: The statistical linear models for predicting a patient’s RDI values have been developed for the first time. The present study is the first study to report a significant effect of age on RDI for a female patient. CLINICAL IMPLICATIONS: The developed model would be particularly useful to predict a patient’s RDI value based on an easily measurable demographic information and anthropometric dimensions. For example, we can anticipate that if a NC of a female patient has been increased by one inch, then RDI value of the patient would be increased by 2.9. Note that the developed models are only applicable to patients, not healthy subjects. DISCLOSURES: No relevant relationships by Jangwoon Park, source=Web Response No relevant relationships by Alaa Sheta, source=Web Response Speaker/Speaker's Bureau relationship with Astra Zaneca, Sunovian, Paratek Please note: $1001 - $5000 Added 01/13/2019 by Salim Surani, source=Web Response, value=Consulting fee