18.3 Evidence-Based Practice, Clinical Practice Guidelines, Patient Safety & Quality Improvement
Key Takeaways
Match study design to the clinical question and appraise bias, uncertainty and applicability. Use AGREE II to evaluate guideline development and presentation.
After a safety event, protect the patient, communicate known facts, and address system conditions and individual behavior using a fair process.
Define outcome, process and balancing measures before PDSA testing. A run-chart signal prompts investigation of change and possible causes.
Use device-specific NHSN event and denominator definitions and distinguish surveillance classification from bedside diagnosis.
Evidence-Based Practice, Patient Safety and Quality Improvement
Quick Answer: Match the study design to the question and evaluate bias, precision and applicability. After a safety event, protect the patient, communicate what happened and identify system changes. For improvement, define measures before testing a change, then track the outcomes over time using consistent numerators and denominators.
Ask a question and appraise the evidence
Use PICOT to specify the population, intervention, comparison, outcome and time. For example, in invasively ventilated children in the PICU, does daily structured assessment of extubation readiness, compared with current practice, change time to successful extubation within 14 days? Measure reintubation and unplanned extubation as well. Define successful extubation and follow-up before interpreting a study.
| Question | Design and appraisal considerations |
|---|---|
| Does an intervention improve outcomes? | Start with a relevant synthesis of randomized trials or a well-designed trial. Check randomization, attrition, outcome definitions, confidence intervals, consistency and how closely the patients and intervention match local practice. A systematic review inherits limitations of its included studies. |
| What is the course of illness, or what risk accompanies an exposure? | A representative cohort can estimate prognosis or examine associations. Assess selection, measurement, confounding and loss to follow-up. |
| How accurately does a test identify disease? | Look for comparison with a suitable reference standard in patients representative of those tested in practice. Check how each test was interpreted and whether everyone received the reference standard. |
| How do children and families experience a decision? | Qualitative research can clarify priorities or barriers. Check sampling, data collection and how findings were developed from participants' accounts. |
For a hypothetical 14-day comparison with 12 events among 100 controls and 8 among 100 treated children, the observed risks are 12% and 8%. The absolute reduction is 4 percentage points, the risk ratio is about 0.67 and the number needed to treat is 25 over 14 days if the estimated effect is causal and applies. Check the confidence interval and harms before adopting the intervention.
Use AGREE II to appraise a guideline. Its six domains address scope and purpose, stakeholder involvement, rigour, clarity, applicability and editorial independence. Check how recommendations were derived, whether relevant families were represented, and whether resources and monitoring are feasible. A quality rating for the guideline does not replace appraisal of the underlying studies. Oxford CEBM evidence levels · AGREE II manual
Respond to a safety event and investigate contributing conditions
Stabilize the child, contain ongoing risk and notify the responsible teams. Document clinical facts, report through the safety system and communicate known facts and next steps to the family. Arrange follow-up as the review clarifies what happened. Reason's Swiss Cheese model illustrates how weaknesses across several layers of defense can align. Look for both frontline actions and underlying conditions, such as staffing, labeling, the order screen and storage design.
In a Just Culture review, assess the person's action and the circumstances. An unintended slip calls for support and redesign of the conditions that enabled it. An at-risk choice calls for coaching and attention to workflow incentives and how staff understand the risk. Conscious disregard of a substantial unjustifiable risk may warrant discipline under the applicable process. The severity of the patient's injury alone cannot classify behavior. Follow applicable reporting law and policy.
Use a multidisciplinary root cause analysis and action (RCA²) to establish the timeline, test contributing factors and select changes with assigned owners and measures. A fishbone diagram groups possible contributors, and repeated “why” questions help explore causal pathways. Confirm each link with evidence. For example, if an infant received a tenfold IV epinephrine overdose, examine the administered product, available concentrations, ordering and preparation steps. The 1 mg/mL product is ten times as concentrated as 0.1 mg/mL, the concentration in the pediatric IV/IO cardiac arrest algorithm. If both were stocked next to each other, redesign storage and make the appropriate ready-to-administer concentration available. Track stocking errors and unintended delays after the change. Standardization or a well-tested forcing function can reduce reliance on memory. Use education and checklists to support the new workflow.
The Joint Commission sentinel-event policy covers serious patient safety events and some specified events regardless of outcome. Determine whether the event meets the applicable hospital definition and complete the required internal analysis and action. Self-reporting to the Joint Commission is voluntary. If submitting through its ordinary RCA/action plan process, its published timeline is 45 business days from awareness, with a separate late-report process. Other mandatory reports may apply under state, federal or program rules. Check the policy in effect for the event. The revised Joint Commission and NQF alignment takes effect in January 2027. AHRQ disclosure tool · Reason's model · Marx on Just Culture · IHI RCA² · AHA pediatric cardiac arrest · Joint Commission July 2026 hospital policy · Joint Commission submission guidance
Test change and interpret data over time
The IHI Model for Improvement asks what the team aims to achieve, how it will recognize improvement and what change it can test. For a PICU central-line project, specify the target rate and deadline. Track CLABSIs per 1,000 line-days as the outcome, adherence to line-care steps as a process measure and harms such as skin injury or delayed access as balancing measures.
Check for a disparity before spreading a change. Suppose 90 of 100 families who prefer English and 12 of 20 families who prefer another language receive documented discharge teach-back. The overall 85% rate hides a 30-percentage-point gap (90% versus 60%). Check missing language and documentation data and review the workflow with families and interpreter staff. Test scheduling a qualified interpreter when discharge teaching is planned, then track subgroup rates and discharge delays using the same definitions. IHI disparity measurement guidance
- Plan: Predict what will happen, choose a small test and define data collection.
- Do: Test the change and record unexpected observations.
- Study: Compare results with the prediction, including unintended effects.
- Act: Adopt, adapt or abandon the change. Test under additional conditions before spreading it.
Use a run chart with time on the horizontal axis and a baseline median. In the Perla, Provost and Murray method, a shift is at least six consecutive non-median points on one side of the median, and a trend is at least five increasing or decreasing points. Points on the median are skipped for a shift. The method also considers the number of runs and unusual points. A signal prompts investigation of what changed. It cannot by itself establish that the intervention caused the difference. A control chart uses control limits calculated for the data and chart type. For event counts with varying device-days, a u chart can account for changing exposure when its count-model assumptions fit. Use the selected chart's published signal rules.
In the illustrated PICU run chart, the first 12 monthly rates have a median of 5.0 per 1,000 central-line days. The six later points are below that median. The first post-change month had 2 events in 800 central-line days, or 2.5 per 1,000. Before attributing the shift to a bundle, check that the surveillance definitions, sampling, case mix and device use are comparable. To combine months, sum events and sum device-days before dividing.
Local improvement projects and studies designed to generate generalizable knowledge can overlap. If the project may involve human-subjects research, consult the institution's review process before proceeding. IHI testing changes · Run chart methods · NIST control chart guidance · HHS QI and research guidance
Apply device prevention and surveillance definitions
| Measure | Surveillance and denominator | Prevention decision |
|---|---|---|
| CLABSI | Apply NHSN laboratory-confirmed bloodstream infection and secondary-source rules. For event attribution, day 1 starts with inpatient placement or first inpatient access of a line present on admission. A replacement can carry forward the count if no full calendar day passed without a line. An eligible line has reached day 3 and is present on the event date or the day before. Count one central-line day per patient per counted calendar day even with multiple lines. Denominator collection includes qualifying lines present on admission, even if not yet accessed. Clinical catheter-related bloodstream infection uses a different diagnostic framework. | Use aseptic insertion and maximal sterile barriers. Select site and antiseptic for patient age, risks and device. Use product-specific application and drying instructions, disinfect hubs before access, protect dressings and remove unnecessary lines. In NICU patients, weigh chlorhexidine benefits and risks by gestational age, postnatal age and skin maturity. |
| CAUTI | Apply the applicable NHSN UTI criteria, including the age-specific pathway and culture requirements. Eligible indwelling urinary catheter duration is more than two consecutive calendar days in an inpatient location, with the catheter present on the event date or removed the day before. Count urinary catheter-days according to the protocol. A positive urine culture alone does not establish symptomatic CAUTI. | Use a catheter only for an appropriate indication. Insert aseptically, maintain a closed system and unobstructed flow, keep the bag below the bladder and review need daily. Avoid routine irrigation. |
| PedVAE and PedVAP | PedVAE uses sustained worsening in minimum oxygen or mean airway pressure after at least two calendar days of stability or improvement. The January 2026 checklist uses a rise of at least 0.25 in minimum FiO₂ or 4 cm H₂O in minimum mean airway pressure for at least two calendar days, with protocol-specific baselines and pressure floors. Use ventilator-days for the rate. PedVAP uses the separate PNEU surveillance protocol, available in-plan in pediatric locations. Diagnose suspected pneumonia clinically. | Elevate the head of the bed when appropriate, provide oral care based on dentition, handle condensate safely and assess readiness for liberation. Avoid routine oral chlorhexidine and individualize sedation interruption and subglottic drainage. |
For device-associated rates, use the numerator and corresponding exposure for the same location and time. The NHSN standardized infection ratio instead divides observed by predicted infections from the applicable model. Coordinate surveillance and reporting with infection prevention, using the current NHSN monthly reporting plan and any applicable CMS or jurisdictional requirements.
2026 NHSN Patient Safety Component Manual · NHSN bloodstream infection FAQ · SHEA central-line guidance · CDC NICU CLABSI guidance · SHEA CAUTI guidance · NHSN pediatric ventilation surveillance · SHEA ventilator prevention guidance
A 3-year-old receives ten times the intended IV midazolam dose and has been stabilized. The review confirms that the nurse inadvertently selected a look-alike vial from an adjacent bin and that the ordering and dispensing systems did not flag the concentration mismatch. What response best fits Just Culture?
Classify the act as reckless based on the dose and terminate the nurse before investigating the workflow.
Classify it as an at-risk shortcut and use remedial education as the only response.
Treat the confirmed inadvertent selection as human error, support the clinician, review contributing conditions and redesign storage and verification.
Treat the lack of an alert as proof that no safety report or disclosure is needed.
A PICU plots monthly CLABSIs per 1,000 central-line days against a baseline median of 5.0. The first month after a line-care change has 2 CLABSIs in 800 central-line days. The next five monthly rates also fall below the baseline median. What is the rate for that first month, and how should the team interpret the pattern?
0.25 per 1,000 central-line days. A run chart cannot identify any signal without control limits.
2.5 per 1,000 central-line days. The first point alone proves that the intervention caused the improvement.
4.0 per 1,000 central-line days. Use the baseline median as the denominator instead of observed device-days.
2.5 per 1,000 central-line days. Six consecutive points below the median signal a shift. Investigate surveillance consistency and other changes before attributing the effect.
A council is considering a pediatric septic shock guideline for local adoption. It finds a systematic review of randomized trials and a multicenter retrospective cohort study. How should it evaluate the evidence and the guideline?
Use AGREE II to rate participant blinding in the cohort and adopt whichever study receives a higher score.
Appraise the trial review for bias, precision and applicability, examine the cohort for confounding and relevance, and use AGREE II to assess the guideline’s development and applicability.
Use the cohort as definitive proof of treatment effect because it represents routine practice.
Adopt the recommendation solely because a meta-analysis exists, regardless of its population, outcomes or uncertainty.
Sections you finish are checked off in the contents.