This space will be used to summarise journal club discussions, which take place at the monthly data science hub working group meetings. 

Journal Club 1: January 2025

Paper being discussed: A novel digital health approach to improving global pediatric sepsis care in Bangladesh using wearable technology and machine learning. Garbern, S. C., et al. 30 October 2024, PLOS Digital Health. 

In this month’s data science meeting, we hosted our first journal club discussion. This discussion centered around wearable technology. We focused on one study about using wearable devices to monitor vitals among pediatric sepsis patients in a Bangladeshi institution. This study explored the feasibility of using a wireless wearable device to gather continuous data on vital signs, then used this information along with other clinical data to develop a diagnostic model for predicting advanced sepsis. Authors found one of their diagnostic models to be more sensitive compared to clinicians in diagnosing advanced sepsis (model sensitivity 0.83; clinician sensitivity 0.76). Authors also noted that they were able to utilise wearable devices without significant issues in data capture or data quality.

Following a summary of this study, the data science hub working group discussed the following questions: 

  • Can wearable devices play a significant role in capturing data in low resource settings? Are there barriers to use other than those which were mentioned in this study?
  • What opportunities does real-time monitoring of symptoms present in terms of early diagnosis? 
  • What are potential barriers and opportunities associated with adopting machine-learning models for diagnosis in low resource settings? 

We noted the opportunities that digital technologies and machine learning methodologies can bring to resource limited settings. However, we also discussed limits of such technologies - do we really need patient vitals by the minute? Would this ultimately add to nurses’ workload or reduce it? All of these are important considerations for implementing new technologies, particularly in resource-limited clinical settings. 

You may find this journal club's presentation file here. 

Journal Club 2: February 2025

Paper being discussed: Evaluation of Data Errors and Monitoring Activities in a Trial in Japan Using a Risk‑Based Approach Including Central Monitoring and Site Risk Assessment. Kondo, H., et al. July 2021, Therapeutic Innovation & Regulatory Science. 

In this month’s data science meeting, we hosted our second journal club discussion. Narshion Ngao led a discussion on risk based monitoring, which is an approach used in clinical trial settings. This methodology identifies, assesses, and mitigates risks related to patient data and safety. It's highly relevant in clinical trials because of their intense resource ulitization; risk-based monitoring can help allocate resources more effectively. However, there is this myth that risk-based monitoring is only to be used in large and complex trials, and the technology is too expensive. In the study by Kondo et al., the authors evaluated if simple risk-based monitoring without advanced technology could produce satisfactory results in terms of managing data and safety risks.

Following a summary of this study, the data science hub working group discussed the following points: 

  • ICH E6 (R2) guidelines – shifts emphasis to reliability of results.
  • It's important to ascertain what process is more effective & efficient.
  • The correction rate for on-site monitoring was 2.5%, that for transcription error was 0.7%, and that for lack of data entry was 0.1%. We thus ascertained that the contributions were made purely by source data verification (SDV), and the correction rate was likely lower than that in the previous study.
  • This suggests that SDV should have a greater focus on critical data.

You may find this journal club's presentation file here.