Interpretable machine learning models for hospital readmission prediction: a two-step extracted regression tree approach.

BACKGROUND: Advanced machine learning models have received wide attention in assisting medical decision making due to the greater accuracy they can achieve. However, their limited interpretability imposes barriers for practitioners to adopt them. Recent advancements in interpretable machine learning tools allow us to look inside the black box of advanced prediction methods to extract interpretable models while maintaining similar prediction accuracy, but few studies have investigated the specific hospital readmission prediction problem with this spirit. METHODS: Our goal is to develop a machine-learning (ML) algorithm that can predict 30- and 90- day hospital readmissions as accurately as black box algorithms while providing medically interpretable insights into readmission risk factors. Leveraging a state-of-art interpretable ML model, we use a two-step Extracted Regression Tree approach to achieve this goal. In the first step, we train a black box prediction algorithm. In the second step, we extract a regression tree from the output of the black box algorithm that allows direct interpretation of medically relevant risk factors. We use data from a large teaching hospital in Asia to learn the ML model and verify our two-step approach. RESULTS: The two-step method can obtain similar prediction performance as the best black box model, such as Neural Networks, measured by three metrics: accuracy, the Area Under the Curve (AUC) and the Area Under the Precision-Recall Curve (AUPRC), while maintaining interpretability. Further, to examine whether the prediction results match the known medical insights (i.e., the model is truly interpretable and produces reasonable results), we show that key readmission risk factors extracted by the two-step approach are consistent with those found in the medical literature. CONCLUSIONS: The proposed two-step approach yields meaningful prediction results that are both accurate and interpretable. This study suggests a viable means to improve the trust of machine learning based models in clinical practice for predicting readmissions through the two-step approach.

Fatty acid metabolism and C9 aldehyde biosynthesis are involved in ε-poly-l-lysine-induced citrus fruit resistance to Penicillium digitatum.

Citrus fruit were easily infected by Penicillium digitatum, and caused green mold rapidly, resulting in enormous post-harvest losses. ε-poly-l-lysine (ε-PL) was generally regarded as a safe (GRAS) substance. Besides, it was proved to have a dual effect on harming fungi and triggering fruit defense responses. Fatty acid metabolism is closely related to fruit defense response. However, little is known about how ε-PL affected fatty acid metabolism in citrus fruit. Here, we found that ε-PL increased the expression of CsFATA, CsACSL, CsFAD2, CsFAD3, CsLOX2S, and CsHPL in fatty acid metabolism, decreasing oleic acid levels and enhancing linoleic and linolenic acid levels. Additionally, ε-PL enriched the activities of LOX and HPL during the oxidative decomposition of fatty acids, and activating C9 aldehyde biosynthesis. Interestingly, ε-PL combined with (2E,4E)-nonadienal (C9 aldehyde) would improve the inhibitory effect against Penicillium digitatum. And the combined bio-fungicide significantly delayed the citrus green mold compared to single concentrations of the individual components. These results suggested that ε-PL improved citrus fruit defense responses through fatty acid-mediated defense responses. Combined bio-fungicide consisting of ε-PL and (2E,4E)-nonadienal have an excellent prospect for controlling citrus green mold.

LEAP model-based analysis to low-carbon transformation path in the power sector: a case study of Guangdong-Hong Kong-Macao Greater Bay Area.

As a major carbon emitter, the power sector plays a crucial role in realizing the goal of carbon peaking and carbon neutrality. This study constructed a low-carbon power system based on the LEAP model (LEAP-GBA) with 2020 as a statistic base aiming of exploring the low-carbon transformation pathway of the power sector in the Guangdong-Hong Kong, and Macao Greater Bay Area (GBA). Five scenarios are set up to simulate the demand, power generation structure, carbon emissions, and power generation costs in the power sector under different scenarios. The results indicate that total electricity demand will peak after 2050, with 80% of it coming from industry, buildings and residential use. To achieve net-zero emissions from the power sector in the GBA, a future power generation mix dominated by nuclear and renewable energy generation and supplemented by fossil energy generation equipped with CCUS technologies. BECCS technology and nuclear power are the key to realize zero carbon emissions from the power sector in the GBA, so it should be the first to promote BECCS technology testing and commercial application, improve the deployment of nuclear power sites, and push forward the construction of nuclear power and technology improvement in the next 40 years.

Characteristics of the composite film of arabinoxylan and starch granules in simulated wheat endosperm.

We found that cell wall components of wheat grains differed significantly across different grain-filling stages; specifically, we observed significant differences in water content and water migration rate (p < 0.05). A composite film of arabinoxylan and starch granules was prepared to simulate wheat endosperm structure. Scanning electron microscopy (SEM), X-ray diffractometer (XRD), and thermogravimetric analysis (TGA) showed that the crystallinity and structural stability of the film increased with increasing starch content. Water diffusion experiments of the films revealed that the water diffusion rate gradually decreased with increasing starch content. Therefore, the water mobility of the starch endosperm was lower than that of the aleurone layer. These findings provide a basis for further studies in the context of wheat grain water regulation.

Effects of lamellar organization and arabinoxylan substitution rate on the properties of films simulating wheat grain aleurone cell wall.

Herein, we evaluated the properties of alternate arabinoxylan (AX)/(1→3) (1→4)-ß-D-glucan (BG) multilayer films. AX was extracted from wheat at three growth stages and single-component and alternate overlapping multilayer films were prepared. The physical properties, water diffusion rate, and water mobility of multilayer films during water absorption and desorption were studied. There were significant differences in the AX content and arabinose-to-xylose ratio at different growth stages. The LAX/BG multilayer films showed excellent thermal stability and mechanical properties with an increase in the relative humidity. The AX multilayer films with a low substitution rate showed a better water-binding capacity, whereas water molecules in films with a high substitution rate showed higher mobility. Therefore, a low substitution rate AX and AX/BG composite structure can improve the thermodynamic properties of multilayer films, but limit water mobility. We provide new insights on the physicochemical properties and water-regulation effects of wheat cell wall.

Variation in cell wall structure and composition of wheat grain based on geography and regulatory effect of cell wall on water mobility.

Wheat grain from 12 different regions in China was used to study variations in the cell wall structure and chemical composition based on geography. The mobility and migration rate of water in wheat grain during moisture absorption and drying were determined under different relative humidity conditions. Depending on the geography, variations were noted in the thickness and component content of the wheat grain cell wall. Cell wall thickness was positively correlated with the total arabinoxylan (TAX) content. Cell wall thickness and TAX content of the aleurone layer were positively correlated with altitude and negatively correlated with longitude. The water migration rate decreased with the increase of cell wall thickness and TAX content. Nuclear magnetic resonance (NMR) results revealed that grains with thick aleurone cell wall showed increased molecular mobility of water. These findings lay the foundation for further study of water regulation in wheat cell wall.