Search

BackgroundTraditional offline assessment of suicide probability is time consuming and difficult in convincing at-risk individuals to participate. Identifying individuals with high suicide probability through online social media has an advantage in its efficiency and potential to reach out to hidden individuals, yet little research has been focused on this specific field. ObjectiveThe objective of this study was to apply two classification models, Simple Logistic Regression (SLR) and Random Forest (RF), to examine the feasibility and effectiveness of identifying high suicide possibility microblog users in China through profile and linguistic features extracted from Internet-based data. MethodsThere were nine hundred and nine Chinese microblog users that completed an Internet survey, and those scoring one SD above the mean of the total Suicide Probability Scale (SPS) score, as well as one SD above the mean in each of the four subscale scores in the participant sample were labeled as high-risk individuals, respectively. Profile and linguistic features were fed into two machine learning algorithms (SLR and RF) to train the model that aims to identify high-risk individuals in general suicide probability and in its four dimensions. Models were trained and then tested by 5-fold cross validation; in which both training set and test set were generated under the stratified random sampling rule from the whole sample. There were three classic performance metrics (Precision, Recall, F1 measure) and a specifically defined metric “Screening Efficiency” that were adopted to evaluate model effectiveness. ResultsClassification performance was generally matched between SLR and RF. Given the best performance of the classification models, we were able to retrieve over 70% of the labeled high-risk individuals in overall suicide probability as well as in the four dimensions. Screening Efficiency of most models varied from 1/4 to 1/2. Precision of the models was generally below 30%. ConclusionsIndividuals in China with high suicide probability are recognizable by profile and text-based information from microblogs. Although there is still much space to improve the performance of classification models in the future, this study may shed light on preliminary screening of risky individuals via machine learning algorithms, which can work side-by-side with expert scrutiny to increase efficiency in large-scale-surveillance of suicide probability from online social media.

Clinical paper searching is a major task for clinical researchers to collect authoritative and up-to-date evidences to support their research works and clinical practices. Currently, this task needs huge amount of labor work. Researchers usually spend a lot of time searching on the online repository and iterating many times to get the final paper list. Systematic review is a special case, in which the paper searching process is a critical step. To address this challenge, this paper introduces a method to streamline the iterative paper searching process. It automatically selects the most probably matched papers, and then generates new search strategy. All the intermediate results are visualized based on the paper citation graph. It assembles technologies such as PageRank and Topic-based clustering to accelerate the paper searching tasks. The precision, recall, and execution time of the proposed method are then evaluated by comparing with published systematic reviews.

Clinical trials are key and essential processes for researchers to develop new treatments as well as evaluate their effectiveness and safety, whilst more than half of all clinical trials experience delays, which leads to a considerable amount of cost. In this paper, we present a cost-effective framework to reduce the time and monetary cost in the stage of recruiting and screening eligible clinical trial participants. By leveraging patients' observed conditions and the cost of medical examinations, the proposed framework uses collaborative filtering techniques to predict the utilized cost for the to-do medical examinations and then rank patients and medical examinations. The preliminary experiment results indicate that the framework is promising to reduce the cost spent on medical examinations by three quarters or even more and accelerate the recruitment process in the screening stage.

Recent advances in big data analytics provide more flexible, efficient, and open tools for researchers to gain insight from healthcare data. Whilst many tools require researchers to develop programs with programming languages like Python, R and so on, which is not a skill set grasped by many researchers in the healthcare data analytics area. To make data science more approachable, we explored existing tools and developed a practice that can help data scientists convert existing analytics pipelines to user-friendly analytics APPs with rich interactions and features of real-time analysis. With this practice, data scientists can develop customized analytics pipelines as APPs in Jupyter Notebook and disseminate them to other researchers easily, and researchers can benefit from the shared notebook to perform analysis tasks or reproduce research results much more easily.

With the rapid growth of clinical data and knowledge, feature construction for clinical analysis becomes increasingly important and challenging. Given a clinical dataset with up to hundreds or thousands of columns, the traditional manual feature construction process is usually too labour intensive to generate a full spectrum of features with potential values. As a result, advanced large-scale data analysis technologies, such as feature selection for predictive modelling, cannot be fully utilized for clinical data analysis. In this paper, we propose an automatic feature construction framework for clinical data analysis, namely, Feature++. It leverages available public knowledge to understand the semantics of the clinical data, and is able to integrate external data sources to automatically construct new features based on predefined rules and clinical knowledge. We demonstrate the effectiveness of Feature++ in a typical predictive modelling use case with a public clinical dataset, and the results suggest that the proposed approach is able to fulfil typical feature construction tasks with minimal dataset specific configurations, so that more accurate models can be obtained from various clinical datasets in a more efficient way.

Recent advances in cloud computing and machine learning made it more convenient for researchers to gain insights from massive healthcare data, while performing analyses on healthcare data in current practice still lacks efficiency for researchers. What's more, collaborating among different researchers and sharing analysis results are challenging issues. In this paper, we developed a practice to make analytics process collaborative and analysis results reproducible by exploiting and extending Jupyter Notebook. After applying this practice in our use cases, we can perform analyses and deliver results with less efforts in shorter time comparing to our previous practice.

Objective: To investigate how suicide microblog users in China "act" and "speak" differently from other microblog users without suicide idea., Methods: The suicide group consisted of 31 Chinese microblog users identified as suicide via online information provided by a Sina microblog user, and the control group consisted of 30 active microblog users without suicide idea screened by using suicide-related psychological scales. The differences in 10 microblog use behavioral characteristics and 88 linguistic characteristics between the suicide group and the control group were compared with normality test and rank sum test respectively., Results: In the behavioral characteristics, the suicide group used hyperlinks and "@" less frequently than the control group [0.04 (0.04) vs. 0.06 (0.04), P=0.029; 0.60 (0.27) vs. 0.69 (0.18), P=0.028], and was more self-focused [0.47 (0.25) vs. 0.30 (0.10), P=0.010]. In the linguistic characteristics, the suicide group showed less frequency in using measure word, work related word and apostrophe than the control group (P<0.05), and showed more frequency in using pronoun, personal pronoun, third person singular, non-specific pronoun, word expressing social experience, word expressing anxiety, word expressing exclusion, sexual word, religious word, second person singular, human being related word, negative emotion related word, anger related word, sadness or death related word (P<0.05)., Conclusion: Suicides seemed to interact less with others, showed more self-concern and more negative expressions, use more cognitively exclusive, death-related, religion-related words, and use less work-related words. The results of this study might be helpful for the research on suicide among netizen.