Sample Size Analysis for Machine Learning Clinical Validation Studies.

BACKGROUND: Before integrating new machine learning (ML) into clinical practice, algorithms must undergo validation. Validation studies require sample size estimates. Unlike hypothesis testing studies seeking a p-value, the goal of validating predictive models is obtaining estimates of model performance. There is no standard tool for determining sample size estimates for clinical validation studies for machine learning models. METHODS: Our open-source method, Sample Size Analysis for Machine Learning (SSAML) was described and was tested in three previously published models: brain age to predict mortality (Cox Proportional Hazard), COVID hospitalization risk prediction (ordinal regression), and seizure risk forecasting (deep learning). RESULTS: Minimum sample sizes were obtained in each dataset using standardized criteria. DISCUSSION: SSAML provides a formal expectation of precision and accuracy at a desired confidence level. SSAML is open-source and agnostic to data type and ML model. It can be used for clinical validation studies of ML models.

Classification of neurologic outcomes from medical notes using natural language processing.

Neurologic disability level at hospital discharge is an important outcome in many clinical research studies. Outside of clinical trials, neurologic outcomes must typically be extracted by labor intensive manual review of clinical notes in the electronic health record (EHR). To overcome this challenge, we set out to develop a natural language processing (NLP) approach that automatically reads clinical notes to determine neurologic outcomes, to make it possible to conduct larger scale neurologic outcomes studies. We obtained 7314 notes from 3632 patients hospitalized at two large Boston hospitals between January 2012 and June 2020, including discharge summaries (3485), occupational therapy (1472) and physical therapy (2357) notes. Fourteen clinical experts reviewed notes to assign scores on the Glasgow Outcome Scale (GOS) with 4 classes, namely 'good recovery', 'moderate disability', 'severe disability', and 'death' and on the Modified Rankin Scale (mRS), with 7 classes, namely 'no symptoms', 'no significant disability', 'slight disability', 'moderate disability', 'moderately severe disability', 'severe disability', and 'death'. For 428 patients' notes, 2 experts scored the cases generating interrater reliability estimates for GOS and mRS. After preprocessing and extracting features from the notes, we trained a multiclass logistic regression model using LASSO regularization and 5-fold cross validation for hyperparameter tuning. The model performed well on the test set, achieving a micro average area under the receiver operating characteristic and F-score of 0.94 (95% CI 0.93-0.95) and 0.77 (0.75-0.80) for GOS, and 0.90 (0.89-0.91) and 0.59 (0.57-0.62) for mRS, respectively. Our work demonstrates that an NLP algorithm can accurately assign neurologic outcomes based on free text clinical notes. This algorithm increases the scale of research on neurological outcomes that is possible with EHR data.

Identifying Patients With Delirium Based on Unstructured Clinical Notes: Observational Study.

BACKGROUND: Delirium in hospitalized patients is a syndrome of acute brain dysfunction. Diagnostic (International Classification of Diseases [ICD]) codes are often used in studies using electronic health records (EHRs), but they are inaccurate. OBJECTIVE: We sought to develop a more accurate method using natural language processing (NLP) to detect delirium episodes on the basis of unstructured clinical notes. METHODS: We collected 1.5 million notes from >10,000 patients from among 9 hospitals. Seven experts iteratively labeled 200,471 sentences. Using these, we trained three NLP classifiers: Support Vector Machine, Recurrent Neural Networks, and Transformer. Testing was performed using an external data set. We also evaluated associations with delirium billing (ICD) codes, medications, orders for restraints and sitters, direct assessments (Confusion Assessment Method [CAM] scores), and in-hospital mortality. F1 scores, confusion matrices, and areas under the receiver operating characteristic curve (AUCs) were used to compare NLP models. We used the φ coefficient to measure associations with other delirium indicators. RESULTS: The transformer NLP performed best on the following parameters: micro F1=0.978, macro F1=0.918, positive AUC=0.984, and negative AUC=0.992. NLP detections exhibited higher correlations (φ) than ICD codes with deliriogenic medications (0.194 vs 0.073 for ICD codes), restraints and sitter orders (0.358 vs 0.177), mortality (0.216 vs 0.000), and CAM scores (0.256 vs -0.028). CONCLUSIONS: Clinical notes are an attractive alternative to ICD codes for EHR delirium studies but require automated methods. Our NLP model detects delirium with high accuracy, similar to manual chart review. Our NLP approach can provide more accurate determination of delirium for large-scale EHR-based studies regarding delirium, quality improvement, and clinical trails.

Prolonged Intubation in Patients With Prior Cerebrovascular Disease and COVID-19.

Objectives: Patients with comorbidities are at increased risk for poor outcomes in COVID-19, yet data on patients with prior neurological disease remains limited. Our objective was to determine the odds of critical illness and duration of mechanical ventilation in patients with prior cerebrovascular disease and COVID-19. Methods: A observational study of 1,128 consecutive adult patients admitted to an academic center in Boston, Massachusetts, and diagnosed with laboratory-confirmed COVID-19. We tested the association between prior cerebrovascular disease and critical illness, defined as mechanical ventilation (MV) or death by day 28, using logistic regression with inverse probability weighting of the propensity score. Among intubated patients, we estimated the cumulative incidence of successful extubation without death over 45 days using competing risk analysis. Results: Of the 1,128 adults with COVID-19, 350 (36%) were critically ill by day 28. The median age of patients was 59 years (SD: 18 years) and 640 (57%) were men. As of June 2nd, 2020, 127 (11%) patients had died. A total of 177 patients (16%) had a prior cerebrovascular disease. Prior cerebrovascular disease was significantly associated with critical illness (OR = 1.54, 95% CI = 1.14-2.07), lower rate of successful extubation (cause-specific HR = 0.57, 95% CI = 0.33-0.98), and increased duration of intubation (restricted mean time difference = 4.02 days, 95% CI = 0.34-10.92) compared to patients without cerebrovascular disease. Interpretation: Prior cerebrovascular disease adversely affects COVID-19 outcomes in hospitalized patients. Further study is required to determine if this subpopulation requires closer monitoring for disease progression during COVID-19.

CoVA: An Acuity Score for Outpatient Screening that Predicts Coronavirus Disease 2019 Prognosis.

BACKGROUND: We sought to develop an automatable score to predict hospitalization, critical illness, or death for patients at risk for coronavirus disease 2019 (COVID-19) presenting for urgent care. METHODS: We developed the COVID-19 Acuity Score (CoVA) based on a single-center study of adult outpatients seen in respiratory illness clinics or the emergency department. Data were extracted from the Partners Enterprise Data Warehouse, and split into development (n = 9381, 7 March-2 May) and prospective (n = 2205, 3-14 May) cohorts. Outcomes were hospitalization, critical illness (intensive care unit or ventilation), or death within 7 days. Calibration was assessed using the expected-to-observed event ratio (E/O). Discrimination was assessed by area under the receiver operating curve (AUC). RESULTS: In the prospective cohort, 26.1%, 6.3%, and 0.5% of patients experienced hospitalization, critical illness, or death, respectively. CoVA showed excellent performance in prospective validation for hospitalization (expected-to-observed ratio [E/O]: 1.01; AUC: 0.76), for critical illness (E/O: 1.03; AUC: 0.79), and for death (E/O: 1.63; AUC: 0.93). Among 30 predictors, the top 5 were age, diastolic blood pressure, blood oxygen saturation, COVID-19 testing status, and respiratory rate. CONCLUSIONS: CoVA is a prospectively validated automatable score for the outpatient setting to predict adverse events related to COVID-19 infection.

Classification of the Disposition of Patients Hospitalized with COVID-19: Reading Discharge Summaries Using Natural Language Processing.

BACKGROUND: Medical notes are a rich source of patient data; however, the nature of unstructured text has largely precluded the use of these data for large retrospective analyses. Transforming clinical text into structured data can enable large-scale research studies with electronic health records (EHR) data. Natural language processing (NLP) can be used for text information retrieval, reducing the need for labor-intensive chart review. Here we present an application of NLP to large-scale analysis of medical records at 2 large hospitals for patients hospitalized with COVID-19. OBJECTIVE: Our study goal was to develop an NLP pipeline to classify the discharge disposition (home, inpatient rehabilitation, skilled nursing inpatient facility [SNIF], and death) of patients hospitalized with COVID-19 based on hospital discharge summary notes. METHODS: Text mining and feature engineering were applied to unstructured text from hospital discharge summaries. The study included patients with COVID-19 discharged from 2 hospitals in the Boston, Massachusetts area (Massachusetts General Hospital and Brigham and Women's Hospital) between March 10, 2020, and June 30, 2020. The data were divided into a training set (70%) and hold-out test set (30%). Discharge summaries were represented as bags-of-words consisting of single words (unigrams), bigrams, and trigrams. The number of features was reduced during training by excluding n-grams that occurred in fewer than 10% of discharge summaries, and further reduced using least absolute shrinkage and selection operator (LASSO) regularization while training a multiclass logistic regression model. Model performance was evaluated using the hold-out test set. RESULTS: The study cohort included 1737 adult patients (median age 61 [SD 18] years; 55% men; 45% White and 16% Black; 14% nonsurvivors and 61% discharged home). The model selected 179 from a vocabulary of 1056 engineered features, consisting of combinations of unigrams, bigrams, and trigrams. The top features contributing most to the classification by the model (for each outcome) were the following: "appointments specialty," "home health," and "home care" (home); "intubate" and "ARDS" (inpatient rehabilitation); "service" (SNIF); "brief assessment" and "covid" (death). The model achieved a micro-average area under the receiver operating characteristic curve value of 0.98 (95% CI 0.97-0.98) and average precision of 0.81 (95% CI 0.75-0.84) in the testing set for prediction of discharge disposition. CONCLUSIONS: A supervised learning-based NLP approach is able to classify the discharge disposition of patients hospitalized with COVID-19. This approach has the potential to accelerate and increase the scale of research on patients' discharge disposition that is possible with EHR data.