Barriers to understanding natural variability.

By: Bruce J. West

In: Journal of Evaluation in Clinical Practice

Date: 11 March 2010

Key Concepts:

• Most decision making revolves around pattern recognition [prediction].
  • Data => information => patterns => knowledge, models.
  • Comparing patterns in the brain with patterns in the environment.
  • Humans make decisions based on the compatibility of these patterns.
• Old approach: homeostasis = the ideal (average) equilibrium physiological configuration for survival.
  • Human body evolved to optimize for survival.
  • Resists change.
  • Identify best range of operation for cardiovascular, respiratory and other physiological systems.
  • Most important measurement is the average value, assuming limited variability.
  • Dysfunction is a function of deviation outside the normal distribution.
• Associated with a mechanical world view.
  • View that dominates physical, social and life sciences.
  • Mechanical processes have a “best”, or average, value.
  • Deviations from the average are errors.
  • Small errors are more common than large ones.
    • Gaussian distribution (thin tails).
  • Individual behavior varies, population characteristics are stable and predictable.
    • Order, regularity emerges out of disorder.
  • Average value characterizes the phenomenon of interest.
• However, much of the world does not work this way.
  • Non-linear dynamics: small changes can have large effects.
  • Complex networks; unstable, unpredictable processes.
  • Small number of elements represent disproportionate share of outcomes (eg, wealth).
  • Limited ability to control or predict.
  • Failure of average to capture the properties of a phenomenon.
  • Pareto distribution (power laws, fat tails).
• New approach: homeostasis = variability necessary for adaptation.
  • Body has to adapt to rapid, short-term, complex changes in the environment.
  • Accommodate change, instead of resisting it.
  • Patterns of variability, not average value, are indicators of health.
    • Too much = no good.
    • Too little = no good.
  • A decline in variability signals deterioration in health, ability to adapt.
• Complex networks: variability, rate of change matter.
  • Previously, control of the world focused on control of man-made machines.
  • Similar mechanistic approach to non-linear networks and processes not effective.
    • Different measures: variability instead of average.
    • Process matters: Rate of change versus range of outcomes.