Cost-effectiveness analysis on implementing newborn hearing screening programmes in a low- to middle-income country.

OBJECTIVE: To analyse the cost-effectiveness (CE) of implementing different newborn hearing screening protocols in a low- to middle-income country. DESIGN: A decision analytical model with a 78-year time horizon. STUDY SAMPLE: Direct medical, direct non-medical and indirect costs were collected from 126 subjects in southern Thailand. Various protocols involving universal newborn hearing screening (UNHS) and targeted newborn hearing screening (TNHS), using two technologies, namely automated otoacoustic emissions (aOAEs) and automated auditory brainstem responses (aABRs), were evaluated. Incremental cost-effectiveness ratios (ICERs) were calculated for all protocols in United States dollars (US$)/quality-adjusted life year (QALY) gained. Also, probabilistic sensitivity analyses with 1000 trials for each specific protocol were performed. RESULTS: The ICERs of UNHS with aOAE, UNHS with aABR, TNHS with aABR and UNHS with optimised baseline parameters were 3702, 3545, 1545 and 2483 US$/QALY gained, respectively. With the CE threshold of 5000 US$/QALY gained, the chances of ICERs to be cost-effective for UNHS with aOAE, UNHS with aABR, TNHS with aABR and UNHS with optimised baseline parameters were 72, 77, 93 and 94%, respectively. CONCLUSIONS: All screening protocols were considered as cost-effective, and a very high chance of being cost-effective for UNHS could be achieved when certain baseline parameters were optimised.

Robust machine learning method for imputing missing values in audiograms collected in children.

OBJECTIVE: To assess the accuracy and reliability of a machine learning (ML) algorithm for predicting the full audiograms of hearing-impaired children relative to the common approach (CA). DESIGN: Retrospective study. STUDY SAMPLE: There were 206 audiograms included from 206 children with sensorineural hearing loss. Nested cross-validation was used for evaluating the performance of the CA and ML. Six audiogram prediction simulations were performed in which either one or two thresholds across 0.5-4 kHz from complete audiograms in the dataset were labelled. Missing thresholds at the remaining frequencies were then predicted using the CA and ML in each simulation. The accuracy of the ML algorithm was determined by comparing the median average absolute threshold differences between the CA and ML using Wilcoxon signed-rank test. The reliability between runs of the ML was also assessed with Cronbach's alphas. RESULTS: The median average absolute threshold differences in ML (5-8 dBHL) were statistically significantly lower than those in CA (6.25-10 dBHL) in all six simulations (p value < 0.05). The ML algorithm was also found to be reliable to predict the audiograms in all six simulations (α > 0.9). CONCLUSION: Using the ML to predict the children's audiograms was reliable and more accurate than using the CA.

CBCT-to-CT Translation Using Registration-Based Generative Adversarial Networks in Patients with Head and Neck Cancer.

Recently, deep learning with generative adversarial networks (GANs) has been applied in multi-domain image-to-image translation. This study aims to improve the image quality of cone-beam computed tomography (CBCT) by generating synthetic CT (sCT) that maintains the patient's anatomy as in CBCT, while having the image quality of CT. As CBCT and CT are acquired at different time points, it is challenging to obtain paired images with aligned anatomy for supervised training. To address this limitation, the study incorporated a registration network (RegNet) into GAN during training. RegNet can dynamically estimate the correct labels, allowing supervised learning with noisy labels. The study developed and evaluated the approach using imaging data from 146 patients with head and neck cancer. The results showed that GAN trained with RegNet performed better than those trained without RegNet. Specifically, in the UNIT model trained with RegNet, the mean absolute error (MAE) was reduced from 40.46 to 37.21, the root mean-square error (RMSE) was reduced from 119.45 to 108.86, the peak signal-to-noise ratio (PSNR) was increased from 28.67 to 29.55, and the structural similarity index (SSIM) was increased from 0.8630 to 0.8791. The sCT generated from the model had fewer artifacts and retained the anatomical information as in CBCT.

Prediction of Acid-Base and Potassium Imbalances in Intensive Care Patients Using Machine Learning Techniques.

Acid-base disorders occur when the body's normal pH is out of balance. They can be caused by problems with kidney or respiratory function or by an excess of acids or bases that the body cannot properly eliminate. Acid-base and potassium imbalances are mechanistically linked because acid-base imbalances can alter the transport of potassium. Both acid-base and potassium imbalances are common in critically ill patients. This study investigated machine learning models for predicting the occurrence of acid-base and potassium imbalances in intensive care patients. We used an institutional dataset of 1089 patients with 87 variables, including vital signs, general appearance, and laboratory results. Gradient boosting (GB) was able to predict nine clinical conditions related to acid-base and potassium imbalances: mortality (AUROC = 0.9822), hypocapnia (AUROC = 0.7524), hypercapnia (AUROC = 0.8228), hypokalemia (AUROC = 0.9191), hyperkalemia (AUROC = 0.9565), respiratory acidosis (AUROC = 0.8125), respiratory alkalosis (AUROC = 0.7685), metabolic acidosis (AUROC = 0.8682), and metabolic alkalosis (AUROC = 0.8284). Some predictions remained relatively robust even when the prediction window was increased. Additionally, the decision-making process was made more interpretable and transparent through the use of SHAP analysis. Overall, the results suggest that machine learning could be a useful tool to gain insight into the condition of intensive care patients and assist in the management of acid-base and potassium imbalances.

AI-Assisted Assessment of Wound Tissue with Automatic Color and Measurement Calibration on Images Taken with a Smartphone.

Wound assessment is essential for evaluating wound healing. One cornerstone of wound care practice is the use of clinical guidelines that mandate regular documentation, including wound size and wound tissue composition, to determine the rate of wound healing. The traditional method requires wound care professionals to manually measure the wound area and tissue composition, which is time-consuming, costly, and difficult to reproduce. In this work, we propose an approach for automatic wound assessment that incorporates automatic color and measurement calibration and artificial intelligence algorithms. Our approach enables the comparison of images taken at different times, even if they were taken under different lighting conditions, distances, lenses, and camera sensors. We designed a calibration chart and developed automatic algorithms for color and measurement calibration. The wound area and wound composition on the images were annotated by three physicians with more than ten years of experience. Deep learning models were then developed to mimic what the physicians did on the images. We examined two network variants, U-Net with EfficientNet and U-Net with MobileNetV2, on wound images with a size of 1024 × 1024 pixels. Our best-performing algorithm achieved a mean intersection over union (IoU) of 0.6964, 0.3957, 0.6421, and 0.1552 for segmenting a wound area, epithelialization area, granulation tissue, and necrotic tissue, respectively. Our approach was able to accurately segment the wound area and granulation tissue but was inconsistent with respect to the epithelialization area and necrotic tissue. The calibration chart, which helps calibrate colors and scales, improved the performance of the algorithm. The approach could provide a thorough assessment of the wound, which could help clinicians tailor treatment to the patient's condition.

Teamwork Management and Benefit of Telemedicine in COVID-19 Outbreak Control on an Offshore Vessel in the Gulf of Thailand, Songkhla Province, Thailand: A Case Report.

In May 2021, there was a COVID-19 outbreak on board a construction support ship traveling from India to Thailand. Controlling the outbreak on this offshore vessel from 11 May to 2 June 2021 was applied. This case report describes the teamwork management of COVID-19 control on the vessel in the Gulf of Thailand. We summarized the COVID-19 outbreak control process on board, including active COVID-19-infected cases (CoIC) and close contacts (CoCC) identification, isolation, quarantine, treatment, and clinical monitoring using telemedicine to report their health measurements twice daily, including emergency conditions if they occurred. Active COVID-19 cases were identified by two rounds of reverse transcription polymerase chain reaction (RT-PCR) tests in all crew members, in which 7 of 29 (24.1%) showed positive results. Both the CoIC and CoCC were strictly and absolutely isolated and quarantined on the vessel. No serious medical conditions were reported during the monitoring. The third-round RT-PCR tests were conducted, and all tested negative one week later. Teamwork management in proactive COVID-19 case identification, isolation, comprehensive treatment, and close monitoring of health conditions using telemedicine devices is beneficial for controlling the COVID-19 outbreak on board.

Multi-label classification of symptom terms from free-text bilingual adverse drug reaction reports using natural language processing.

Allergic reactions to medication range from mild to severe or even life-threatening. Proper documentation of patient allergy information is critical for safe prescription, avoiding drug interactions, and reducing healthcare costs. Allergy information is regularly obtained during the medical interview, but is often poorly documented in electronic health records (EHRs). While many EHRs allow for structured adverse drug reaction (ADR) reporting, a free-text entry is still common. The resulting information is neither interoperable nor easily reusable for other applications, such as clinical decision support systems and prescription alerts. Current approaches require pharmacists to review and code ADRs documented by healthcare professionals. Recently, the effectiveness of machine algorithms in natural language processing (NLP) has been widely demonstrated. Our study aims to develop and evaluate different NLP algorithms that can encode unstructured ADRs stored in EHRs into institutional symptom terms. Our dataset consists of 79,712 pharmacist-reviewed drug allergy records. We evaluated three NLP techniques: Naive Bayes-Support Vector Machine (NB-SVM), Universal Language Model Fine-tuning (ULMFiT), and Bidirectional Encoder Representations from Transformers (BERT). We tested different general-domain pre-trained BERT models, including mBERT, XLM-RoBERTa, and WanchanBERTa, as well as our domain-specific AllergyRoBERTa, which was pre-trained from scratch on our corpus. Overall, BERT models had the highest performance. NB-SVM outperformed ULMFiT and BERT for several symptom terms that are not frequently coded. The ensemble model achieved an exact match ratio of 95.33%, a F1 score of 98.88%, and a mean average precision of 97.07% for the 36 most frequently coded symptom terms. The model was then further developed into a symptom term suggestion system and achieved a Krippendorff's alpha agreement coefficient of 0.7081 in prospective testing with pharmacists. Some degree of automation could both accelerate the availability of allergy information and reduce the efforts for human coding.

Drug Recommendation from Diagnosis Codes: Classification vs. Collaborative Filtering Approaches.

Over time, large amounts of clinical data have accumulated in electronic health records (EHRs), making it difficult for healthcare professionals to navigate and make patient-centered decisions. This underscores the need for healthcare recommendation systems that help medical professionals make faster and more accurate decisions. This study addresses drug recommendation systems that generate an appropriate list of drugs that match patients' diagnoses. Currently, recommendations are manually prepared by physicians, but this is difficult for patients with multiple comorbidities. We explored approaches to drug recommendations based on elderly patients with diabetes, hypertension, and cardiovascular disease who visited primary-care clinics and often had multiple conditions. We examined both collaborative filtering approaches and traditional machine-learning classifiers. The hybrid model between the two yielded a recall at 5 of 76.61%, a precision at 5 of 46.20%, a macro-averaged area under the curve of 74.52%, and an average physician agreement of 47.50%. Although collaborative filtering is widely used in recommendation systems, our results showed that it consistently underperformed traditional classification. Collaborative filtering was sensitive to class imbalances and favored the more popular classes. This study highlighted challenges that need to be addressed when developing recommendation systems in EHRs.

Integrative Effects between a Bubble and Seal Program and Workers' Compliance to Health Advice on Successful COVID-19 Transmission Control in a Factory in Southern Thailand.

Applying health measures to prevent COVID-19 transmission caused disruption of businesses. A practical plan to balance public health and business sustainability during the pandemic was needed. Herein, we describe a "Bubble and Seal" (B&S) program implemented in a frozen seafood factory in southern Thailand. We enrolled 1539 workers who lived in the factory dormitories. First, the workers who had a high fatality risk were triaged by RT-PCR tests, quarantined and treated if they had COVID-19. Newly diagnosed or suspected COVID-19 workers underwent the same practices. The non-quarantined workers were regulated to work and live in their groups without contact across the groups. Workers' personal hygiene and preventive measures were strongly stressed. Between the 6th and 9th weeks of the program, the post-COVID-19 infection status (PCIS) of all participants was evaluated by mass COVID-19 antibody or RT-PCR tests. Finally, 91.8% of the workers showed positive PCIS, which was above the number required for program exit. Although no workers had received a vaccination, there was only one case of severe COVID-19 pneumonia, and no evidence of COVID-19 spreading to the surrounding communities. Implementation of the B&S program and workers' adherence to health advice was the key to this success.

Feature Explanations in Recurrent Neural Networks for Predicting Risk of Mortality in Intensive Care Patients.

Critical care staff are presented with a large amount of data, which made it difficult to systematically evaluate. Early detection of patients whose condition is deteriorating could reduce mortality, improve treatment outcomes, and allow a better use of healthcare resources. In this study, we propose a data-driven framework for predicting the risk of mortality that combines high-accuracy recurrent neural networks with interpretable explanations. Our model processes time-series of vital signs and laboratory observations to predict the probability of a patient's mortality in the intensive care unit (ICU). We investigated our approach on three public critical care databases: Multiparameter Intelligent Monitoring in Intensive Care III (MIMIC-III), MIMIC-IV, and eICU. Our models achieved an area under the receiver operating characteristic curve (AUC) of 0.87-0.91. Our approach was not only able to provide the predicted mortality risk but also to recognize and explain the historical contributions of the associated factors to the prediction. The explanations provided by our model were consistent with the literature. Patients may benefit from early intervention if their clinical observations in the ICU are continuously monitored in real time.

Molecular Classification Models for Triple Negative Breast Cancer Subtype Using Machine Learning.

Triple negative breast cancer (TNBC) lacks well-defined molecular targets and is highly heterogenous, making treatment challenging. Using gene expression analysis, TNBC has been classified into four different subtypes: basal-like immune-activated (BLIA), basal-like immune-suppressed (BLIS), mesenchymal (MES), and luminal androgen receptor (LAR). However, there is currently no standardized method for classifying TNBC subtypes. We attempted to define a gene signature for each subtype, and to develop a classification method based on machine learning (ML) for TNBC subtyping. In these experiments, gene expression microarray data for TNBC patients were downloaded from the Gene Expression Omnibus database. Differentially expressed genes unique to 198 known TNBC cases were identified and selected as a training gene set to train in seven different classification models. We produced a training set consisting of 719 DEGs selected from uniquely expressed genes of all four subtypes. The highest average accuracy of classification of the BLIA, BLIS, MES, and LAR subtypes was achieved by the SVM algorithm (accuracy 95-98.8%; AUC 0.99-1.00). For model validation, we used 334 samples of unknown TNBC subtypes, of which 97 (29.04%), 73 (21.86%), 39 (11.68%) and 59 (17.66%) were predicted to be BLIA, BLIS, MES, and LAR, respectively. However, 66 TNBC samples (19.76%) could not be assigned to any subtype. These samples contained only three upregulated genes (EN1, PROM1, and CCL2). Each TNBC subtype had a unique gene expression pattern, which was confirmed by identification of DEGs and pathway analysis. These results indicated that our training gene set was suitable for development of classification models, and that the SVM algorithm could classify TNBC into four unique subtypes. Accurate and consistent classification of the TNBC subtypes is essential for personalized treatment and prognosis of TNBC.

Non-Contact Assessment of Peripheral Artery Haemodynamics Using Infrared Video Thermography.

Skin temperature has long been used as a natural indicator of vascular diseases in the extremities. Considerable correlation between oscillations in skin surface temperature and oscillations of skin blood flow has previously been demonstrated. We hypothesised that the impairment of blood flow in stenotic (subcutaneous) peripheral arteries would influence cutaneous temperature such that, by measuring gradients in the temperature distribution over skin surfaces, one may be able to diagnose or quantify the progression of vascular conditions in whose pathogenesis a reduction in subcutaneous blood perfusion plays a critical role (e.g. peripheral artery disease). As proof of principle, this study investigates the local changes in the skin temperature of healthy humans (15 male, [Formula: see text] years old, BMI [Formula: see text] kg/m 2) undergoing two physical challenges designed to vary their haemodynamic status. Skin temperature was measured in four central regions (forehead, neck, chest, and left shoulder) and four peripheral regions (left upper arm, forearm, wrist, and hand) using an infrared thermal camera. We compare inter-region patterns. Median temperature over the peripheral regions decreased from baseline after both challenges (maximum decrease: [Formula: see text] °C at 60 s after exercise; [Formula: see text] and [Formula: see text] °C at 180 s of cold-water immersion; [Formula: see text]). Median temperature over the central regions showed no significant changes. Our results show that the non-contact measurement of perfusion-related changes in peripheral temperature from infrared video data is feasible. Further research will be directed towards the thermographic study of patients with symptomatic peripheral vascular disease.

Cardio-respiratory signal extraction from video camera data for continuous non-contact vital sign monitoring using deep learning.

Non-contact vital sign monitoring enables the estimation of vital signs, such as heart rate, respiratory rate and oxygen saturation (SpO2), by measuring subtle color changes on the skin surface using a video camera. For patients in a hospital ward, the main challenges in the development of continuous and robust non-contact monitoring techniques are the identification of time periods and the segmentation of skin regions of interest (ROIs) from which vital signs can be estimated. We propose a deep learning framework to tackle these challenges. APPROACH: This paper presents two convolutional neural network (CNN) models. The first network was designed for detecting the presence of a patient and segmenting the patient's skin area. The second network combined the output from the first network with optical flow for identifying time periods of clinical intervention so that these periods can be excluded from the estimation of vital signs. Both networks were trained using video recordings from a clinical study involving 15 pre-term infants conducted in the high dependency area of the neonatal intensive care unit (NICU) of the John Radcliffe Hospital in Oxford, UK. MAIN RESULTS: Our proposed methods achieved an accuracy of 98.8% for patient detection, a mean intersection-over-union (IOU) score of 88.6% for skin segmentation and an accuracy of 94.5% for clinical intervention detection using two-fold cross validation. Our deep learning models produced accurate results and were robust to different skin tones, changes in light conditions, pose variations and different clinical interventions by medical staff and family visitors. SIGNIFICANCE: Our approach allows cardio-respiratory signals to be continuously derived from the patient's skin during which the patient is present and no clinical intervention is undertaken.

Assessment of Signal Processing Methods for Measuring the Respiratory Rate in the Neonatal Intensive Care Unit.

Knowledge of the pathological instabilities in the breathing pattern can provide valuable insights into the cardiorespiratory status of the critically-ill infant as well as their maturation level. This paper is concerned with the measurement of respiratory rate in premature infants. We compare the rates estimated from the chest impedance pneumogram, the ECG-derived respiratory rhythms, and the PPG-derived respiratory rhythms against those measured in the reference standard of breath detection provided by attending clinical staff during 165 manual breath counts. We demonstrate that accurate RR estimates can be produced from all sources for RR in the 40-80 bpm (breaths per min) range. We also conclude that the use of indirect methods based on the ECG or the PPG poses a fundamental challenge in this population due to their poor behavior at fast breathing rates (upward of 80 bpm).

Non-contact physiological monitoring of preterm infants in the Neonatal Intensive Care Unit.

The implementation of video-based non-contact technologies to monitor the vital signs of preterm infants in the hospital presents several challenges, such as the detection of the presence or the absence of a patient in the video frame, robustness to changes in lighting conditions, automated identification of suitable time periods and regions of interest from which vital signs can be estimated. We carried out a clinical study to evaluate the accuracy and the proportion of time that heart rate and respiratory rate can be estimated from preterm infants using only a video camera in a clinical environment, without interfering with regular patient care. A total of 426.6 h of video and reference vital signs were recorded for 90 sessions from 30 preterm infants in the Neonatal Intensive Care Unit (NICU) of the John Radcliffe Hospital in Oxford. Each preterm infant was recorded under regular ambient light during daytime for up to four consecutive days. We developed multi-task deep learning algorithms to automatically segment skin areas and to estimate vital signs only when the infant was present in the field of view of the video camera and no clinical interventions were undertaken. We propose signal quality assessment algorithms for both heart rate and respiratory rate to discriminate between clinically acceptable and noisy signals. The mean absolute error between the reference and camera-derived heart rates was 2.3 beats/min for over 76% of the time for which the reference and camera data were valid. The mean absolute error between the reference and camera-derived respiratory rate was 3.5 breaths/min for over 82% of the time. Accurate estimates of heart rate and respiratory rate could be derived for at least 90% of the time, if gaps of up to 30 seconds with no estimates were allowed.

Non-invasive stroke volume estimation by transthoracic electrical bioimpedance versus Doppler echocardiography in healthy volunteers.

Thoracic electrical bioimpedance (TEB) and transthoracic echocardiography (TTE) are non-invasive methods to estimate stroke volume (SV) and cardiac output (CO). Thoracic electrical bioimpedance is not in widespread clinical use with reports of inaccurate cardiac output estimation compared to invasive monitors, particularly in non-healthy populations. We explore its use as a trend monitor by comparing it against thoracic echocardiography in fifteen healthy volunteers undergoing two physical challenges designed to vary cardiac output. Of all paired values, 54.6% showed gross trend agreement and only 1.9% showed direct disagreement between the two monitors. Our results show thoracic bioimpedance may have a role as a non-invasive cardiac output trend monitor in healthy volunteer studies.