Machine Learning-Based Prediction of Malnutrition in Surgical In-Patients: A Validation Pilot Study.

BACKGROUND: Malnutrition in hospitalised patients can lead to serious complications, worse patient outcomes and longer hospital stays. State-of-the-art screening methods rely on scores, which need additional manual assessments causing higher workload. OBJECTIVES: The aim of this prospective study was to validate a machine learning (ML)-based approach for an automated prediction of malnutrition in hospitalised patients. METHODS: For 159 surgical in-patients, an assessment of malnutrition by dieticians was compared to the ML-based prediction conducted in the evening of admission. RESULTS: The model achieved an accuracy of 83.0% and an AUROC of 0.833 in the prospective validation cohort. CONCLUSION: The results of this pilot study indicate that an automated malnutrition screening could replace manual screening tools in hospitals.

Machine Learning Based Risk Prediction for Major Adverse Cardiovascular Events for ELGA-Authorized Clinics1.

BACKGROUND: Artificial Intelligence (AI) has had an important impact on many industries as well as the field of medical diagnostics. In healthcare, AI techniques such as case-based reasoning and data driven machine learning (ML) algorithms have been used to support decision-making processes for complex tasks. This is used to assist medical professionals in making clinical decisions. A way of supporting clinicians is providing predicted prognoses of various ML models. OBJECTIVES: Training an ML model based on the data of a hospital and using it on another hospital have some challenges. METHODS: In this research, we applied data analysis to discover required data filters on a hospital's EHR data for training a model for another hospital. RESULTS: We applied experiments on real-world data of ELGA (Austrian health record system) and KAGes (a public healthcare provider of 20+ hospitals in Austria). In this scenario, we train the prediction model for ELGA- authorized health service providers using the KAGes data since we do not have access to the complete ELGA data. CONCLUSION: Finally, we observed that filtering the data with both feature and value selection increases the classification performance of the prediction model, which is trained for another system.

Frail People in LABLand: Development of an Easy-to-Use Machine Learning Model to Identify Frail People in Hospitals Based on Laboratory Data.

BACKGROUND: Frail individuals are very vulnerable to stressors, which often lead to adverse outcomes. To ensure an adequate therapy, a holistic diagnostic approach is needed which is provided in geriatric wards. It is important to identify frail individuals outside the geriatric ward as well to ensure that they also benefit from the holistic approach. OBJECTIVES: The goal of this study was to develop a machine learning model to identify frail individuals in hospitals. The model should be applicable without additional effort, quickly and in many different places in the healthcare system. METHODS: We used Gradient Boosting Decision Trees (GBDT) to predict a frailty target derived from a gold standard assessment. The used features were laboratory values, age and sex. We also identified the most important features. RESULTS: The best GBDT achieved an AUROC of 0.696. The most important laboratory values are urea, creatinine, granulocytes, chloride and calcium. CONCLUSION: The model performance is acceptable, but insufficient for clinical use. Additional laboratory values or the laboratory history could improve the performance.

Development of an Architecture to Implement Machine Learning Based Risk Prediction in Clinical Routine: A Service-Oriented Approach.

BACKGROUND: Patients at risk of developing a disease have to be identified at an early stage to enable prevention. One way of early detection is the use of machine learning based prediction models trained on electronic health records. OBJECTIVES: The aim of this project was to develop a software solution to predict cardiovascular and nephrological events using machine learning models. In addition, a risk verification interface for health care professionals was established. METHODS: In order to meet the requirements, different tools were analysed. Based on this, a software architecture was created, which was designed to be as modular as possible. RESULTS: A software was realised that is able to automatically calculate and display risks using machine learning models. Furthermore, predictions can be verified via an interface adapted to the need of health care professionals, which shows data required for prediction. CONCLUSION: Due to the modularised software architecture and the status-based calculation process, different technologies could be applied. This facilitates the installation of the software at multiple health care providers, for which adjustments need to be carried out at one part of the software only.

Machine Learning Based Risk Prediction for Major Adverse Cardiovascular Events.

BACKGROUND: Patients with major adverse cardiovascular events (MACE) such as myocardial infarction or stroke suffer from frequent hospitalizations and have high mortality rates. By identifying patients at risk at an early stage, MACE can be prevented with the right interventions. OBJECTIVES: The aim of this study was to develop machine learning-based models for the 5-year risk prediction of MACE. METHODS: The data used for modelling included electronic medical records of more than 128,000 patients including 29,262 patients with MACE. A feature selection based on filter and embedded methods resulted in 826 features for modelling. Different machine learning methods were used for modelling on the training data. RESULTS: A random forest model achieved the best calibration and discriminative performance on a separate test data set with an AUROC of 0.88. CONCLUSION: The developed risk prediction models achieved an excellent performance in the test data. Future research is needed to determine the performance of these models and their clinical benefit in prospective settings.

Black carbon and other light-absorbing impurities in snow in the Chilean Andes.

Vertical profiles of black carbon (BC) and other light-absorbing impurities were measured in seasonal snow and permanent snowfields in the Chilean Andes during Austral winters 2015 and 2016, at 22 sites between latitudes 18°S and 41°S. The samples were analyzed for spectrally-resolved visible light absorption. For surface snow, the average mass mixing ratio of BC was 15 ng/g in northern Chile (18-33°S), 28 ng/g near Santiago (a major city near latitude 33°S, where urban pollution plays a significant role), and 13 ng/g in southern Chile (33-41°S). The regional average vertically-integrated loading of BC was 207 µg/m2 in the north, 780 µg/m2 near Santiago, and 2500 µg/m2 in the south, where the snow season was longer and the snow was deeper. For samples collected at locations where there had been no new snowfall for a week or more, the BC concentration in surface snow was high (~10-100 ng/g) and the sub-surface snow was comparatively clean, indicating the dominance of dry deposition of BC. Mean albedo reductions due to light-absorbing impurities were 0.0150, 0.0160, and 0.0077 for snow grain radii of 100 µm for northern Chile, the region near Santiago, and southern Chile; respective mean radiative forcings for the winter months were 2.8, 1.4, and 0.6 W/m2. In northern Chile, our measurements indicate that light-absorption by impurities in snow was dominated by dust rather than BC.

Impact of Orientation on the Vitamin D Weighted Exposure of a Human in an Urban Environment.

The vitamin D3-weighted UV exposure of a human with vertical posture was calculated for urban locations to investigate the impact of orientation and obstructions on the exposure. Human exposure was calculated by using the 3D geometry of a human and integrating the radiance, i.e., the radiant energy from the direct solar beam and the diffuse sky radiation from different incident and azimuth angles. Obstructions of the sky are derived from hemispherical images, which are recorded by a digital camera with a fisheye lens. Due to the low reflectivity of most surfaces in the UV range, the radiance from obstructed sky regions was neglected. For spring equinox (21 March), the exposure of a human model with winter clothing in an environment where obstructions cover 40% of the sky varies by up to 25%, depending on the orientation of the human model to the sun. The calculation of the accumulated vitamin D3-weighted exposure of a human with winter clothing walking during lunch break shows that human exposure is reduced by the obstruction of buildings and vegetation by 40%.

Is Multidirectional UV Exposure Responsible for Increasing Melanoma Prevalence with Altitude? A Hypothesis Based on Calculations with a 3D-Human Exposure Model.

In a recent study, melanoma incidence rates for Austrian inhabitants living at higher altitudes were found to increase by as much as 30% per 100 m altitude. This strong increase cannot simply be explained by the known increase of erythemally-weighted irradiance with altitude, which ranges between 0.5% and 4% per 100 m. We assume that the discrepancy is partially explainable by upwelling UV radiation; e.g., reflected by snow-covered surfaces. Therefore, we present an approach where the human UV exposure is derived by integrating incident radiation over the 3D geometry of a human body, which enables us to take upwelling radiation into account. Calculating upwelling and downwelling radiance with a radiative transfer model for a snow-free valley and for snow-covered mountain terrain (with albedo of 0.6) yields an increase in UV exposure by 10% per 100 m altitude. The results imply that upwelling radiation plays a significant role in the increase of melanoma incidence with altitude.

Solar Simulators for Healthy Vitamin D Synthesis.

BACKGROUND/AIM: The angular distribution of solar radiance and its spectral characteristics is required for the determination of vitamin D3 production in humans. MATERIALS AND METHODS: The vitamin D3 weighted exposure can be calculated by integrating the incident solar spectral radiance over all relevant parts of the human body. A novel instrument allowing simultaneous measurements of spectral radiance from more than 100 directions has been developed. A large solar simulator for controlled experiments is described. RESULTS: In summer it is relatively easy to obtain sufficient vitamin D because sun exposure times are short. In winter solstice vitamin D3 cannot be obtained with realistic clothing even if the exposure were extended to all daylight hours. CONCLUSION: Improved and controlled experiments to determine vitamin D3 production are required to assess the positive effects of solar UV radiation and to assess its natural variability.

Measuring high-resolution sky luminance distributions with a CCD camera.

We describe how sky luminance can be derived from a newly developed hemispherical sky imager (HSI) system. The system contains a commercial compact charge coupled device (CCD) camera equipped with a fish-eye lens. The projection of the camera system has been found to be nearly equidistant. The luminance from the high dynamic range images has been calculated and then validated with luminance data measured by a CCD array spectroradiometer. The deviation between both datasets is less than 10% for cloudless and completely overcast skies, and differs by no more than 20% for all sky conditions. The global illuminance derived from the HSI pictures deviates by less than 5% and 20% under cloudless and cloudy skies for solar zenith angles less than 80°, respectively. This system is therefore capable of measuring sky luminance with the high spatial and temporal resolution of more than a million pixels and every 20 s respectively.

A novel method to calculate solar UV exposure relevant to vitamin D production in humans.

We present a novel method to calculate vitamin D3 -weighted exposure by integrating the incident solar spectral radiance over all relevant parts of the human body. Earlier investigations are based on the irradiance on surfaces, whereas our calculated exposure of a voxel model of a human takes into account the complex geometry of the radiation field. Assuming that sufficient vitamin D3 (1000 international units) can be produced within the human body in one minute for a completely uncovered body in vertical posture in summer at midlatitudes (e.g. Rome, June 21, noon, UV index of 10), we calculate the exposure times needed in other situations or seasons to gain enough vitamin D3 . Our calculations show that the UV index is not a good indicator for the exposure which depends on the orientation of the body (e.g. vertical (standing) or horizontal (lying down) posture). Without clothing the exposure is dominated by diffuse sky radiation and it is nearly irrelevant how the body in vertical posture is oriented toward the sun. At the winter solstice (December 21, noon, cloudy) at least in central Europe sufficient vitamin D3 cannot be obtained with realistic clothing, even if the exposure were extended to all daylight hours.