Root Cause Analysis and the London Protocol: An Introduction (101)

In brief: Root cause analysis (RCA) is a structured, retrospective investigation of a patient safety incident that asks what happened, how it happened, why it happened and what must change so that it does not happen again. The London Protocol, first published in 2004 and revised in 2025, is a systems analysis version of RCA built on James Reason’s organisational accident model and a framework of contributory factors.

Root cause analysis and the London Protocol: a general introduction

In aviation and other safety-critical industries, accidents and near misses are investigated exhaustively and the lessons are disseminated widely. In healthcare, learning from incidents has historically been fragmentary and uncertain (Vincent & Adams, 2006). Incident investigation is the discipline that closes this gap: it turns a single harmful event into knowledge about how the system of care actually works.

What root cause analysis is, and what it is not

In the United States, the most familiar approach is the root cause analysis method required by The Joint Commission for sentinel events, which grew out of Total Quality Management. Charles Vincent and colleagues argue that the label itself is misleading: incidents usually arise from a chain of events and many contributory factors, so the search for a single “root” oversimplifies. They prefer the term systems analysis (Vincent & Adams, 2006). The aim is not to find the person who made the last error, but to understand the conditions that made the error likely.

A second criticism concerns what happens after the analysis. In 2015 the US National Patient Safety Foundation renamed the process Root Cause Analyses and Actions (RCA²) to stress that prevention requires action, and introduced an Action Hierarchy that grades proposed actions from weak (training, reminders, new policies) to strong (forcing functions, standardisation, physical redesign) (NPSF, 2015). The Joint Commission now asks teams to identify at least one stronger or intermediate strength action for each root cause (The Joint Commission, 2016).

The London Protocol in five ideas

1. A theory of accidents. The protocol is based on Reason’s organisational accident model: unsafe acts at the sharp end are shaped by error-producing conditions and by latent failures in the organisation (Reason, 1990; Taylor-Adams & Vincent, 2004).
2. Care delivery problems (CDPs). The first analytic step is to identify actions or omissions by staff in the process of care where care deviated beyond safe limits of practice.
3. Clinical context. For each CDP the investigator records the patient’s condition at that moment and other patient factors that affected care.
4. Contributory factors. The investigator then asks why each CDP occurred, using a framework of contributory factors (Vincent et al., 1998). Factors that reflect general problems in a unit, rather than one-off circumstances, are the targets for change.
5. A structured process. Records review, a chronology and structured interviews with staff answer three questions in turn: what happened (outcome and chronology), how it happened (the CDPs) and why it happened (the contributory factors). Recommendations and an action plan follow.

Table note: the seven levels follow the framework of Vincent et al. (1998) used by the London Protocol; examples in the first five rows are drawn from Vincent and Adams (2006), the rest are illustrative.

Where RCA sits among incident analysis methods

RCA and the London Protocol are linear-to-systemic methods suited to most hospital incidents. Methods such as AcciMap and STAMP/CAST analyse incidents as failures of control across many organisational levels and are often preferred for complex, multi-agency events; HFACS classifies human factors in a fixed taxonomy. The new edition of the London Protocol was developed after a narrative review of these methods (Vincent et al., 2025). For a map of more than 180 safety science models and tools, see the Safety Science Entity Atlas.

Common pitfalls

• Stopping at the person closest to the event and recommending retraining or a reminder memo, the weakest actions in the hierarchy.
• Hindsight bias: judging decisions with information that was not available at the time.
• Excluding patients and families, who often hold information no one else has; the 2025 edition gives this much more weight.
• Writing recommendations without owners, deadlines or a way to measure whether they worked.
• Treating every incident with the same heavy investigation. England’s Patient Safety Incident Response Framework (PSIRF) moved the NHS away from investigating by default towards proportionate, learning-focused responses (NHS England, 2022).

Does root cause analysis work?

RCA has been required for serious events in many health systems for more than two decades, yet evidence that it reduces harm is thin. Peerally et al. (2017) set out eight problems with the way RCA is used in healthcare. The first is its name: by implying that a single root cause, or a small number of causes, can be found, it promotes a reductionist view. Others include investigations of variable quality, too little feedback to those who reported the incident and weak follow-through, so that organisational forgetting remains widespread.

Empirical work points the same way. A review of 302 RCAs from one US academic medical centre found that the proposed solutions were mostly education, process changes and enforcement of existing policies, weaker actions that are less likely to stop the same event happening again (Kellogg et al., 2017). Card (2017) argued that the popular “5 whys” technique pushes teams down a single causal chain and gives results that depend on who is asking, and recommended that it not be used for RCA in healthcare.

The scale of the problem: sentinel events

The Joint Commission received 1,575 sentinel event reports in 2024. Patient falls were the most frequent (776 events, 49%), followed by wrong surgery, delays in treatment, suicide and unintended retention of foreign objects, each around 8% of reports; 21% of all events were associated with the patient’s death. Because reporting is largely voluntary, the Commission warns that these data cannot show the true frequency of events or trends over time (The Joint Commission, 2025).

A toolbox: incident analysis methods compared

How PSIRF changes the response

Under England’s Patient Safety Incident Response Framework, organisations choose from a set of learning responses instead of investigating every incident in the same way. NHS England’s learning response toolkit includes after action reviews, swarm huddles held as soon as possible after an incident, multidisciplinary team reviews that look for learning across several similar incidents, and patient safety incident investigations (PSIIs) for incidents that indicate significant risk and potential for new learning. The Systems Engineering Initiative for Patient Safety (SEIPS) model is used to structure the discussion (NHS England, n.d.). The London Protocol fits most naturally inside a PSII, where its interviews and contributory factor framework provide depth.

An illustrative example

A patient receives a second dose of an anticoagulant after transfer from the emergency department to a ward. A person-focused review would stop at the nurse who gave the second dose. A London Protocol analysis would build a chronology and identify two care delivery problems: the dose given in the emergency department was not visible before transfer, and medicines were not reconciled on arrival. It would then look for contributory factors: an electronic record that did not show emergency department administrations on the ward view (task and technology), a handover held during an admissions peak (work environment) and no standard for medicines reconciliation at transfer (organisation and management). The strongest action would be a system change, such as making administrations visible across settings, rather than a reminder to staff.

Note: this case is invented for teaching and does not describe a real incident.

Related reading on this site: the article summary The London Protocol: Advancing Clinical Incident Analysis, Harvard Medical Practice Study II, The Safety Case: An Introduction (101) and the Adverse Events & Incident Analysis collection.

Sources for this introduction

• Card, A. J. (2017). The problem with ‘5 whys’. BMJ Quality & Safety, 26(8), 671–677. https://doi.org/10.1136/bmjqs-2016-005849
• Kellogg, K. M., Hettinger, Z., Shah, M., Wears, R. L., Sellers, C. R., Squires, M., & Fairbanks, R. J. (2017). Our current approach to root cause analysis: Is it contributing to our failure to improve patient safety? BMJ Quality & Safety, 26(5), 381–387. https://doi.org/10.1136/bmjqs-2016-005991
• National Patient Safety Foundation. (2015). RCA²: Improving root cause analyses and actions to prevent harm. NPSF. https://www.ihi.org/resources/tools/rca2-improving-root-cause-analyses-and-actions-prevent-harm
• NHS England. (2022). Patient Safety Incident Response Framework. https://www.england.nhs.uk/patient-safety/incident-response-framework/
• NHS England. (n.d.). Patient safety learning response toolkit. https://www.england.nhs.uk/publication/patient-safety-learning-response-toolkit/
• Peerally, M. F., Carr, S., Waring, J., & Dixon-Woods, M. (2017). The problem with root cause analysis. BMJ Quality & Safety, 26(5), 417–422. https://doi.org/10.1136/bmjqs-2016-005511
• Reason, J. (1990). Human error. Cambridge University Press.
• Taylor-Adams, S., & Vincent, C. (2004). Systems analysis of clinical incidents: The London Protocol. Clinical Safety Research Unit, Imperial College London.
• The Joint Commission. (2016). Guidelines for development of strong corrective actions. The Joint Commission.
• The Joint Commission. (2025). Sentinel event data: 2024 annual review. https://digitalassets.jointcommission.org/api/public/content/eac7511986c0442a9c1ae04b1aa02cc0
• Vincent, C., & Adams, S. (2006). The London Protocol. HealthManagement, 8(3). https://healthmanagement.org/c/hospital/issuearticle/the-london-protocol
• Vincent, C., Taylor-Adams, S., & Stanhope, N. (1998). Framework for analysing risk and safety in clinical medicine. BMJ, 316(7138), 1154–1157. https://doi.org/10.1136/bmj.316.7138.1154