Clinical Practice Guidelines

What a guideline is. Clinical practice guidelines (CPGs) are systematically developed statements that link the best available evidence to explicit, actionable recommendations for specific clinical circumstances. They sit at the junction of evidence-based medicine, implementation science and medicolegal practice, and they recur in basic-science and clinical vivas because they test whether a candidate can appraise evidence rather than merely recall it.

Why appraisal is not optional. This page teaches AGREE II and GRADE, and its own citations say why you need them; the figure is worse than most readers expect. When 279 guidelines published between 1985 and 1997 were scored against a 25-item methodological standard, mean adherence was 43.1%, about 10.8 of the 25 items (Shaneyfelt, JAMA 1999, PMID 10349893). The worst-performing domain was identification and synthesis of the evidence, at 33.6%, the very step a guideline exists to perform, and adherence improved only modestly across the period studied, from 36.9% to 50.4%.

The examinable position follows from that. A guideline is an artefact to be appraised, not an authority to be obeyed, and "which guideline says so" is a weaker answer than "how was that guideline made". That is the justification for the critical appraisal skills tested alongside this topic, and for knowing where a recommendation sits in the levels of evidence and which study design produced it.

The three pillars. The exam-relevant understanding rests on how a guideline grades its evidence and recommendations (GRADE), how its methodological quality is appraised (AGREE II), and why good guidelines still fail at the bedside (implementation science). A consultant-level answer treats a guideline as a population-level tool that informs but never dictates the care of the individual in front of you.

The components. A guideline has five essential components, and each can be inspected on its own:

1. Clinical question (PICO): population, intervention, comparison, outcome
2. Evidence summary: a systematic review of the available evidence
3. Evidence quality assessment: GRADE or a similar system
4. Recommendation statements: clear, actionable guidance
5. Rationale: the explanation linking the evidence to the recommendation

The question. PICO frames the question the evidence summary answers.

Types of guideline. By scope, a guideline is condition-specific (a single disease or injury, such as ACL rupture), procedure-specific (a single intervention, such as TKA) or cross-cutting (applying across conditions, such as VTE prophylaxis). By developer, it comes from a professional society (AAOS, AOA), a government agency (NHMRC, NICE), a health system (hospital-specific) or a Cochrane group.

AGREE II. The instrument assesses the methodological rigour and transparency of a guideline across six domains and 23 items, and it is the gold standard for guideline appraisal. It answers how the guideline was made, and each domain asks a short set of questions of the document.

Scoring. Each item is rated 1-7, and domain scores are calculated as a percentage of the maximum possible. The appraisal closes with an overall assessment: would you recommend this guideline for use? Yes, yes with modifications, or no.

The rigour threshold. A score above 60% in Rigour of Development indicates a methodologically sound guideline. Many orthopaedic guidelines score poorly on this domain, so always check the evidence grading and the systematic review methodology before applying the recommendations.

The quick assessment. Before a formal AGREE II appraisal, four questions screen a guideline:

1. Who developed it? (reputation, expertise)
2. When was it updated? (currency; within 5 years is ideal)
3. What evidence grading was used? (GRADE preferred)
4. Are conflicts declared? (funding source, disclosures)

A living guideline is continuously surveilled and updated as new evidence emerges, rather than on a fixed 3-5 year cycle.

Finding guidelines. Start with the specialty society guidelines, check the government or national guideline databases, then search PubMed for "clinical practice guideline" plus the topic. The TRIP Database searches multiple guideline databases at once; the other main sources each have a characteristic strength and limitation.

Why guidelines fail. A high-quality guideline changes nothing unless it changes practice, and the reasons it does not are diagnosable. There are four barriers, and each has its own remedy.

Strategies. Passive dissemination, publication and mailing, has low effectiveness. The active strategies are:

• Education and dissemination through multiple channels
• Clinical decision support: electronic alerts, order sets
• Audit and feedback: compare performance to the guideline and feed the result back
• Academic detailing: one-on-one education with opinion leaders
• Local champions and opinion leaders
• Multifaceted interventions: combining strategies, the most effective

What the evidence says about uptake. The reviews of what actually changes practice (Grimshaw 2004; Cabana 1999) point to a small number of consistent factors:

Read "multifaceted is most effective" with care: Grimshaw found no reliable relationship between the number of components in a multifaceted intervention and its effect: more is not automatically better, and fit to the local barrier matters more. The same review of 235 studies and 309 comparisons found only about 29% reported any cost information (PMID 14960256). Implementation is not free and is poorly costed, and a recommendation without an implementation plan is a wish.

A guideline mandated is not a guideline implemented. The WHO surgical safety checklist is the best worked example on this page. It has three parts: sign in (identity, consent, site marking), time out (team briefing, antibiotic timing) and sign out (counts, recovery plan). Introducing the 19-item checklist across eight hospitals in eight cities was associated with inpatient death falling from 1.5% to 0.8% (p=0.003) and complications from 11.0% to 7.0% (p less than 0.001) (Haynes 2009, PMID 19144931).

Quote it with its companion. Those eight hospitals volunteered for a WHO improvement programme and there was no concurrent control; when Ontario mandated checklists across 101 hospitals, the effect vanished, with mortality 0.71% to 0.65% (OR 0.91, 95% CI 0.80-1.03, p=0.13) and complications 3.86% to 3.82% (OR 0.97, p=0.29) across 215,711 procedures (Urbach 2014, PMID 24620866). The difference between the two results is the whole of implementation science.

Measuring whether it worked. Implementation is judged on process measures (guideline awareness, compliance rates) and outcome measures (complication rates, patient-reported outcomes, cost-effectiveness). The benefits it is meant to deliver are reduced variation in care, improved adherence to best practice and measurable outcome improvements; VTE guidelines, for instance, reduced symptomatic PE rates. In postoperative care the usual quality indicators are VTE prophylaxis timing, antibiotic compliance, SSI rates and readmission rates.

What guidelines address. Indications for surgery, VTE and antibiotic prophylaxis, perioperative care protocols and enhanced recovery pathways. Specific surgical techniques, implant selection and approach comparisons are often not addressed.

AAOS examples.

Postoperative care. Guideline-directed postoperative care covers VTE prophylaxis (duration and agent), antibiotic prophylaxis, pain management protocols and ERAS pathways.

VTE prophylaxis by guideline. The agents and durations differ by body:

Not cookbook medicine. Guidelines inform decisions; they do not dictate them. Individual patient factors may override a recommendation, and rare complications or comorbidities may not be addressed at all. Wise clinicians use a guideline as the starting point and then individualise on patient-specific factors.

When a guideline does not apply. Deviation is expected when:

• The patient is atypical (age, comorbidities) or has a contraindication
• The patient's preferences differ from the guideline's assumptions
• Local resources are unavailable
• New evidence has been published since the guideline was developed

How guidelines go wrong. Misapplication means applying a guideline to the wrong population, using an outdated one, or ignoring individual factors. Overreliance is cookbook medicine, defensive practice and ignoring clinical judgement. The guidelines themselves carry inherent limits (evidence gaps, lag time in development, a population rather than individual focus) and quality concerns (conflicts of interest, industry influence, variable methodology).

The medicolegal position. Guidelines are evidence of, but do not define, the standard of care.

• Courts (Bolam/Bolitho in the UK, comparable standards elsewhere) ask whether a responsible body of practitioners would have acted similarly; a guideline is strong evidence of that body's practice but is not automatically binding.
• Documented, reasoned deviation, based on patient-specific factors and shared decision-making, is defensible. Unaware deviation is far harder to defend.
• Guidelines can cut both ways: following an outdated or poor-quality guideline is not a defence if a reasonable clinician would have recognised it as superseded.
• The safest position is awareness of the relevant guideline, an explicit rationale when departing from it, and a contemporaneous note of the shared decision.

Differential: Guideline vs Related Evidence Documents

A frequent viva trap is conflating a clinical practice guideline with other documents that look similar but carry different authority. Know the distinctions.

Controversies and Areas of Uncertainty

Major Guideline Developers Worldwide

Side-by-Side: Where Guidelines Genuinely Differ

Registry Evidence Feeding Guidelines

Arthroplasty and fixation guidance increasingly draws on national joint registries rather than RCTs alone:

• NJR (England, Wales, NI, IoM, Guernsey), AJRR (US), AOANJRR (Australia), SHAR (Sweden), Norwegian and NZJR registries supply implant survival and revision-rate data on millions of procedures.
• Registries provide external validity and rare-event detection (e.g. metal-on-metal failure) that trials cannot, but are observational - they inform but cannot prove causation.
• Guideline panels combine registry survival data with RCT functional outcomes and GRADE the resulting body of evidence.

High- vs Limited-Resource Practice Variation

International guidelines require local adaptation; the ADAPTE and GRADE-ADOLOPMENT frameworks formalise this.

Key drivers of variation:

• Resource setting: availability of DOACs, intra-operative imaging, implant range, and theatre capacity changes what is feasible (a strong recommendation is meaningless if the drug or implant is unavailable).
• Local epidemiology: injury patterns, comorbidity burden, and antimicrobial resistance differ by region and alter the benefit-harm balance.
• Local registry data: revision rates for a given implant in the local population may justify departing from an international default.

Example: an international guideline may default to a DOAC for thromboprophylaxis, but a limited-resource setting may justifiably adopt aspirin plus mechanical prophylaxis based on cost, monitoring capacity, and acceptable local outcome data - a defensible local adaptation, not a deviation in error.

Two separate judgements. GRADE rates the quality of the evidence (High, Moderate, Low, Very Low) separately from the strength of the recommendation (Strong or Weak, for or against). Quality is confidence in the effect estimate; strength answers whether we should do it. The recommendation side is taken in the next section.

The four levels. Each level carries a statement of how much further research would change the picture:

Where a body of evidence starts. A randomised trial starts at High; an observational study starts at Low. That baseline is established before any judgement is applied, and the explicit modifiers then move it down or up.

Downgrading. Five factors lower the rating.

Upgrading. The mirror-image factors raise the rating, and they apply typically to observational studies with exceptional features.

Oxford levels of evidence. A separate scheme ranks study designs by level:

View larger

Two strengths. A recommendation is strong or weak, for or against. Strong means the benefits clearly outweigh the harms; weak (also called conditional) means the balance of benefits and harms is close. The wording is the signal, "we recommend" against "we suggest", and surgeons must recognise the difference.

When low-quality evidence still earns a strong recommendation. The situations are (1) benefits that clearly outweigh harms even with the uncertainty, (2) ethical considerations that mandate action, and (3) resource implications that favour the intervention. The standing example is prophylactic antibiotics for open fractures: a strong recommendation despite limited RCT evidence.

Evidence to decision. Beyond evidence quality, the GRADE evidence-to-decision framework weighs six things:

• Benefits and harms: the balance of desirable and undesirable effects
• Values and preferences: how important the outcomes are to patients
• Resources: cost and feasibility
• Equity: impact on health disparities
• Acceptability: to stakeholders
• Feasibility: implementation barriers

Two guidelines can therefore make different recommendations from the same evidence, according to local values, resources or priorities.

Applying the strength in practice. A strong recommendation is applied to most patients unless contraindicated: "We recommend VTE prophylaxis for major orthopaedic surgery" means giving prophylaxis to essentially all patients. A weak recommendation requires shared decision-making: a recommendation worded "We suggest arthroscopic debridement may be considered for mechanical symptoms in early OA" means discussing the alternatives, because patient values matter. Where there is no recommendation the evidence is insufficient to guide practice; use clinical judgement and tell the patient about the uncertainty.

Shared decision-making. For a weak recommendation the management algorithm is explicitly shared decision-making:

1. Present the options, including the option of no intervention.
2. Communicate the evidence: the magnitude of benefit and harm in absolute terms (natural frequencies, not relative risk), plus the GRADE confidence in those estimates.
3. Elicit values and preferences: what outcomes matter most to this patient (function versus avoiding surgery versus avoiding bleeding).
4. Reach and document a shared decision that integrates the evidence with the patient's goals.

Decision aids improve knowledge and reduce decisional conflict; they are the practical tool that operationalises a weak recommendation at the bedside.