A method for hand-foot-mouth disease prediction using GeoDetector and LSTM model in Guangxi, China

Dual-attention-based recurrent neural network for hand-foot-mouth disease prediction in Korea

Hand-foot-mouth disease (HFMD) is a viral disease that occurs primarily in children. Meteorological factors have a significant impact on its popularity annually in Korea. This study proposes a new HFMD prediction model using a dual-attention-based recurrent neural network (DA-RNN) and important weather factors for HFMD in Korea. First, suspected cases of HFMD in each state were predicted using meteorological factors from the DA-RNN. Second, the weather factors were divided into six categories: temperature, wind, rainfall, day length, humidity, and air pollution to conduct sensitivity analysis. Because of this prediction, the proposed model showed the best performance in predicting the number of suspected HFMD cases in a week compared with other RNN methods. Sensitivity analysis showed that air pollution and rainfall play an important role in HFMD in Korea. This model provides information for HFMD prevention and control and can be extended to predict other infectious diseases.

Prediction of outpatients with conjunctivitis in Xinjiang based on LSTM and GRU models

BACKGROUND

Reasonable and accurate forecasting of outpatient visits helps hospital managers optimize the allocation of medical resources, facilitates fine hospital management, and is of great significance in improving hospital efficiency and treatment capacity.

METHODS

Based on conjunctivitis outpatient data from the First Affiliated Hospital of Xinjiang Medical University Ophthalmology from 2017/1/1 to 2019/12/31, this paper built and evaluated Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models for outpatient visits prediction.

RESULTS

In predicting the number of conjunctivitis visits over the next 31 days, the LSTM model had a root mean square error (RMSE) of 2.86 and a mean absolute error (MAE) of 2.39, the GRU model has an RMSE of 2.60 and an MAE of 1.99.

CONCLUSIONS

The GRU method can better predict trends in hospital outpatient flow over time, thus providing decision support for medical staff and outpatient management.

Study of the influence of meteorological factors on HFMD and prediction based on the LSTM algorithm in Fuzhou, China

BACKGROUND

This study adopted complete meteorological indicators, including eight items, to explore their impact on hand, foot, and mouth disease (HFMD) in Fuzhou, and predict the incidence of HFMD through the long short-term memory (LSTM) neural network algorithm of artificial intelligence.

METHOD

A distributed lag nonlinear model (DLNM) was used to analyse the influence of meteorological factors on HFMD in Fuzhou from 2010 to 2021. Then, the numbers of HFMD cases in 2019, 2020 and 2021 were predicted using the LSTM model through multifactor single-step and multistep rolling methods. The root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and symmetric mean absolute percentage error (SMAPE) were used to evaluate the accuracy of the model predictions.

RESULTS

Overall, the effect of daily precipitation on HFMD was not significant. Low (4 hPa) and high (≥ 21 hPa) daily air pressure difference (PRSD) and low (< 7 °C) and high (> 12 °C) daily air temperature difference (TEMD) were risk factors for HFMD. The RMSE, MAE, MAPE and SMAPE of using the weekly multifactor data to predict the cases of HFMD on the following day, from 2019 to 2021, were lower than those of using the daily multifactor data to predict the cases of HFMD on the following day. In particular, the RMSE, MAE, MAPE and SMAPE of using weekly multifactor data to predict the following week's daily average cases of HFMD were much lower, and similar results were also found in urban and rural areas, which indicating that this approach was more accurate.

CONCLUSION

This study's LSTM models combined with meteorological factors (excluding PRE) can be used to accurately predict HFMD in Fuzhou, especially the method of predicting the daily average cases of HFMD in the following week using weekly multifactor data.

Forecasting upper respiratory tract infection burden using high-dimensional time series data and forecast combinations

Upper respiratory tract infections (URTIs) represent a large strain on primary health resources. To mitigate URTI transmission and public health burdens, it is important to pre-empt and provide forward guidance on URTI burden, while taking into account various facets which influence URTI transmission. This is so that appropriate public health measures can be taken to mitigate strain on primary care resources. This study describes a new approach to forecasting URTIs which can be used for national public health resource planning. Specifically, using environmental and disease data comprising more than 1000 dimensions, we developed sub-models which optimizes model explainability, in-sample model fit, predictive accuracy and combines many weaker predictors over a 2-month time horizon to generate direct, point forecasts over a 1-8 week ahead forecast horizon. Predictive performance was evaluated using rolling out-of-sample forecast assessment within both periods with/without structural breaks in transmission over the period of 2012-2022. We showed that forecast combinations of 5 other forecasting models had better and more consistent predictive performance than other modelling approaches, over periods with and without structural breaks in transmission dynamics. Furthermore, epidemiological analysis on high dimensional data was enabled using post-selection inference, to show the dynamic association between lower temperature, increases in past relative humidity and absolute humidity and increased URTIs attendance. The methods proposed can be used for outbreak preparedness and guide healthcare resource planning, in both stable periods of transmission and periods where structural breaks in data occur.

A comparative study of three models to analyze the impact of air pollutants on the number of pulmonary tuberculosis cases in Urumqi, Xinjiang

In this paper, we separately constructed ARIMA, ARIMAX, and RNN models to determine whether there exists an impact of the air pollutants (such as PM2.5, PM10, CO, O3, NO2, and SO2) on the number of pulmonary tuberculosis cases from January 2014 to December 2018 in Urumqi, Xinjiang. In addition, by using a new comprehensive evaluation index DISO to compare the performance of three models, it was demonstrated that ARIMAX (1,1,2) × (0,1,1)12 + PM2.5 (lag = 12) model was the optimal one, which was applied to predict the number of pulmonary tuberculosis cases in Urumqi from January 2019 to December 2019. The predicting results were in good agreement with the actual pulmonary tuberculosis cases and shown that pulmonary tuberculosis cases obviously declined, which indicated that the policies of environmental protection and universal health checkups in Urumqi have been very effective in recent years.

Impact of Tree Cover Loss on Carbon Emission: A Learning-Based Analysis

Describing the processes leading to deforestation is essential for the development and implementation of the forest policies. In this work, two different learning models were developed in order to identify the best possible model for the assessment of the deforestation causes and trends. We developed autoregressive integrated moving average (ARIMA) model and long short-term memory (LSTM) independently in order to see the trend between tree cover loss and carbon dioxide emission. This study includes the twenty-year data of Pakistan on tree cover loss and carbon emission from the Global Forest Watch (GFW) platform, a known platform to get numerical data. Minimum mean absolute error (MAE) for the prediction of tree cover loss and carbon emission obtained through ARIMA model is 0.89 and 0.95, respectively. The minimum MAE given by LSTM model is 0.33 and 0.43, respectively. There is no such kind of study conducted in order to identify the increase in carbon emission due to tree cover loss most specifically in Pakistan. The results endorsed that one of the main causes of increase in the pollution in the environment in terms of carbon emission is due to tree cover loss.

A Hybrid Model for Coronavirus Disease 2019 Forecasting Based on Ensemble Empirical Mode Decomposition and Deep Learning

The novel coronavirus pneumonia that began to spread in 2019 is still raging and has placed a burden on medical systems and governments in various countries. For policymaking and medical resource decisions, a good prediction model is necessary to monitor and evaluate the trends of the epidemic. We used a long short-term memory (LSTM) model and the improved hybrid model based on ensemble empirical mode decomposition (EEMD) to predict COVID-19 trends; Methods: The data were collected from the Harvard Dataverse. Epidemic data from 21 January 2020 to 25 April 2021 for California, the most severely affected state in the United States, were used to develop an LSTM model and an EEMD-LSTM hybrid model, which is an LSTM model combined with ensemble empirical mode decomposition. In this study, ninety percent of the data were adopted to fit the models as a training set, while the subsequent 10% were used to test the prediction effect of the models. The mean absolute percentage error, mean absolute error, and root mean square error were used to evaluate the prediction performances of the models; Results: The results indicated that the number of confirmed cases in California was increasing as of 25 April 2021, with no obvious evidence of a sharp decline. On 25 April 2021, the LSTM model predicted 3666418 confirmed cases, whereas the EEMD-LSTM predicted 3681150. The mean absolute percentage errors for the LSTM and EEMD-LSTM models were 0.0151 and 0.0051, respectively. The mean absolute and root mean square errors were 5.58 × 10⁴ and 5.63 × 10⁴ for the LSTM model and 1.9 × 10⁴ and 2.43 × 10⁴ for the EEMD-LSTM model, respectively; Conclusions: The results showed the advantage of an EEMD-LSTM model over a single LSTM model, and the established EEMD-LSTM model may be suitable for monitoring and evaluating the epidemic situation and providing quantitative analysis evidence for epidemic prevention and control.

Study on the influence of meteorological factors on influenza in different regions and predictions based on an LSTM algorithm

BACKGROUND

Influenza epidemics pose a threat to human health. It has been reported that meteorological factors (MFs) are associated with influenza. This study aimed to explore the similarities and differences between the influences of more comprehensive MFs on influenza in cities with different economic, geographical and climatic characteristics in Fujian Province. Then, the information was used to predict the daily number of cases of influenza in various cities based on MFs to provide bases for early warning systems and outbreak prevention.

METHOD

Distributed lag nonlinear models (DLNMs) were used to analyse the influence of MFs on influenza in different regions of Fujian Province from 2010 to 2021. Long short-term memory (LSTM) was used to train and model daily cases of influenza in 2010-2018, 2010-2019, and 2010-2020 based on meteorological daily values. Daily cases of influenza in 2019, 2020 and 2021 were predicted. The root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and symmetric mean absolute percentage error (SMAPE) were used to quantify the accuracy of model predictions.

RESULTS

The cumulative effect of low and high values of air pressure (PRS), air temperature (TEM), air temperature difference (TEMD) and sunshine duration (SSD) on the risk of influenza was obvious. Low (< 979 hPa), medium (983 to 987 hPa) and high (> 112 hPa) PRS were associated with a higher risk of influenza in women, children aged 0 to 12 years, and rural populations. Low (< 9 °C) and high (> 23 °C) TEM were risk factors for influenza in four cities. Wind speed (WIN) had a more significant effect on the risk of influenza in the ≥ 60-year-old group. Low (< 40%) and high (> 80%) relative humidity (RHU) in Fuzhou and Xiamen had a significant effect on influenza. When PRS was between 1005-1015 hPa, RHU > 60%, PRE was low, TEM was between 10-20 °C, and WIN was low, the interaction between different MFs and influenza was most obvious. The RMSE, MAE, MAPE, and SMAPE evaluation indices of the predictions in 2019, 2020 and 2021 were low, and the prediction accuracy was high.

CONCLUSION

All eight MFs studied had an impact on influenza in four cities, but there were similarities and differences. The LSTM model, combined with these eight MFs, was highly accurate in predicting the daily cases of influenza. These MFs and prediction models could be incorporated into the influenza early warning and prediction system of each city and used as a reference to formulate prevention strategies for relevant departments.

Forecasting of non-accidental, cardiovascular, and respiratory mortality with environmental exposures adopting machine learning approaches

Environmental exposure constantly changes with time and various interactions that can affect health outcomes. Machine learning (ML) or deep learning (DL) algorithms have been used to solve complex problems, such as multiple exposures and their interactions. This study developed predictive models for cause-specific mortality using ML and DL algorithms with the daily or hourly measured meteorological and air pollution data. The ML algorithm improved the performance compared to the conventional methods, even though the optimal algorithm depended on the adverse health outcomes. The best algorithms were extreme gradient boosting, ridge, and elastic net, respectively, for non-accidental, cardiovascular, and respiratory mortality with daily measurement; they were superior to the generalized additive model reducing a mean absolute error by 4.7%, 4.9%, and 16.8%, respectively. With hourly measurements, the ML model tended to outperform the conventional models, even though hourly data, instead of daily data, did not enhance the performance in some models. The proposed model allows a better understanding and development of robust predictive models for health outcomes using multiple environmental exposures.

Comparison of ARIMA model, DNN model and LSTM model in predicting disease burden of occupational pneumoconiosis in Tianjin, China

BACKGROUND

This study aims to explore appropriate model for predicting the disease burden of pneumoconiosis in Tianjin by comparing the prediction effects of Autoregressive Integrated Moving Average (ARIMA) model, Deep Neural Networks (DNN) model and multivariate Long Short-Term Memory Neural Network (LSTM) models.

METHODS

Disability adjusted life year (DALY) was used to evaluate the disease burden of occupational pneumoconiosis. ARIMA model, DNN model and multivariate LSTM model were used to establish prediction model. Three performance evaluation metrics including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) were used to compare the prediction effects of the three models.

RESULTS

From 1990 to 2021, there were 10,694 cases of pneumoconiosis patients in Tianjin, resulting in a total of 112,725.52 person-years of DALY. During this period, the annual DALY showed a fluctuating trend, but it had a strong correlation with the number of pneumoconiosis patients, the average age of onset, the average age of receiving dust and the gross industrial product, and had a significant nonlinear relationship with them. The comparison of prediction results showed that the performance of multivariate LSTM model and DNN model is much better than that of traditional ARIMA model. Compared with the DNN model, the multivariate LSTM model performed better in the training set, showing lower RMES (42.30 vs. 380.96), MAE (29.53 vs. 231.20) and MAPE (1.63% vs. 2.93%), but performed less stable than the DNN on the test set, showing slightly higher RMSE (1309.14 vs. 656.44), MAE (886.98 vs. 594.47) and MAPE (36.86% vs. 22.43%).

CONCLUSION

The machine learning techniques of DNN and LSTM are an innovative method to accurately and efficiently predict the burden of pneumoconiosis with the simplest data. It has great application prospects in the monitoring and early warning system of occupational disease burden.

Predicting the Number of Reported Pulmonary Tuberculosis in Guiyang, China, Based on Time Series Analysis Techniques

Tuberculosis (TB) is one of the world's deadliest infectious disease killers today, and despite China's increasing efforts to prevent and control TB, the TB epidemic is still very serious. In the context of the COVID-19 pandemic, if reliable forecasts of TB epidemic trends can be made, they can help policymakers with early warning and contribute to the prevention and control of TB. In this study, we collected monthly reports of pulmonary tuberculosis (PTB) in Guiyang, China, from January 1, 2010 to December 31, 2020, and monthly meteorological data for the same period, and used LASSO regression to screen four meteorological factors that had an influence on the monthly reports of PTB in Guiyang, including sunshine hours, relative humidity, average atmospheric pressure, and annual highest temperature, of which relative humidity (6-month lag) and average atmospheric pressure (7-month lag) have a lagging effect with the number of TB reports in Guiyang. Based on these data, we constructed ARIMA, Holt-Winters (additive and multiplicative), ARIMAX (with meteorological factors), LSTM, and multivariable LSTM (with meteorological factors). We found that the addition of meteorological factors significantly improved the performance of the time series prediction model, which, after comprehensive consideration, included the ARIMAX (1,1,1) (0,1,2)₁₂ model with a lag of 7 months at the average atmospheric pressure, outperforms the other models in terms of both fit (RMSE = 37.570, MAPE = 10.164%, MAE = 28.511) and forecast sensitivity (RMSE = 20.724, MAPE = 6.901%, MAE = 17.306), so the ARIMAX (1,1,1) (0,1,2)₁₂ model with a lag of 7 months can be used as a predictor tool for predicting the number of monthly reports of PTB in Guiyang, China.

Vulnerability to typhoons: A comparison of consequence and driving factors between Typhoon Hato (2017) and Typhoon Mangkhut (2018)

Typhoon disasters have caused casualties, property loss, and other negative impacts to social and economic development. Vulnerability is an important component of typhoon risk. However, little is known about the contributions of vulnerability factors and their interaction effects on typhoon-induced losses at a fine scale. Focusing on the vulnerability measures of Typhoon Hato in 2017 and Typhoon Mangkhut in 2018, this study aims to quantify the contribution and interactive effects of physical and socioeconomic factors on vulnerability based on the GeoDetector method and determine the factors that account for most of the change in vulnerability. The results show that from Typhoon Hato in 2017 to Typhoon Mangkhut in 2018, the vulnerability of the economy and houses decrease on average. Rain intensity and wind intensity are the dominant factors of disaster loss for Typhoon Hato and Typhoon Mangkhut, respectively. Vegetation cover and landform explain vulnerability better than average slope in most instances. For different loss types, the dominant socioeconomic vulnerability factor is different. For both typhoons, emergency transfer has a higher determining power (q) ranking for the population vulnerability, while the percentage of the GDP made up of primary industry have higher q ranking for economic vulnerability. The dominant interaction effects between two vulnerability factors differ depending on the typhoon and loss type but show a nonlinear enhancement effect in most cases. Moreover, changes in the maximum 4-hour accumulated rainfall account for most of the change in vulnerability between Hato and Mangkhut. Overall, the results can be conducive to understanding the complexity of vulnerability to typhoons and provide a reference for possible indicators for vulnerability assessment models, and determining the reasons for changes in vulnerability can be constructive to the formulation of specific policies for disaster prevention and mitigation.

Early Warning of Infectious Diseases in Hospitals Based on Multi-Self-Regression Deep Neural Network

Objective. Infectious diseases usually spread rapidly. This study aims to develop a model that can provide fine-grained early warnings of infectious diseases using real hospital data combined with disease transmission characteristics, weather, and other multi-source data. Methods. Based on daily data reported for infectious diseases collected from several large general hospitals in China between 2012 and 2020, seven common infectious diseases in medical institutions were screened and a multi self-regression deep (MSRD) neural network was constructed. Using a recurrent neural network as the basic structure, the model can effectively model the epidemiological trend of infectious diseases by considering the current influencing conditions while taking into account the historical development characteristics in time-series data. The fitting and prediction accuracy of the model were evaluated using mean absolute error (MAE) and root mean squared error. Results. The proposed approach is significantly better than the existing infectious disease dynamics model, susceptible-exposed-infected-removed (SEIR), as it addresses the concerns of difficult-to-obtain quantitative data such as latent population, overfitting of long time series, and considering only a single series of the number of sick people without considering the epidemiological characteristics of infectious diseases. We also compare certain machine learning methods in this study. Experimental results demonstrate that the proposed approach achieves an MAE of 0.6928 and 1.3782 for hand, foot, and mouth disease and influenza, respectively. Conclusion. The MRSD-based infectious disease prediction model proposed in this paper can provide daily and instantaneous updates and accurate predictions for epidemic trends.

Prediction of hand, foot, and mouth disease epidemics in Japan using a long short-term memory approach

Hand, foot, and mouth disease (HFMD) is a common febrile illness caused by enteroviruses in the Picornaviridae family. The major symptoms of HFMD are fever and a vesicular rash on the hand, foot, or oral mucosa. Acute meningitis and encephalitis are observed in rare cases. HFMD epidemics occur annually in Japan, usually in the summer season. Relatively large-scale outbreaks have occurred every two years since 2011. In this study, the epidemic patterns of HFMD in Japan are predicted four weeks in advance using a deep learning method. The time-series data were analyzed by a long short-term memory (LSTM) approach called a Recurrent Neural Network. The LSTM model was trained on the numbers of weekly HFMD cases in each prefecture. These data are reported in the Infectious Diseases Weekly Report, which compiles the national surveillance data from web sites at the National Institute of Infectious Diseases, Japan, under the Infectious Diseases Control Law. Consequently, our trained LSTM model distinguishes between relatively large-scale and small-scale epidemics. The trained model predicted the HFMD epidemics in 2018 and 2019, indicating that the LSTM approach can estimate the future epidemic patterns of HFMD in Japan.

Spatio-temporal stratified associations between urban human activities and crime patterns: a case study in San Francisco around the COVID-19 stay-at-home mandate

Crime changes have been reported as a result of human routine activity shifting due to containment policies, such as stay-at-home (SAH) mandates during the COVID-19 pandemic. However, the way in which the manifestation of crime in both space and time is affected by dynamic human activities has not been explored in depth in empirical studies. Here, we aim to quantitatively measure the spatio-temporal stratified associations between crime patterns and human activities in the context of an unstable period of the ever-changing socio-demographic backcloth. We propose an analytical framework to detect the stratified associations between dynamic human activities and crimes in urban areas. In a case study of San Francisco, United States, we first identify human activity zones (HAZs) based on the similarity of daily footfall signatures on census block groups (CBGs). Then, we examine the spatial associations between crime spatial distributions at the CBG-level and the HAZs using spatial stratified heterogeneity statistical measurements. Thirdly, we use different temporal observation scales around the effective date of the SAH mandate during the COVID-19 pandemic to investigate the dynamic nature of the associations. The results reveal that the spatial patterns of most crime types are statistically significantly associated with that of human activities zones. Property crime exhibits a higher stratified association than violent crime across all temporal scales. Further, the strongest association is obtained with the eight-week time span centred around the SAH order. These findings not only enhance our understanding of the relationships between urban crime and human activities, but also offer insights into that tailored crime intervention strategies need to consider human activity variables.

Deep learning time series prediction models in surveillance data of hepatitis incidence in China

Background

Precise incidence prediction of Hepatitis infectious disease is critical for early prevention and better government strategic planning. In this paper, we presented different prediction models using deep learning methods based on the monthly incidence of Hepatitis through a national public health surveillance system in China mainland.

Methods

We assessed and compared the performance of three deep learning methods, namely, Long Short-Term Memory (LSTM) prediction model, Recurrent Neural Network (RNN) prediction model, and Back Propagation Neural Network (BPNN) prediction model. The data collected from 2005 to 2018 were used for the training and prediction model, while the data are split via 5-Fold cross-validation. The performance was evaluated based on three metrics: mean square error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE).

Results

Among the year 2005–2018, 20,924,951 cases and 11,892 deaths were supervised in the system. Hepatitis B (HB) is the most disease-causing incidence and death, and the proportion is greater than 70 percent, while the percentage of the incidence and deaths is decreased much in 2018 compared with 2005. Based on the measured errors and the visualization of the three neural networks, there is no one model predicting the incidence cases that can be completely superior to other models. When predicting the number of incidence cases for HB, the performance ranking of the three models from high to low is LSTM, BPNN, RNN, while it is LSTM, RNN, BPNN for Hepatitis C (HC). while the MAE, MSE and MAPE of the LSTM model for HB, HC are 3.84*10⁻⁰⁶, 3.08*10⁻¹¹, 4.981, 8.84*10⁻⁰⁶, 1.98*10⁻¹²,5.8519, respectively.

Conclusions

The deep learning time series predictive models show their significance to forecast the Hepatitis incidence and have the potential to assist the decision-makers in making efficient decisions for the early detection of the disease incidents, which would significantly promote Hepatitis disease control and management.

Predicting the impact of climate change on the re-emergence of malaria cases in China using LSTMSeq2Seq deep learning model: a modelling and prediction analysis study

OBJECTIVES

Malaria is a vector-borne disease that remains a serious public health problem due to its climatic sensitivity. Accurate prediction of malaria re-emergence is very important in taking corresponding effective measures. This study aims to investigate the impact of climatic factors on the re-emergence of malaria in mainland China.

DESIGN

A modelling study.

SETTING AND PARTICIPANTS

Monthly malaria cases for four Plasmodium species (P. falciparum, P. malariae, P. vivax and other Plasmodium) and monthly climate data were collected for 31 provinces; malaria cases from 2004 to 2016 were obtained from the Chinese centre for disease control and prevention and climate parameters from China meteorological data service centre. We conducted analyses at the aggregate level, and there was no involvement of confidential information.

PRIMARY AND SECONDARY OUTCOME MEASURES

The long short-term memory sequence-to-sequence (LSTMSeq2Seq) deep neural network model was used to predict the re-emergence of malaria cases from 2004 to 2016, based on the influence of climatic factors. We trained and tested the extreme gradient boosting (XGBoost), gated recurrent unit, LSTM, LSTMSeq2Seq models using monthly malaria cases and corresponding meteorological data in 31 provinces of China. Then we compared the predictive performance of models using root mean squared error (RMSE) and mean absolute error evaluation measures.

RESULTS

The proposed LSTMSeq2Seq model reduced the mean RMSE of the predictions by 19.05% to 33.93%, 18.4% to 33.59%, 17.6% to 26.67% and 13.28% to 21.34%, for P. falciparum, P. vivax, P. malariae, and other plasmodia, respectively, as compared with other candidate models. The LSTMSeq2Seq model achieved an average prediction accuracy of 87.3%.

CONCLUSIONS

The LSTMSeq2Seq model significantly improved the prediction of malaria re-emergence based on the influence of climatic factors. Therefore, the LSTMSeq2Seq model can be effectively applied in the malaria re-emergence prediction.

Driving Mechanism of Habitat Quality at Different Grid-Scales in a Metropolitan City

Urban ecosystem dysfunction, habitat fragmentation, and biodiversity loss caused by rapid urbanization have threatened sustainable urban development. Urban habitat quality is one of the important indicators for assessing the urban ecological environment. Therefore, it is of great practical significance to carry out a study on the driving mechanism of urban habitat quality and integrate the results into urban planning. In this study, taking Zhengzhou, China, as an example, the InVEST model was used to analyze the spatial differentiation characteristics of urban habitat quality and Geodetector software was adopted to explore the driving mechanism of habitat quality at different grid-scales. The results show the following: (1) LUCC, altitude, slope, surface roughness, relief amplitude, population, nighttime light, and NDVI are the dominant factors affecting the spatial differentiation of habitat quality. Among them, the impacts of slope, surface roughness, population, nighttime light, and NDVI on habitat quality are highly sensitive to varying grid-scales. At the grid-scale of 1000 to 1250 m, the impacts of the dominant factors on habitat quality is closer to the mean level of multiple scales. (2) The impact of each factor on the spatial distribution of habitat quality is different, and the difference between most factors has always been significant regardless of the variation of grid-scales. The superimposed impact of two factors on the spatial distribution of habitat quality is greater than the impact of the single factor. (3) Combined with the research results and the local conditions of Zhengzhou, we put forward some directions of habitat protection around adjusting urban land use structure, applying nature-based solutions and establishing a systematic thinking model for multi-level urban habitat sustainability.

Effectiveness of cascading time series models based on meteorological factors in improving health risk prediction

Meteorological factors, which are periodic and regular in a long run, have an unignorable impact on human health. Accurate health risk prediction based on meteorological factors is essential for optimal allocation of resource in healthcare units. However, due to the non-stationary and non-linear nature of the original hospitalization sequence, traditional methods are less robust in predicting it. This study aims to investigate hospital admission prediction models using time series pre-processing algorithms and deep learning approach based on meteorological factors. Using the electronic medical record data from Panyu Central Hospital and meteorological data of Panyu district from 2003 to 2019, 46,089 eligible patients with lower respiratory tract infections (LRTIs) and four meteorological factors were identified to build and evaluate the prediction models. A novel hybrid model, Cascade GAM-CEEMDAN-LSTM Model (CGCLM), was established in combination with generalized additive model (GAM), complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and long-short term memory (LSTM) networks for predicting daily admissions of patients with LRTIs. The experimental results show that CGCLM multistep method proposed in this paper outperforms single LSTM model in the prediction of health risk time series at different time window sizes. Moreover, our results also indicate that CGCLM has the best prediction performance when the time window is set to 61 days (RMSE = 1.12, MAE = 0.87, R² = 0.93). Adequate extraction of exposure-response relationships between meteorological factors and diseases and suitable handling of sequence pre-processing have an important role in time series prediction. This hybrid climate-based model for predicting LRTIs disease can also be extended to time series prediction of other epidemic disease.

Comparison of ARIMA and LSTM for prediction of hemorrhagic fever at different time scales in China

Objectives

This study intends to build and compare two kinds of forecasting models at different time scales for hemorrhagic fever incidence in China.

Methods

Autoregressive Integrated Moving Average (ARIMA) and Long Short-Term Memory Neural Network (LSTM) were adopted to fit monthly, weekly and daily incidence of hemorrhagic fever in China from 2013 to 2018. The two models, combined and uncombined with rolling forecasts, were used to predict the incidence in 2019 to examine their stability and applicability.

Results

ARIMA (2, 1, 1) (0, 1, 1)₁₂, ARIMA (1, 1, 3) (1, 1, 1)₅₂ and ARIMA (5, 0, 1) were selected as the best fitting ARIMA model for monthly, weekly and daily incidence series, respectively. The LSTM model with 64 neurons and Stochastic Gradient Descent (SGDM) for monthly incidence, 8 neurons and Adaptive Moment Estimation (Adam) for weekly incidence, and 64 neurons and Root Mean Square Prop (RMSprop) for daily incidence were selected as the best fitting LSTM models. The values of root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) of the models combined with rolling forecasts in 2019 were lower than those of the direct forecasting models for both ARIMA and LSTM. It was shown from the forecasting performance in 2019 that ARIMA was better than LSTM for monthly and weekly forecasting while the LSTM was better than ARIMA for daily forecasting in rolling forecasting models.

Conclusions

Both ARIMA and LSTM could be used to build a prediction model for the incidence of hemorrhagic fever. Different models might be more suitable for the incidence prediction at different time scales. The findings can provide a good reference for future selection of prediction models and establishments of early warning systems for hemorrhagic fever.

Analysis and prediction of hand, foot and mouth disease incidence in China using Random Forest and XGBoost

Hand, foot and mouth disease (HFMD) is an increasingly serious public health problem, and it has caused an outbreak in China every year since 2008. Predicting the incidence of HFMD and analyzing its influential factors are of great significance to its prevention. Now, machine learning has shown advantages in infectious disease models, but there are few studies on HFMD incidence based on machine learning that cover all the provinces in mainland China. In this study, we proposed two different machine learning algorithms, Random Forest and eXtreme Gradient Boosting (XGBoost), to perform our analysis and prediction. We first used Random Forest to examine the association between HFMD incidence and potential influential factors for 31 provinces in mainland China. Next, we established Random Forest and XGBoost prediction models using meteorological and social factors as the predictors. Finally, we applied our prediction models in four different regions of mainland China and evaluated the performance of them. Our results show that: 1) Meteorological factors and social factors jointly affect the incidence of HFMD in mainland China. Average temperature and population density are the two most significant influential factors; 2) Population flux has different delayed effect in affecting HFMD incidence in different regions. From a national perspective, the model using population flux data delayed for one month has better prediction performance; 3) The prediction capability of XGBoost model was better than that of Random Forest model from the overall perspective. XGBoost model is more suitable for predicting the incidence of HFMD in mainland China.

Predicting ICD-9 Codes Using Self-Report of Patients

The International Classification of Diseases (ICD) is a globally recognized medical classification system that aids in the identification of diseases and the regulation of health trends. The ICD framework makes it easy to keep track of records and evaluate medical data for evidence-based decision-making. Several methods have predicted ICD-9 codes based on the discharge summary, clinical notes, and nursing notes. In our study, our approach only utilizes the subjective component to predict ICD-9 codes. Data cleaning and segmentation, and Natural Language Processing (NLP) techniques are applied on the subjective component during the pre-processing. Our study builds the Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU) to develop a model for predicting ICD-9 codes. The ICD-9 codes contain different ICD levels such as chapter, block, three-digit code, and full code. The GRU model scores the highest recall of 57.91% in the chapter level and the top-10 experiment has a recall of 67.37%. Based on the subjective component, the model can help patients in the form of a remote assistance tool.

Time series analysis of hemorrhagic fever with renal syndrome in mainland China by using an XGBoost forecasting model

BACKGROUND

Hemorrhagic fever with renal syndrome (HFRS) is still attracting public attention because of its outbreak in various cities in China. Predicting future outbreaks or epidemics disease based on past incidence data can help health departments take targeted measures to prevent diseases in advance. In this study, we propose a multistep prediction strategy based on extreme gradient boosting (XGBoost) for HFRS as an extension of the one-step prediction model. Moreover, the fitting and prediction accuracy of the XGBoost model will be compared with the autoregressive integrated moving average (ARIMA) model by different evaluation indicators.

METHODS

We collected HFRS incidence data from 2004 to 2018 of mainland China. The data from 2004 to 2017 were divided into training sets to establish the seasonal ARIMA model and XGBoost model, while the 2018 data were used to test the prediction performance. In the multistep XGBoost forecasting model, one-hot encoding was used to handle seasonal features. Furthermore, a series of evaluation indices were performed to evaluate the accuracy of the multistep forecast XGBoost model.

RESULTS

There were 200,237 HFRS cases in China from 2004 to 2018. A long-term downward trend and bimodal seasonality were identified in the original time series. According to the minimum corrected akaike information criterion (CAIC) value, the optimal ARIMA (3, 1, 0) × (1, 1, 0)12 model is selected. The index ME, RMSE, MAE, MPE, MAPE, and MASE indices of the XGBoost model were higher than those of the ARIMA model in the fitting part, whereas the RMSE of the XGBoost model was lower. The prediction performance evaluation indicators (MAE, MPE, MAPE, RMSE and MASE) of the one-step prediction and multistep prediction XGBoost model were all notably lower than those of the ARIMA model.

CONCLUSIONS

The multistep XGBoost prediction model showed a much better prediction accuracy and model stability than the multistep ARIMA prediction model. The XGBoost model performed better in predicting complicated and nonlinear data like HFRS. Additionally, Multistep prediction models are more practical than one-step prediction models in forecasting infectious diseases.

Spatial-temporal heterogeneity and meteorological factors of hand-foot-and-mouth disease in Xinjiang, China from 2008 to 2016

The study aims to depict the temporal and spatial distributions of hand-foot-and-mouth disease (HFMD) in Xinjiang, China and reveal the relationships between the incidence of HFMD and meteorological factors in Xinjiang. With the national surveillance data of HFMD in Xinjiang and meteorological parameters in the study area from 2008 to 2016, in GeoDetector Model, we examined the effects of meteorological factors on the incidence of HFMD in Xinjiang, China, tested the spatial-temporal heterogeneity of HFMD risk, and explored the temporal-spatial patterns of HFMD through the spatial autocorrelation analysis. From 2008 to 2016, the HFMD distribution showed a distinct seasonal pattern and HFMD cases typically occurred from May to July and peaked in June in Xinjiang. Relative humidity, precipitation, barometric pressure and temperature had the more significant influences on the incidence of HFMD than other meteorological factors with the explanatory power of 0.30, 0.29, 0.29 and 0.21 (P<0.000). The interaction between any two meteorological factors had a nonlinear enhancement effect on the risk of HFMD. The relative risk in Northern Xinjiang was higher than that in Southern Xinjiang. Global spatial autocorrelation analysis results indicated a fluctuating trend over these years: the positive spatial dependency on the incidence of HFMD in 2008, 2010, 2012, 2014 and 2015, the negative spatial autocorrelation in 2009 and a random distribution pattern in 2011, 2013 and 2016. Our findings revealed the correlation between meteorological factors and the incidence of HFMD in Xinjiang. The correlation showed obvious spatiotemporal heterogeneity. The study provides the basis for the government to control HFMD based on meteorological information. The risk of HFMD can be predicted with appropriate meteorological factors for HFMD prevention and control.

Daily mean temperature and HFMD: risk assessment and attributable fraction identification in Ningbo China

BACKGROUND

Hand, foot, and mouth disease (HFMD) remains a significant public health issue, especially in developing countries. Many studies have reported the association between environmental temperature and HFMD. However, the results are highly heterogeneous in different regions. In addition, there are few studies on the attributable risk of HFMD due to temperature.

OBJECTIVES

The study aimed to assess the association between temperature and HFMD incidence and to evaluate the attributable burden of HFMD due to temperature in Ningbo China.

METHODS

The research used daily incidence of HFMD from 2014 to 2017 and distributed lag non-linear model (DLNM) to investigate the effects of daily mean temperature (Tmean) on HFMD incidence from lag 0 to 30 days, after controlling potential confounders. The lag effects and cumulative relative risk (CRR) were analyzed. Attributable fraction (AF) of HFMD incidence due to temperature was calculated. Stratified analysis by gender and age were also conducted.

RESULTS

The significant associations between Tmean and HFMD incidence were observed in Ningbo for lag 0-30. Two peaks were observed at both low (5-11 °C) and high (16-29 °C) temperature scales. For low temperature scale, the highest CRR was 2.22 (95% CI: 1.61-3.07) at 7 °C on lag 0-30. For high temperature scale, the highest CRR was 3.54 (95% CI: 2.58-4.88) at 24 °C on lag 0-30. The AF due to low and high temperature was 5.23% (95% CI: 3.10-7.14%) and 39.55% (95% CI: 30.91-45.51%), respectively. There was no significant difference between gender- and age-specific AFs, even though the school-age and female children had slightly higher AF values.

CONCLUSIONS

The result indicates that both high and low temperatures were associated with daily incidence of HFMD, and more burdens were caused by heat in Ningbo.

Comparison of ARIMA and LSTM in Forecasting the Incidence of HFMD Combined and Uncombined with Exogenous Meteorological Variables in Ningbo, China

Background: This study intends to identify the best model for predicting the incidence of hand, foot and mouth disease (HFMD) in Ningbo by comparing Autoregressive Integrated Moving Average (ARIMA) and Long Short-Term Memory Neural Network (LSTM) models combined and uncombined with exogenous meteorological variables. Methods: The data of daily HFMD incidence in Ningbo from January 2014 to November 2017 were set as the training set, and the data of December 2017 were set as the test set. ARIMA and LSTM models combined and uncombined with exogenous meteorological variables were adopted to fit the daily incidence of HFMD by using the data of the training set. The forecasting performances of the four fitted models were verified by using the data of the test set. Root mean square error (RMSE) was selected as the main measure to evaluate the performance of the models. Results: The RMSE for multivariate LSTM, univariate LSTM, ARIMA and ARIMAX (Autoregressive Integrated Moving Average Model with Exogenous Input Variables) was 10.78, 11.20, 12.43 and 14.73, respectively. The LSTM model with exogenous meteorological variables has the best performance among the four models and meteorological variables can increase the prediction accuracy of LSTM model. For the ARIMA model, exogenous meteorological variables did not increase the prediction accuracy but became the interference factor of the model. Conclusions: Multivariate LSTM is the best among the four models to fit the daily incidence of HFMD in Ningbo. It can provide a scientific method to build the HFMD early warning system and the methodology can also be applied to other communicable diseases.

Spatiotemporal characteristics and driving forces of terrorist attacks in Belt and Road regions

To achieve the strategic goals of the Belt and Road Initiative (BRI), it is necessary to deepen our understanding of terrorist attacks in BRI countries. First, we selected data for terrorist attacks in BRI regions from 1998 to 2017 from the Global Terrorism Database and analyzed their time distribution using trend analysis and wavelet analysis. Then, we used honeycomb hexagons to present the spatial distribution characteristics. Finally, based on the Fragile States Index, we used GeoDetector to analyze the driving forces of the terrorist attacks. The following conclusions were obtained: (1) During 1998–2017, the number of events was the highest on Mondays and the lowest on Fridays. In addition, the incidence of events was high between Monday and Thursday but was the lowest on Fridays and Saturdays. The number of events was the largest in January, May, July, and November and was the lowest in June and September; the incidence of terrorist attacks from April to May and July to August was high. (2) Terrorist attacks showed a 10-year cycle during the study period. Terrorist attacks in the last 10 years of the study period were broader in scope and higher in number compared with the previous 10 years. In addition, China, Russia, Saudi Arabia, and northeastern Europe saw many new terrorist attacks during the latter 10 years. (3) The number of terrorist attacks by bombing/explosion was the largest, followed by armed attack; assassination, kidnapping, and infrastructure attacks were the least frequent. The core areas of the terrorist attacks were Iraq, Israel, Afghanistan, Pakistan, and India. (4) The driving force analysis revealed that the indicators “security apparatus,” “human flight and brain drain,” and “external intervention” contributed the most to BRI terrorist attacks.

Spatiotemporal characteristics and driving forces of terrorist attacks in Belt and Road regions

To achieve the strategic goals of the Belt and Road Initiative (BRI), it is necessary to deepen our understanding of terrorist attacks in BRI countries. First, we selected data for terrorist attacks in BRI regions from 1998 to 2017 from the Global Terrorism Database and analyzed their time distribution using trend analysis and wavelet analysis. Then, we used honeycomb hexagons to present the spatial distribution characteristics. Finally, based on the Fragile States Index, we used GeoDetector to analyze the driving forces of the terrorist attacks. The following conclusions were obtained: (1) During 1998–2017, the number of events was the highest on Mondays and the lowest on Fridays. In addition, the incidence of events was high between Monday and Thursday but was the lowest on Fridays and Saturdays. The number of events was the largest in January, May, July, and November and was the lowest in June and September; the incidence of terrorist attacks from April to May and July to August was high. (2) Terrorist attacks showed a 10-year cycle during the study period. Terrorist attacks in the last 10 years of the study period were broader in scope and higher in number compared with the previous 10 years. In addition, China, Russia, Saudi Arabia, and northeastern Europe saw many new terrorist attacks during the latter 10 years. (3) The number of terrorist attacks by bombing/explosion was the largest, followed by armed attack; assassination, kidnapping, and infrastructure attacks were the least frequent. The core areas of the terrorist attacks were Iraq, Israel, Afghanistan, Pakistan, and India. (4) The driving force analysis revealed that the indicators “security apparatus,” “human flight and brain drain,” and “external intervention” contributed the most to BRI terrorist attacks.

Trend analysis and forecast of daily reported incidence of hand, foot and mouth disease in Hubei, China by Prophet model

Hand, foot, and mouth disease (HFMD) is common among children below 5 years. HFMD has a high incidence in Hubei Province, China. In this study, the Prophet model was used to forecast the incidence of HFMD in comparison with the autoregressive-integrated moving average (ARIMA) model, and HFMD incidence was decomposed into trends, yearly, weekly seasonality and holiday effect. The Prophet model fitted better than the ARIMA model in daily reported incidence of HFMD. The HFMD incidence forecast by the Prophet model showed that two peaks occurred in 2019, with the higher peak in May and the lower peak in December. Periodically changing patterns of HFMD incidence were observed after decomposing the time-series into its major components. In specific, multi-year variability of HFMD incidence was found, and the slow-down increasing point of HFMD incidence was identified. Relatively high HFMD incidences appeared in May and on Mondays. The effect of Spring Festival on HFMD incidence was much stronger than that of other holidays. This study showed the potential of the Prophet model to detect seasonality in HFMD incidence. Our next goal is to incorporate climate variables into the Prophet model to produce an accurate forecast of HFMD incidence.

Associations between ambient air pollution and daily incidence of pediatric hand, foot and mouth disease in Ningbo, 2014-2016: a distributed lag nonlinear model

Hand, foot and mouth disease (HFMD) has high prevalence around the world, with serious consequences for children. Due to the long survival period of HFMD virus in ambient air, air pollutants may play a critical role in HFMD epidemics. We collected data on daily cases of HFMD among children aged 0–14 years in Ningbo City between 2014 and 2016. Distributed lag nonlinear models were used to assess the effects of particulate matter (PM_2.5), sulphur dioxide (SO₂), nitrogen dioxide (NO₂) and ozone (O₃) on the daily incidence of HFMD among children, with analyses stratified by gender and age. Compared with moderate levels of air pollution, high SO₂ levels had a relative risk (RR) of 2.32 (95% CI 1.42–3.79) and high NO₂ levels had a RR of 2.01 (95% CI 1.22–3.31). The RR of O₃ was 2.12 (95% CI 1.47–3.05) and that of PM_2.5 was 0.77 (95% CI 0.64–0.92) at moderate levels of air pollution. Specifically, high levels of SO₂ and NO₂ had RRs of 2.39 (95% CI 1.44–3.96) and 2.02 (95% CI 1.21–3.39), respectively, among 0–4-year-old children, while high O₃ had an RR of 2.31 (95% CI 1.09–4.89) among 5–14-year-old children. Our findings suggest significant associations of high SO₂ and NO₂ levels and moderate O₃ levels in HFMD epidemics, and also indicate that air pollution causes lagged effects on HFMD epidemics. Our study provides practical and useful data for targeted prevention and control of HMFD based on environmental evidence.