newsfilter.io
Podcast, Interview

a16z Podcast | Health Data -- A Feedback Loop for Humanity

  • Q.bio aims to build technology to measure, digitize, and simulate human physiology to drive healthcare improvements and cost reductions over time.
  • The scope of treatable and diagnosable conditions is expected to expand as technological capabilities advance.
  • Healthcare systems are projected to shift from fee-for-service models to value-based models to incentivize prevention and establish metrics for tracking population health.
  • Continuous, regular monitoring is identified as a necessary requirement to validate progress within value-based care systems.
  • Early disease detection technology is anticipated to be adopted more frequently if the capacity to identify conditions exists.
  • Time-series data analysis is expected to improve diagnostic accuracy, reduce false positives, and increase test sensitivity and specificity compared to single-point measurements.
  • Future diagnostic approaches will involve deconflating the measurement of physiological systems from the analysis of those measurements.
  • Data scientists are predicted to integrate variables offering 70% predictive accuracy into multivariate models rather than discarding them.
  • Tracking deltas or changes over time within a system is considered inherently more sensitive than relying on absolute thresholds.
  • Information content in a complete physiological state is estimated to be a million trillion times greater than that of the genome.
  • Patients are expected to own, control, and utilize non-invasively gathered bodily data to resolve data value "chicken and egg" problems.
  • Gum disease is highlighted as a longitudinal indicator for cardiovascular disease due to the volume of available oral health data.
  • Redesigning delivery systems to focus on primary care and the doctor-patient relationship is expected to make healthcare better and cheaper.
  • Society will likely recognize the need for "data donors" to anonymously contribute health outcomes and longitudinal profiles.
  • A dataset of 1,000 people is estimated to be sufficient for building predictive models, while 1 million data donors represents an ideal scenario.
  • Government sponsorship of a public data repository, potentially accessible via DMV registration, is proposed to facilitate data donation.
  • Storing original sensor readings is expected to enable future re-analysis as computational technology becomes more sophisticated.
  • A network effect from data donation is predicted to benefit every person born in the future.