A comprehensive benchmark for COVID-19 predictive modeling using electronic health records in intensive care.

The COVID-19 pandemic highlighted the need for predictive deep-learning models in health care. However, practical prediction task design, fair comparison, and model selection for clinical applications remain a challenge. To address this, we introduce and evaluate two new prediction tasks-outcome-specific length-of-stay and early-mortality prediction for COVID-19 patients in intensive care-which better reflect clinical realities. We developed evaluation metrics, model adaptation designs, and open-source data preprocessing pipelines for these tasks while also evaluating 18 predictive models, including clinical scoring methods and traditional machine-learning, basic deep-learning, and advanced deep-learning models, tailored for electronic health record (EHR) data. Benchmarking results from two real-world COVID-19 EHR datasets are provided, and all results and trained models have been released on an online platform for use by clinicians and researchers. Our efforts contribute to the advancement of deep-learning and machine-learning research in pandemic predictive modeling.

Mortality prediction with adaptive feature importance recalibration for peritoneal dialysis patients.

The study aims to develop AICare, an interpretable mortality prediction model, using electronic medical records (EMR) from follow-up visits for end-stage renal disease (ESRD) patients. AICare includes a multichannel feature extraction module and an adaptive feature importance recalibration module. It integrates dynamic records and static features to perform personalized health context representation learning. The dataset encompasses 13,091 visits and demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An additional public dataset of 4,789 visits from 1,363 hemodialysis (HD) patients is also considered. AICare outperforms traditional deep learning models in mortality prediction while retaining interpretability. It uncovers mortality-feature relationships and variations in feature importance and provides reference values. An AI-doctor interaction system is developed for visualizing patients' health trajectories and risk indicators.

The Combined Effects of Hourly Multi-Pollutant on the Risk of Ambulance Emergency Calls: A Seven-Year Time Series Study.

BACKGROUND: Ambulance emergency calls (AECs) are seen as a more suitable metric for syndromic surveillance due to their heightened sensitivity in reflecting the health impacts of air pollutants. Limited evidence has emphasized the combined effect of hourly air pollutants on AECs. This study aims to investigate the combined effects of multipollutants (i.e., PM2.5, PM10, Ozone, NO2, and SO2) on all-cause and cause-specific AECs by using the quantile g-computation method. METHODS: We used ambulance emergency dispatch data, air pollutant data, and meteorological data from between 1 January 2013 and 31 December 2019 in Shenzhen, China, to estimate the associations of hourly multipollutants with AECs. We followed a two-stage analytic protocol, including the distributed lag nonlinear model, to examine the predominant lag for each air pollutant, as well as the quantile g-computation model to determine the associations of air pollutant mixtures with all-cause and cause-specific AECs. RESULTS: A total of 3,022,164 patients were identified during the study period in Shenzhen. We found that each interquartile range increment in the concentrations of PM2.5, PM10, Ozone, NO2, and SO2 in 0-8 h, 0-8 h, 0-48 h, 0-28 h, and 0-24 h was associated with the highest risk of AECs. Each interquartile range increase in the mixture of air pollutants was significantly associated with a 1.67% (95% CI, 0.12-3.12%) increase in the risk of all-cause AECs, a 1.81% (95% CI, 0.25-3.39%) increase in the risk of vascular AECs, a 1.77% (95% CI, 0.44-3.11%) increase in reproductive AECs, and a 2.12% (95% CI, 0.56-3.71%) increase in AECs due to injuries. CONCLUSIONS: We found combined effects of pollutant mixtures associated with an increased risk of AECs across various causes. These findings highlight the importance of targeted policies and interventions to reduce air pollution, particularly for PM, Ozone, and NO2 emissions.

Associations Between Hourly Ambient Particulate Matter Air Pollution and Ambulance Emergency Calls: Time-Stratified Case-Crossover Study.

BACKGROUND: Associations between short-term exposure to ambient particulate matter (PM) air pollutants and mortality or hospital admissions have been well-documented in previous studies. Less is known about the associations of hourly exposure to PM air pollutants with ambulance emergency calls (AECs) for all causes and specific causes by conducting a case-crossover study. In addition, different patterns of AECs may be attributed to different seasons and daytime or nighttime periods. OBJECTIVE: In this study, we quantified the risk of all-cause and cause-specific AECs associated with hourly PM air pollutants between January 1, 2013, and December 31, 2019, in Shenzhen, China. We also examined whether the observed associations of PM air pollutants with AECs for all causes differed across strata defined by sex, age, season, and the time of day. METHODS: We used ambulance emergency dispatch data and environmental data between January 1, 2013, and December 31, 2019, from the Shenzhen Ambulance Emergency Centre and the National Environmental Monitor Station to conduct a time-stratified case-crossover study to estimate the associations of air pollutants (ie, PM with an aerodynamic diameter less than 2.5 µm [PM2.5] or 10 µm [PM10]) with all-cause and cause-specific AECs. We generated a well-established, distributed lag nonlinear model for nonlinear concentration response and nonlinear lag-response functions. We used conditional logistic regression to estimate odds ratios with 95% CIs, adjusted for public holidays, season, the time of day, the day of the week, hourly temperature, and hourly humidity, to examine the association of all-cause and cause-specific AECs with hourly air pollutant concentrations. RESULTS: A total of 3,022,164 patients were identified during the study period in Shenzhen. Each IQR increase in PM2.5 (24.0 µg/m3) and PM10 (34.0 µg/m3) concentrations over 24 hours was associated with an increased risk of AECs (PM2.5: all-cause, 1.8%, 95% CI 0.8%-2.4%; PM10: all-cause, 2.0%, 95% CI 1.1%-2.9%). We observed a stronger association of all-cause AECs with PM2.5 and PM10 in the daytime than in the nighttime (PM2.5: daytime, 1.7%, 95% CI 0.5%-3.0%; nighttime, 1.4%, 95% CI 0.3%-2.6%; PM10: daytime, 2.1%, 95% CI 0.9%-3.4%; nighttime, 1.7%, 95% CI 0.6%-2.8%) and in the older group than in the younger group (PM2.5: 18-64 years, 1.4%, 95% CI 0.6%-2.1%; ≥65 years, 1.6%, 95% CI 0.6%-2.6%; PM10: 18-64 years, 1.8%, 95% CI 0.9%-2.6%; ≥65 years, 2.0%, 95% CI 1.1%-3.0%). CONCLUSIONS: The risk of all-cause AECs increased consistently with increasing concentrations of PM air pollutants, showing a nearly linear relationship with no apparent thresholds. PM air pollution increase was associated with a higher risk of all-cause AECs and cardiovascular diseases-, respiratory diseases-, and reproductive illnesses-related AECs. The results of this study may be valuable to air pollution attributable to the distribution of emergency resources and consistent air pollution control.

Prediction model for elevated intraocular pressure risk after silicone oil filling based on clinical features.

Background: To evaluate risk factors and further develop prediction models for intraocular pressure elevation (IOP) after vitreoretinal surgery with silicone oil tamponade to support clinical management. Methods: A retrospective study analyzed 1,061 eyes of 1,061 consecutive patients that presented to the Jiangsu Province Hospital between December 2015 and December 2020, the IOP was measured from the preoperative visit and at the 1-week, 1-month, 3-month, and 6-month visits, and the final postoperative visit before silicone oil removal. Four machine learning methods were used to carried out the prediction of IOP elevation: Decision Tree, Logistic Regression, Random Forest, and Gradient-Boosted Decision Trees (GBDT) based on features including demographic and clinical characteristics, preoperative factors and surgical factors. Predictors were selected based on the p-value of the univariate analysis. Results: Elevated intraocular pressure developed in 26.01% of the eyes postoperatively. Elevated intraocular pressure primarily occurred within 1-2 weeks after surgery. Additionally, the majority of IOP values were distributed around 25-40 mmHg. GBDT utilizing features with p-values less than 0.5 from the hypothesis testing demonstrated the best predictive performance for 0.7944 in accuracy. The analysis revealed that age, sex, hypertension, diabetes, myopia, retinal detachment, lens status and biological parameters have predictive value. Conclusion: Age, sex, hypertension, diabetes, myopia, retinal detachment, lens status and biological parameters have influence on postoperative intraocular pressure elevation for patients with silicone oil tamponade after pars plana vitrectomy. The prediction model showed promising accuracy for the occurrence of IOP elevation. This may have some reference significance for reducing the incidence of high intraocular pressure after pars plana vitrectomy combined with silicone oil filling.

Estimating Urban Road GPS Environment Friendliness with Bus Trajectories: A City-Scale Approach †.

GPS is taken as the most prevalent positioning system in practice. However, in urban areas, as the GPS satellite signal could be blocked by buildings, the GPS positioning is not accurate due to multi-path errors. Estimating the negative impact of urban environments on GPS accuracy, that is the GPS environment friendliness (GEF) in this paper, will help to predict the GPS errors in different road segments. It enhances user experiences of location-based services and helps to determine where to deploy auxiliary assistant positioning devices. In this paper, we propose a method of processing and analysing massive historical bus GPS trajectory data to estimate the urban road GEF integrated with the contextual information of roads. First, our approach takes full advantage of the particular feature that bus routes are fixed to improve the performance of map matching. In order to estimate the GEF of all roads fairly and reasonably, the method estimates the GPS positioning error of each bus on the roads that are not covered by its route, by taking POIinformation, tag information of roads, and building layout information into account. Finally, we utilize a weighted estimation strategy to calculate the GEF of each road based on the GPS positioning performance of all buses. Based on one month of GPS trajectory data of 4835 buses within the second ring road in Chengdu, China, we estimate the GEF of 8831 different road segments and verify the rationality of the results by satellite maps, street views, and field tests.

How to quantify the relationship between spatial distribution of urban waterbodies and land surface temperature?

Urban waterbodies can effectively mitigate the increasing UHI effects and thus enhance climate resilience of urban areas. To contribute to our limited understanding in cooling effect of waterbodies on surrounding thermal environments, we examine the quantitative relationship between the spatial distribution of urban waterbodies and the land surface temperature (LST) in Wuhan, China. This paper 1) applies two indicators, the fractional water cover and the gravity water index, for measuring the spatial distribution of urban waterbodies; 2) conducts simple linear regression and spatial regression analyses to explore the LST-water relationship at multiple scales; and 3) compares the individual regression results from different land use types. The results show that the spatial distribution of urban waterbodies affects the LST significantly, and the gravity water index sufficiently explains the LST variation at various scales. Furthermore, the impact of urban waterbody distribution on the LST does vary across different land use types. Conclusions from this study provide insights of the cooling effect of urban waterbodies, which can further assist city planners and decision makers in utilizing cooling effects of waterbodies to improve the thermal environment of urban areas.

Mining Patterns of Disease Progression: A Topic-Model-Based Approach.

Knowledge of how diseases progress and transform is crucial for clinical decision making. Frequent pattern mining techniques, such as sequential pattern mining (SPM) algorithms, can automatically extract such knowledge from large collections of electronic medical records (EMR). However, EMR data are usually unorganized and highly noisy. Finding meaningful disease patterns often calls for manual manipulation such as cohort and feature selection on EMR data by medical professionals. In this paper, we propose a topic-model-based SPM approach to find disease progression patterns from diagnostic records. We improve the traditional SPM algorithms by filtering and grouping the diagnosis sequences according to different clinical topics. These topics represent certain clinical conditions with closely related diagnoses, and are detected without prior medical knowledge. The experiment on real-world EMR data shows that our approach is able to find meaningful progression patterns with less noises, and can help quickly identify interesting patterns related to a certain clinical condition with less human effort.

[A matched case-control study on the risk factors of metabolic syndrome among policemen].

OBJECTIVE: To investigate the life style, genetic and occupational risk factors of metabolic syndrome (MS) among policemen. METHODS: 1:4 matched case-control study was used, based on physical examination data of Tianjin Policemen in 2010, 708 patients with MS were randomly selected as cases, which were matched with 2832 healthy controls on the basis of sex and age (+/- 1 year). An epidemiological investigations on the past exposure status of several possible risk factors was conducted, and the data were analyzed with conditional logistic regression. RESULTS: Fifteen factors related to exposure were identified for MS through univariate conditional logistic regression analysis. Multivariate conditional logistic regression analysis suggested that, seven factors, such as family history of hypertension (OR = 2.406, 95% CI: 1.946-2.975), family history of diabetes (OR = 1.301, 95% CI: 1.043-1.623), smoking (OR = 1.357, 95%CI: 1.010-1.823), snoring (OR = 1.268, 95% CI: 1.043-1.543), work intensity (OR = 4.603, 95% CI: 3.767-5.623), occupational stressful events (OR = 1.524, 95% CI: 1.209-1.922), security policemen (OR = 1.453, 95% CI: 1.127-1.872) and criminal investigation policemen (OR = 2.792, 95% CI: 2.168-3.596), could significantly increase the risk of disease development, but dairy products (OR = 0.782, 95% CI: 0.619-0.989) was a protect factor for MS. The results from population attributable risk factors analysis showed that the control of smoking, snoring, work intensity, occupational stressful events can decreased the risk of MS to 16.26%, 11.71%, 56.87% and 8.97%, respectively. CONCLUSION: Metabolic syndrome has became a significant public health problem among policemen, it's necessary to take measures on life style, occupational risk factors for reducing the incidence of MS, and improving the health level among policemen.