Morning Session: Genomics, Medicine, Epidemiology and Data Interpretation

on YouTube ↗01h 51m

" summary

Randomized controlled trials cover only about 10–20% of treatments and just 6% of current asthmatics would have qualified for those trials; observational claims come from things like a pooled dataset of 250 million patients where 10% of diabetics, 11% with depression and 24% with hypertension have unique medication trajectories. Text models using LDA topics on MIMIC‑III (≈40k patients) outperform static acuity scores for 24‑hour mortality but show unequal accuracy by insurance, ethnicity and gender. Cost‑based algorithms under‑enroll sicker Black patients at a 97‑point auto‑enroll threshold (Black share would rise from 17.7% to 36.1% if equalized), while image‑trained models on x‑rays can remove racial pain gaps that clinician KL grades miss; genomic pipelines built on HapMap SNPs and PCA clusters risk reifying race unless structural racism and data gaps are addressed.

AI-written summary · about

" chapters (24)

1. 0:00Start
2. 1:13Evidence in Healthcare and Health?
3. 3:21Human Treatment Pathways Are Shockingly Unique
4. 5:02Opportunity: Machine Learning What Is Healthy
5. 8:54Topic Representation Improves In-Hospital Mortality Prediction
6. 14:31Modeling Mistrust in EOL Care with Clinical Notes
7. 15:23Train a Model to Detect Mistrust in Notes
8. 16:34Understanding Mistrust Metrics
9. 18:00ML+Health = Questions Beyond The Obvious
10. 18:36Understanding Trust Has Real Impact
11. 20:26What is "Doctor-ing"?
12. 20:56Health Data Is A Resource; Treat It That Way.
13. 21:12Create Research with a Resource
14. 25:24CONCEPTS OF RACIAL GROUPS
15. 26:46"Genomic risk markers" for Common Complex Diseases
16. 27:22Genomic markers for CCD
17. 29:52LAYERED INFRASTRUCTURES
18. 31:28The Winner
19. 31:48HAPMAP PHASE 1
20. 32:44HAPMAP DESIGN
21. 33:27Technologies Making Meaning: How Race Became Genomic
22. 34:29How SNPs become populations
23. 35:22Conclusion: Race and Big Data
24. 36:13What are populations?

" related talks

• 34mLongitudinal Phenotypes and Disease Trajectories at Population ScaleSøren Brunak · Computational Challenges in Very Large-Scale 'Omics'
• 30mReducing Bias in Artificial Intelligence...Laura Gutiérrez Funderburk · PyCon US 2022
• 15mInterpretable Machine Learning with RDean Allsopp · satRdays Newcastle 2019
• 07mBig Data and Cardiac RehabilitationLise Worthen-Chaudhari, Michael Gold · Strata Rx Conference 2013
• 56mIntroducing Decision-Making Under Uncertainty to Medical Research:...Don Berry · Optimization and Decision-Making Under Uncertainty (Simons Institute, September 2016)
• 02mTransforming healthcare through precision data scienceSriram Vishwanath · Strata + Hadoop World 2016 - New York, New York
• 06mHealthcare Analytics in the Age of Personalized MedicineAugust Calhoun · Strata Rx Conference 2013
• 01h 23mIntroduction to data mining (especially for healthcare domain)Milena Stojic · DSC Europe 23