UCR SOM - TriNetX Summer Bootcamp 2025, Week 4 – Conducting Outcomes Analyses

Outcomes Analysis Boot Camp Session Daniel Novak, director of student scholarly activities at the University of California Riverside School of Medicine, led the fourth session of the TriNetXSummer Boot Camp 2025, focusing on conducting outcomes analyses. The session covered statistical tests, data navigation in Trinetic, and the importance of propensity score matching to reduce confounders. Index Events and Propensity Matching Daniel explained the concept of index events and characteristics windows in TriNetX, emphasizing their importance for studying disease progression and correlations. He discussed the use of propensity score matching to balance cohorts and control for confounding variables, highlighting its necessity for accurate causal inference. Daniel also demonstrated how to use TriNetX' baseline comparison statistics tool to assess the need for propensity score matching, showing a significant difference between two groups before matching and a much higher overlap after matching. Outcome Exploration and Analysis Tool Daniel explained the concept of exploring outcomes, a useful feature for identifying potential confounders and ensuring enough participants for analysis. He demonstrated how to use the tool to compare diagnoses, procedures, medications, and labs between cohorts, emphasizing its role in avoiding rare outcomes and adding relevant codes to studies. Daniel also highlighted the importance of saving starred outcomes for final analysis and discussed the process of running outcome statistics after selecting covariates and outcomes. He concluded by urging attendees to make informed decisions for the best analysis possible. Propensity Score Matching Explained Daniel discussed the decision to use propensity score matching (PSM) in statistical analyses, explaining its purpose and when it is necessary. He emphasized that PSM helps reduce random variation and controls for confounders, leading to more accurate results. Daniel clarified misconceptions about PSM, such as the belief that it reduces cohort size or that non-significant results after PSM should be reported instead of using matched results. He stressed that PSM is crucial for ensuring the validity of study findings and that any change in statistical significance after PSM indicates the presence of confounding factors. Optimal Study Timeframe in Research Daniel discussed the importance of using propensity score matching (PSM) in scientific research, emphasizing that it should be a theory-driven approach rather than a random exercise. He explained that selecting the appropriate timeframe for a study is crucial, as longer timeframes may dilute the relationship between the event and outcomes due to factors like aging and changing standards of care, while shorter timeframes might miss significant outcomes. Daniel suggested considering the type of disease or condition, existing research, and using tools like the Kaplan-Meier curve to determine the optimal study timeframe. Efficient Timeframe Analysis Strategies Daniel explained how to set up and run multiple timeframes in an analysis, emphasizing the efficiency of duplicating the analysis and adjusting the timeframe rather than recalculating. He discussed the importance of making decisions about index events, outcomes (whether to pool or use individual outcomes), and excluding cases with prior outcomes. Daniel highlighted the trade-offs between single and pooled outcomes, noting that pooled outcomes can confuse analysis but may be necessary for certain research questions. He also stressed the significance of the "Exclude Cases with Prior Outcomes" checkbox in refining the study population. Propensity Score Matching Case Exclusions Daniel explained the implications of excluding cases with prior outcomes before or after propensity score matching. He highlighted that excluding cases post-matching can reduce cohort sizes and affect statistical power, particularly for small cohorts. Daniel suggested considering whether to exclude cases in outcomes or cohort definitions by assessing the impact on cohort sizes and statistical power. He also discussed the importance of making these decisions early in the study design process. Propensity Score Matching Techniques Daniel explained the concept of propensity score matching, highlighting its importance in balancing groups for research analysis. He demonstrated how matching reduces statistical differences between groups, making the data more reliable. Daniel also discussed various measures of association, including risk difference, risk ratio, odds ratio, and hazard ratio, explaining their uses and interpretations. He emphasized the importance of reporting 95% confidence intervals to assess the reliability of results. #TriNetX #ClinicalInformatics #MedicalStudentResearch #RealWorldEvidence #MedicalEducation #HealthDataAnalytics #ClinicalTrials #PropensityScoreMatching #EHRData #ICD10Codes #DigitalHealth #HealthTech #PopulationHealth