<?xml version="1.0" encoding="UTF-8"?>
<opml version="2.0">
  <head>
    <title>006-02 Bayesian updating, decision thresholds, and diagnostic safety</title>
    <ownerName>Integrated Medical Foundations</ownerName>
  </head>
  <body>
    <outline text="Bayesian updating and diagnostic safety">
      <outline text="Evidence changes odds">
        <outline text="Prior odds times likelihood ratio gives posterior odds"/>
        <outline text="Keep the direction and scale of the update"/>
        <outline text="Predictive values answer the inverse question"/>
        <outline text="Likelihood ratios are not fixed">
          <outline text="Disease spectrum and test threshold"/>
          <outline text="Operator skill and timing"/>
          <outline text="Definition of the reference standard"/>
        </outline>
        <outline text="Correlated tests are not independent evidence"/>
      </outline>
      <outline text="Decision thresholds">
        <outline text="Below the testing threshold, investigation harms"/>
        <outline text="Between thresholds, information can change management"/>
        <outline text="Above the treatment threshold, act without waiting"/>
        <outline text="Thresholds move with stakes">
          <outline text="Dangerous and rapidly treatable justifies low-probability testing"/>
          <outline text="Toxic or invasive treatment demands stronger evidence"/>
          <outline text="Reliable follow-up makes watchful waiting safer"/>
        </outline>
        <outline text="Empirical treatment degrades later evidence">
          <outline text="Antimicrobials sterilise cultures"/>
          <outline text="Glucocorticoids suppress inflammatory findings"/>
        </outline>
        <outline text="Value of information: will the result change an action"/>
      </outline>
      <outline text="Causal problem representation">
        <outline text="Semantic qualifiers separate mechanisms">
          <outline text="Acute against chronic"/>
          <outline text="Focal against diffuse"/>
          <outline text="Exertional against random"/>
        </outline>
        <outline text="List what the representation does not explain"/>
        <outline text="Multimorbidity is common, parsimony is not mandatory"/>
        <outline text="Mechanism-based differentials">
          <outline text="Syncope: reflex, orthostatic, cardiac, metabolic, drug"/>
          <outline text="Hypoxaemia: hypoventilation, diffusion, mismatch, shunt"/>
        </outline>
        <outline text="Time course carries causal information"/>
      </outline>
      <outline text="Examination quality">
        <outline text="Technical and biological variance">
          <outline text="Cuff size, position, noise, effort, expectation"/>
        </outline>
        <outline text="Repeat a surprising measurement before explaining it"/>
        <outline text="Interobserver agreement is not diagnostic accuracy"/>
        <outline text="Absence of a sign can be uninformative"/>
        <outline text="Coherent signs from one mechanism beat any single sign"/>
        <outline text="Point-of-care devices need the same discipline"/>
      </outline>
      <outline text="Cognitive forcing and team reasoning">
        <outline text="Pattern recognition is fast when the case is familiar"/>
        <outline text="Analysis is for unfamiliar, high stakes, inconsistent cases"/>
        <outline text="Switch modes deliberately"/>
        <outline text="Diagnostic timeout triggers">
          <outline text="Transfer of care"/>
          <outline text="Unexpected deterioration or treatment failure"/>
          <outline text="A result that does not fit"/>
        </outline>
        <outline text="Team diversity uncovers blind spots if challenge is allowed"/>
        <outline text="Safe handover names escalation criteria and ownership"/>
      </outline>
      <outline text="Diagnostic harm and safety-netting">
        <outline text="Overdiagnosis is not a false positive"/>
        <outline text="Incidental findings carry surveillance and anxiety"/>
        <outline text="Underdiagnosis from access, atypical presentation, fragmented records"/>
        <outline text="Diagnostic overshadowing">
          <outline text="Symptoms wrongly attributed to a known diagnosis"/>
          <outline text="Recheck physical and medication causes"/>
        </outline>
        <outline text="Safety-netting">
          <outline text="Working diagnosis and what remains uncertain"/>
          <outline text="Specific warning symptoms"/>
          <outline text="Where and how urgently to seek help"/>
          <outline text="Who reviews pending tests"/>
        </outline>
        <outline text="Follow-up gives evidence only when designed"/>
        <outline text="The final question is which action is safest at this probability"/>
      </outline>
    </outline>
  </body>
</opml>
