Patient safety and quality improvement in the ICU

Contents

Overview

The ICU is a high-risk environment. Complex patients, multiple invasive interventions, frequent handovers, and time pressure all create conditions for error. Patient safety science aims to understand why errors occur and to design systems that prevent them. Quality improvement (QI) translates that understanding into measurable practice change. Both disciplines share a core principle: most failures are system failures, not individual failures.

Classification of Errors

James Reason's framework distinguishes three error types:

Slips: Correct intention, incorrect execution — a distraction leads to the wrong drug being drawn up despite the clinician knowing the right one.

Lapses: Correct intention, failure of memory — a dose is missed because it was forgotten.

Mistakes: Incorrect intention — an incorrect plan is correctly executed. The clinician believed they were doing the right thing but the underlying reasoning was flawed.

Violations: Deliberate deviation from a rule or protocol. May be routine (normalised deviance) or exceptional.

Understanding the type of error matters for designing interventions: slips and lapses respond to reminders and forcing functions; mistakes require education and feedback; violations require a cultural and systemic response.

Systems Thinking

The Swiss Cheese Model

Reason's Swiss cheese model illustrates how defences in a system have holes (latent conditions). When the holes in multiple defensive layers align, a trajectory exists for an error to cause harm. Improving safety means adding layers of defence and reducing the size of holes — not blaming the individual who happened to be last in the chain.

Latent and Active Failures

Active failures: The actions or omissions by frontline staff that directly cause harm (the "sharp end").

Latent conditions: System-level factors — understaffing, poor equipment design, inadequate training, high workload — that create the conditions for active failures. Addressing latent conditions has a greater systemic safety benefit than focusing on individual behaviour.

High-Reliability Organisations (HROs)

Aviation, nuclear power, and firefighting are studied as examples of high-reliability organisations operating in high-hazard environments with very low failure rates. Key characteristics: preoccupation with failure (near-miss reporting), reluctance to simplify explanations, sensitivity to operations, deference to expertise, and commitment to resilience.

Quality Improvement Methodology

PDSA Cycle

The Plan-Do-Study-Act (PDSA) cycle is the foundational QI methodology:

  • Plan: identify the problem, set an aim, predict the outcome of a change
  • Do: implement the change on a small scale
  • Study: measure the result, compare with the prediction
  • Act: adopt, adapt, or abandon the change based on results

Multiple rapid PDSA cycles allow iterative improvement and learning before wide-scale implementation.

Model for Improvement

Developed by the Institute for Healthcare Improvement (IHI). Asks three questions before entering the PDSA cycle:

  1. What are we trying to accomplish? (Aim statement — specific, measurable, time-limited)
  2. How will we know that a change is an improvement? (Measurement plan)
  3. What change can we make that will result in improvement?

Statistical Process Control (SPC)

Run charts and SPC charts track a measure over time, distinguishing random variation (common cause) from significant change (special cause variation). SPC charts avoid the mistake of responding to every random fluctuation as if it represented a signal.

Measuring Quality in the ICU

Donabedian Framework

Quality is measured across three domains:

Structure: Resources in place — staffing ratios, presence of a consultant 24 hours, bed availability, availability of certain equipment.

Process: What is actually done — bundle compliance rates, time to antibiotics in sepsis, frequency of sedation holds, VTE prophylaxis prescription rate.

Outcome: What happens to patients — standardised mortality rates, ICU length of stay, ventilator-associated pneumonia rate, readmission rate, 30-day mortality.

Process measures are most sensitive to change and most modifiable; outcome measures are the most meaningful but require large numbers and risk adjustment to be interpretable.

Standardised Mortality Ratios

HSMR (Hospital Standardised Mortality Ratio) and SHMI (Summary Hospital-level Mortality Indicator): compare observed deaths with expected deaths (based on case mix), expressed as a ratio. HSMR >100 indicates higher mortality than expected; <100 lower. Interpreting these requires understanding the limitations of case mix adjustment and data quality.

ICNARC: The Intensive Care National Audit and Research Centre (ICNARC) case mix programme provides ICU-specific mortality benchmarking. It risk-adjusts using APACHE II scores and patient demographics to compare units nationally.

ICU Safety Priorities

Care Bundles

A bundle is a small number of evidence-based interventions that, when applied together reliably, improve outcomes more than individual elements alone. The concept of bundling acknowledges that 100% compliance with multiple interventions simultaneously is an ambitious but achievable target.

Ventilator care bundle: Elevate head of bed to 30–45°; twice-daily oral chlorhexidine decontamination; subglottic secretion drainage; sedation hold with spontaneous breathing trial; regular cuff pressure management. Reduces VAP.

Central venous catheter bundle: Hand hygiene; maximal sterile barrier precautions during insertion; chlorhexidine skin antisepsis; optimal site selection (subclavian preferred for low infection risk); daily review of line necessity and prompt removal. Reduces CLABSI.

Sepsis bundle: Hour-1 bundle (bloods, cultures, lactate, antibiotics, fluids). Reviewed and updated by Surviving Sepsis Campaign.

Medication Safety

Medication errors are common in the ICU. High-risk medications (concentrated potassium chloride, insulin, heparin, neuromuscular blocking agents) require double-checking protocols, standardised concentrations, and restricted access. Infusion pump programming errors are a recognised source of serious harm; smart pump technology with dose-error-reduction software reduces this risk.

Never Events

Never events are serious, largely preventable patient safety incidents that should not occur with robust safeguards in place. Examples include wrong-site surgery, misplaced nasogastric tube feeding, and IV potassium chloride administered in error. Never events must be reported to NHS England and trigger mandatory investigation.

Serious Incident Investigation

Root Cause Analysis

Root cause analysis (RCA) is a structured retrospective method for identifying the contributory factors behind a serious incident. The aim is not to assign blame but to understand the chain of events that led to harm.

RCA uses tools such as:

  • Timeline reconstruction: detailed chronological account of events
  • Five whys: repeatedly asking "why" to trace back from the surface event to underlying causes
  • Fishbone (Ishikawa) diagram: categorises contributing factors by domain (staff, environment, equipment, process, patient)

Mortality Review

All ICU deaths should be reviewed in a structured MDT process. The PRISM tool (Preventability, Reviewability, Impact, Sustainability, Mortality) provides a framework. Structured Judgement Review (SJR) is the NHS methodology for mortality review. The aim is to identify potentially avoidable deaths and drive improvements, not to apportion blame.

Incident Reporting

A non-punitive reporting culture is essential. Staff should report near-misses and adverse events freely. The NHS Reporting and Learning System (RLS) collects safety incident reports. Learning from near-misses is often more instructive than learning from harm events, because the volume of near-misses is larger and the counterfactual (what prevented harm this time?) is informative.

Human Factors

Human factors is the science of optimising the interface between people and systems.

Situational awareness (Endsley's model): Perception of elements in the environment → comprehension of their meaning → projection of future status. Loss of situational awareness — typically from cognitive overload, distraction, or anchoring — underlies many serious incidents.

Communication failures: A significant proportion of ICU errors involve communication breakdown. Structured communication tools reduce ambiguity:

  • SBAR: Situation, Background, Assessment, Recommendation — standardises handover communication
  • Read-back: verbal orders repeated back by the recipient to confirm accuracy
  • Closed-loop communication: acknowledgement that a message was received and acted on

Cognitive biases: Anchoring (over-weighting the first diagnosis), premature closure (stopping the diagnostic process too soon), and availability bias (over-weighting recently seen conditions) are common in time-pressured environments.

Fatigue: Night shifts and prolonged shifts impair performance comparably to significant alcohol intoxication. Duty hour regulations, rest facilities, and fatigue risk management systems are important structural mitigations.

Viva Questions

Explain the Swiss cheese model and how it applies to patient safety in the ICU.

Reason's Swiss cheese model describes safety as a series of defensive layers — protocols, guidelines, double-checks, supervision, equipment design, alarms — each of which prevents errors from reaching patients. However, no single layer is perfect: each has holes representing weaknesses or gaps. An error causes harm only when the holes in multiple consecutive layers align simultaneously, allowing a failure trajectory to pass through all defences unimpeded. In the ICU, a common example is a medication error: the prescriber writes an incorrect dose (hole in the prescribing layer), the pharmacist does not review the prescription (hole in the pharmacy check layer), the nurse administers without checking the last dose (hole in the bedside verification layer). No single layer failure caused harm; all three were required to align. The model's practical implication is that improving safety requires adding more defensive layers and reducing the size of the holes — not simply finding and punishing the person at the sharp end, who is usually the last in a chain of latent system failures. Latent conditions — understaffing, unclear protocols, poor equipment design, high workload — create and enlarge holes. A just culture that encourages reporting of near-misses allows identification of latent conditions before they cause harm.

What is the PDSA cycle and how would you use it to reduce catheter-related bloodstream infections in your unit?

The PDSA cycle is the core methodology of quality improvement. Plan: define the problem clearly (CLABSI rate above national benchmark), set a specific measurable aim (reduce CLABSI rate from X per 1000 catheter days to Y within six months), and identify a change to test — for example, implementing a standardised CVC insertion checklist. Do: test the intervention on a small scale, such as one ward or one nurse team, collecting data prospectively. Study: analyse the data — did CLABSI rates fall, was the checklist used consistently, what barriers arose? Act: if the change worked, broaden its application; if it didn't, modify the intervention based on what you learned. Multiple rapid PDSA cycles allow iterative refinement before full-scale implementation. In the CLABSI example, early cycles might reveal that the checklist is not being used because it is not accessible at the bedside — the next cycle would test moving it to the procedure trolley. Using SPC charts to track the CLABSI rate over time distinguishes genuine improvement signals from random variation, ensuring that apparent improvements are real and sustained.

How do human factors contribute to errors in the ICU and what practical steps can mitigate them?

Human factors research demonstrates that errors in high-stakes environments are rarely caused by incompetence or carelessness, but by predictable limitations of human cognition under the pressures inherent to ICU practice — cognitive overload, fatigue, distraction, time pressure, and the need to manage multiple simultaneous complex tasks. Situational awareness failures — losing track of the overall clinical picture — are a common precursor to serious events. Cognitive biases such as anchoring (committing to the first diagnosis and not revising it despite new information) and premature closure (stopping the differential too early) are well described. Fatigue from night shifts and prolonged working hours has effects on cognitive performance comparable to alcohol impairment at levels above legal driving limits. Practical mitigations include: structured communication tools (SBAR for handovers, read-back for verbal orders) to reduce miscommunication; daily safety huddles to maintain team situational awareness; checklists and bundles to reduce reliance on memory in time-pressured situations; forcing functions that make unsafe actions difficult (e.g., concentrated potassium chloride stored separately from standard saline); fatigue risk management programmes including appropriate rest periods and restricted consecutive night shift hours; simulation training to rehearse crisis scenarios in a safe environment; and a non-punitive reporting culture that allows near-misses to be shared and learned from.