Towards Validating the Effectiveness of Obstructive Sleep Apnea Classification from Electronic Health Records Using Machine Learning

Abstract

Obstructive sleep apnea (OSA) is a common, chronic, sleep-related breathing disorder characterized by partial or complete airway obstruction in sleep. The gold standard diagnosis method is polysomnography, which estimates disease severity through the Apnea-Hypopnea Index (AHI). However, this is expensive and not widely accessible to the public. For effective screening, this work implements machine learning algorithms for classification of OSA. The model is trained with routinely acquired clinical data of 1479 records from the Wisconsin Sleep Cohort dataset. Extracted features from the electronic health records include patient demographics, laboratory blood reports, physical measurements, habitual sleep history, comorbidities, and general health questionnaire scores. For distinguishing between OSA and non-OSA patients, feature selection methods reveal the primary important predictors as waist-to-height ratio, waist circumference, neck circumference, body-mass index, lipid accumulation product, excessive daytime sleepiness, daily snoring frequency and snoring volume. Optimal hyperparameters were selected using a hybrid tuning method consisting of Bayesian Optimization and Genetic Algorithms through a five-fold cross-validation strategy. Support vector machines achieved the highest evaluation scores with accuracy: 68.06%, sensitivity: 88.76%, specificity: 40.74%, F1-score: 75.96%, PPV: 66.36% and NPV: 73.33%. We conclude that routine clinical data can be useful in prioritization of patient referral for further sleep studies.

Comparison of Composite Lipid Indices in Patients with Obstructive Sleep Apnoea

“…Although LAP, VAI and TyG were all evaluated, only LAP was included in the final model. 34 In our large group of subjects, lipid parameters were significantly different between the OSA and control groups. To compare the strength of associations between OSA and composite lipid indices we used ROC analysis followed by the DeLong test.…”

Extraction of Sleep Information from Clinical Notes of Patients with Alzheimer’s Disease Using Natural Language Processing

“…Hsiao et al 22 identified patients with sleep disorders using ICD-9 codes 307.4 and 780.5x and explored the association between sleep disorders and autoimmune diseases. ICD codes have also been used to identify obstructive sleep apnea 23, 24 . However, it has been shown that sleep disorders are poorly coded in structured EHR data.…”

Rough-set based learning methods: A case study to assess the relationship between the clinical delivery of cannabinoid medicine for anxiety, depression, sleep, patterns and predictability

“…The paper attempts to find an association between important biomarkers (features) and depression in order to address the problem of patients not disclosing depressive symptoms during a clinical interview. In [11], features such as blood report, physical demographic, sleep history extracted from electronic health records were used to detect obstructive sleep apnea disorder using an SVM classifier with 68.06% score for accuracy measure and 88.76% score for sensitivity measure. Authors of [22] proposed a novel machine learning pipeline containing 6 Classifiers (XGBoost, Random Forest, Support Vector Machine, KNN and a neural network tuned using Bayesian hyperparameter optimization) to predict general anxiety disorder (GAD) and major depressive disorder (MDD) problems.…”

“…Two different classifiers were incorporated (softmax and SVM) with 3D-CNN-SVM giving the best result. Improving diagnosis of depression with XGBOOST machine learning model and a large biomarkers Dutch dataset [20] Towards Validating the Effectiveness of Obstructive Sleep Apnea Classification from Electronic Health Records Using Machine Learning [11] Predicting mental health treatment access among adolescents with elevated depressive symptoms: Machine learning approaches [22] Predicting healthcare trajectories from medical records: A deep learning approach [40] Applying deep neural networks to unstructured text notes in electronic medical records for phenotyping youth depression [39] The use of electronic health records for psychiatric phenotyping and genomics [42] A machine learning approach to modeling PTSD and difficulties in emotion regulation [43] Predicting personalized process-outcome associations in psychotherapy using machine learning approachesA demonstration [44] Table 2.2: Social media as dataset -related papers…”

“…With the availability of large amounts of patient data, machine learning (ML) techniques, specifically, supervised and deep learning classifiers, have made it possible to detect, diagnose and treat mental health disorders. Common dataset formats include: Electronic health records (EHR) [11], Social Media (e.g., twitter, Reddit) [12,13], Image (e.g., MRI) [14] and Audio [15]. [16], present an in-depth review of about 300 papers related to ML and its application in mental health.…”