--- module: 073-02 language: en chapter: 73 title: "Patient Safety, Quality Improvement, Human Factors, and Clinical Systems" module_title: "Incident investigation, diagnostic and medication error, quality measurement, and hazard analysis" source_sha256: c76ce34e5b3ad3e16df3541efe62511729610f68bd6719cdb3b30e717a01727c --- # Safety investigation and quality measurement ## Purpose of safety investigation ### Reconstruct why decisions seemed reasonable ### Find lost chances for detection or recovery ### Neither exoneration nor blame by default ### Evidence for fair accountability and redesign ## Immediate safety first ### Treat the patient, prevent further exposure #### Quarantine products, check similar patients ### Investigation must not delay treatment ### Never alter the original record #### Later notes dated and transparent ### Preserve time-sensitive evidence and logs ### Open disclosure: facts, regret, next steps #### Uncertainty is no reason for silence ## Reliable incident chronology ### Separate times of action, record, result, harm ### Electronic timestamps need interpretation #### Delayed entry, copied notes, clock differences ### Interview separately with psychological safety #### Keep hindsight out of earlier decisions ### Map care across settings and professions #### Transitions expose unowned tasks ### Barrier existed, functioned, or was bypassed ## Root-cause analysis without one root ### Factors from patient to external policy ### Asking why can stop at a convenient human ### Independent, multidisciplinary team ### Recommendations need owner, deadline, measures #### Otherwise documentation, not prevention ### Poor policy compliance is description, not cause ### Causal factor would plausibly change outcome #### Check whether another pathway still harms ## Diagnostic error across the pathway ### From access and symptoms to follow-up ### Ignored disconfirming evidence makes it unsafe #### Track first reasonable diagnostic moment ### Results must reach an explicit owner ### Distributed process, not one wrong thought ### Feedback from autopsy, returns, complaints ## Precise medication incident analysis ### Reconstruct indication, order, route, rate ### Units, decimals, defaults, pump libraries ### Ambiguous order versus prescribing error ### Double-checks weak when not truly independent #### Shared assumption or confirmed calculation ### Outcome severity does not decide importance #### No-harm large errors still deserve learning ## Measuring quality ### Structure, process, outcome, balancing measures #### Balancing detects unintended consequences ### Operational definitions fix numerator, denominator #### Percentages improve by excluding high risk ### Run charts show shifts and trends ### Control charts: common versus special cause #### Reacting to every fluctuation destabilises ### Risk adjustment can hide inequitable care #### Stratify and read adjusted and unadjusted ## Diagnose the problem before the solution ### Define the gap with baseline and observation ### Process maps, Pareto, cause-and-effect diagrams #### Tools generate questions, not causation ### Observe real work with patients and staff ### Prioritise severity, frequency, inequity #### Rare catastrophes justify strong controls ## Proactive hazard analysis ### Failure-mode and effects analysis, step by step #### Risk-priority scores give false precision ### Simulation tests teams and environments #### In situ exposes latent safety threats ### Safety cases for new technology #### Alert fatigue, automation bias, cybersecurity ## Learning from complaints and success ### Complaints reveal dignity, access, delay failures #### Few complaints may mean barriers to speaking ### Safety-II studies how work usually succeeds ### Positive deviance finds better-performing teams #### Workarounds must not excuse system defects ### Findings must lead to visible, monitored action