The debrief carries the effect: what the evidence says about scenario and debrief design
Across 46 samples, debriefed groups outperformed controls by about 25% (d = 0.67), Tannenbaum and Cerasoli, Human Factors, 2013.
By Kim Loza, Head of Product at People Tech Revolution
0.67
debrief effect across 46 samples
Tannenbaum and Cerasoli, 2013
0.68
skills effect for a range of difficulty
Cook et al., 289 studies, 2013
138.8
violent-act cases per 10,000, psychiatric hospitals
BLS Table R8, 2023-24
5 of 1,227
MedEdPORTAL titles naming de-escalation or aggression
PubMed, read September 2026
Where the risk sits
Which settings should the opening scenarios mirror?
The settings where violence concentrates: in 2023-24, US psychiatric hospitals recorded 138.8 violent-act injury cases per 10,000 full-time workers, against 3.4 in ambulatory care.
See the data (cases per 10,000)
| Measure | Cases per 10,000 |
|---|---|
| Psychiatric hospitals | 138.8 |
| Residential disability and mental health | 60.9 |
| Nursing and residential care | 26.5 |
| Hospitals, all | 18.3 |
| General hospitals | 14.9 |
| All health care and social assistance | 11.9 |
| Home health care | 3.7 |
| Ambulatory care | 3.4 |
| All private industry | 2.6 |
Cases with days away from work per 10,000 full-time workers, 2023-24 annualised. Source: BLS Table R8.
Psychiatric hospitals run 40.8 times the ambulatory rate
That is 11.7 times the sector rate of 11.9; residential disability and mental health follow at 60.9.
So what for you: Set the opening scenarios in inpatient mental health and residential care.
The order holds on a wider count
Adding restricted-duty cases, psychiatric hospitals reach 204.5 per 10,000 and residential facilities 84.6, against 17.6 for the sector. On the 2018 coding, psychiatric hospitals stood at 124.9 (BLS).
The other person
Who is the other person in the scenario?
Usually a patient, relative or member of the public: in 2025, 14.47% of NHS staff in England reported physical violence from them, against 1.80% from colleagues.
See the data (percent of staff)
| 2021 | 2022 | 2023 | 2024 | 2025 | |
|---|---|---|---|---|---|
| From patients, relatives or the public | 14.57% | 14.82% | 13.88% | 14.38% | 14.47% |
| From other colleagues | 1.59% | 1.77% | 1.73% | 1.89% | 1.80% |
| From managers | 0.66% | 0.78% | 0.72% | 0.78% | 0.75% |
Percent of staff, last 12 months, 2021 to 2025. Source: NHS Staff Survey briefing 2025.
Violence from the public ran about 8.0 times the colleague rate
The public share held between 13.88% and 14.82% for 5 years; colleagues sat at 1.59% to 1.89%, managers 0.66% to 0.78%.
So what for you: Cast the other person as a patient, relative or visitor, and keep a colleague case.
See the data (percent of staff)
| Measure | Percent of staff |
|---|---|
| Ambulance (operational) | 38.04% |
| Nursing and healthcare assistants | 34.33% |
| Registered nurses and midwives | 22.58% |
| All staff | 14.47% |
| Social care | 13.29% |
| Medical and dental | 11.68% |
| Allied health professionals | 9.12% |
| Wider healthcare team | 3.02% |
Percent of staff, last 12 months, 2025. Source: NHS Staff Survey 2025 detailed spreadsheets.
Ambulance crews and nursing assistants report it most
38.04% of ambulance staff and 34.33% of nursing and healthcare assistants reported it, against 14.47% of all staff.
So what for you: Write ambulance and nursing assistant scenarios with people from those roles.
The debrief
Does a debrief add anything to the scenario itself?
A measurable amount: across 46 samples, debriefed groups outperformed controls by about 25%, and later meta-analyses of structured debriefs report d = 0.79 and 0.92.
See the data (effect size)
| Measure | Effect size |
|---|---|
| Debrief against none (2013) | 0.67 |
| After-action review (2021) | 0.79 |
| After-action review (2022) | 0.92 |
| Deliberate practice (2011) | 0.71 |
| Video added to the debrief (2014) | 0.10 |
Standardised effect size, 2011 to 2022. Source: Tannenbaum and Cerasoli; Keiser and Arthur; Keiser and Arthur; McGaghie et al.; Cheng et al..
The debrief carries an effect of its own
Structured debriefs show 0.67 to 0.92 across fields; adding video showed 0.10, which the review calls negligible.
So what for you: Give every session debrief time and a trained facilitator before adding playback.
Most published debriefs are thinly described
A 2014 review found 177 studies, with 11,511 learners, pairing simulation with a debrief. Against no intervention, effects ran from 0.28 to 2.16; duration and structure were usually incompletely reported.
Where a debrief does most
In a 2022 meta-analysis (overall d = 0.92), the largest effects came on complex tasks that give no feedback of their own. A separate reaction phase or a canned performance review added little.
Structure and scripts
A 2015 review found one generalisable study, of scripted debriefing run by novice instructors. In healthcare, highly and loosely structured reviews performed comparably (Keiser and Arthur, 2021, overall d = 0.79 across all samples).
Gains measured inside the scenario
A Cochrane review found training may raise knowledge (SMD 0.86); skills were very uncertain (0.21) and the long-term effect on aggression unclear (risk ratio 1.14). Price and colleagues found the strongest effects in artificial training scenarios.
Debrief models
Which debrief model should a program use?
A model the facilitators can run the same way every time: common structures move from reaction to analysis to summary.
PEARLS and plus-delta
PEARLS blends learner self-assessment, focused facilitation and directive feedback, with a scripted tool. Plus-delta asks what went well and what to change, a self-assessment method (Sawyer et al.).
Advocacy with inquiry
Rudolph and colleagues combine what the facilitator saw with a curious question about the learner’s reasoning, “debriefing with good judgment”; AHRQ’s dictionary defines advocacy-inquiry the same way.
The TeamSTEPPS debrief checklist
AHRQ’s Pocket Guide lists 10 items: communication clear; roles understood; situation awareness maintained; workload equitable; task assistance requested or offered; errors made or avoided; resources available; what went well; what should improve; what to do differently.
Debrief length
How long should the debrief run?
Long enough for the objectives: effective video-debrief studies ran 10 to 90 minutes, and no measured comparison ties debrief length to scenario length.
Tracing the ratio rule
The rule of 2 to 3 times the scenario is often credited to Fanning and Gaba (2007), whose full text holds no ratio; it reports crew debriefings averaging 31 minutes.
Where the ratio comes from
Bae and colleagues trace it to a 2010 book chapter and ran a 1 hour debrief after a 30 minute scenario. Others cite a paywalled 2010 article (a 2023 paramedic paper) or set a minimum of the scenario length (StatPearls).
What was measured
A 2024 study ran a 40 minute debrief at 2 times its scenario. Cheng and colleagues found short debriefs with expert modelling ahead of long ones (ES 0.21 to 0.74), a difference the review does not treat as established.
Design features
Which design features are worth building?
Difficulty levels, repetition and interactivity: across 289 comparative studies they pooled 0.65 to 0.68 on skills, while higher fidelity added 1% to 2%.
See the data (effect size)
| Measure | Effect size |
|---|---|
| Range of difficulty (20) | 0.68 |
| Repetitive practice (7) | 0.68 |
| Distributed practice (6) | 0.66 |
| Interactivity (89) | 0.65 |
| Multiple learning strategies (70) | 0.62 |
| Individualised learning (59) | 0.52 |
| Mastery learning (3) | 0.45 |
| Feedback (80) | 0.44 |
| Longer time (23) | 0.34 |
| Clinical variation (16) | 0.20 |
| Group training (8) | −0.22 |
Pooled effect size, skills; positive favours the feature, published 2013. Source: Cook et al., Medical Teacher.
Difficulty, repetition and spacing show the largest effects
Clinical variation (0.20) and group training (−0.22) showed no clear gain, and mastery learning rests on 3 studies.
So what for you: Write each scenario at several difficulty levels and plan spaced repeats.
Realism adds little measured learning
Across 24 studies, Norman and colleagues found high-fidelity simulation ahead of low fidelity by 1% to 2% on average. Against no practice, simulation showed a knowledge effect of 1.20 across 35,226 trainees (Cook, 2011).
Practice beats a lecture
Deliberate practice outperformed traditional clinical education across 14 studies (0.71). In a 2026 review, 12 of 56 studies rated strong, and each combined teaching with role-play or simulation.
Headset, live role-play or branching
In a 2023 police study, VR and live-action scenarios each outscored control, with VR comparable to live action. A 2026 review of VR formats found none of its 15 studies evaluated organisational impact; searches found no branching against linear comparison.
How often to repeat
OSHA says high-risk settings may need refreshers “perhaps monthly or quarterly”. A 2025 meta-analysis (SMD 0.85) suggests repeated practice may outperform lectures, without testing it.
Prebrief
What should happen before the scenario starts?
A prebrief on expectations, roles and safety: a 2022 meta-analysis found a larger collaboration effect with a prebrief (0.82 against 0.27).
See the data (effect size)
| Measure | Effect size |
|---|---|
| With a prebrief | 0.82 |
| Without a prebrief | 0.27 |
Standardised effect size, studies to June 2022. Source: Tong et al., Nurse Education Today.
A prebrief showed its gain on collaboration
The gap held across 4,926 students in 42 studies (p = 0.004); knowledge, skill and 5 other outcomes did not differ.
So what for you: Aim the prebrief at the team: roles, ground rules and how to call for help.
What the standard asks for
The INACSL prebriefing standard prepares learners for the content and ground rules; its 2025 revision adds roles, psychological and physical safety, confidentiality and the fiction contract.
A headset needs its own orientation
A 2024 review of VR and AR prebriefs found orientation and time handled inconsistently. Across 21 studies, preparation and briefing showed positive effects on satisfaction and knowledge (Tyerman et al.).
Safety in the room
How do we keep practice safe for staff with a real incident history?
Plan it into the prebrief and debrief, so staff can speak up and ask for help while the scenario keeps its challenge.
Discomfort is part of the design
A review that screened 2,071 records found psychological safety in simulation is not about removing all discomfort. The INACSL debriefing standard asks for privacy, open discussion and trust; Kolbe et al. describe restoring it.
Co-design
Who should write the scenario?
The staff who do the work, with patients and families where possible: published scenarios are scarce, with 5 of 1,227 MedEdPORTAL records naming de-escalation or aggression in the title.
A small published library
PubMed indexes 1,227 MedEdPORTAL records; 0.4% name de-escalation, escalation or aggression in the title, none workplace violence, so most organisations write their own.
Co-designed with staff and patients
The EDITION trial co-designed training across 10 wards; in its uncontrolled evaluation, falls in conflict and containment were associated with study phase, and outcome completion was 68% overall and 76% after training.
Staff rated it acceptable
211 of 214 trainees rated the co-delivered course: median 55 of 63 overall, 33 of 36 for acceptability, 23 of 27 for impact (Grundy et al.). They valued lived-experience co-delivery most.
Patients and families can help
AHRQ notes patients, families and community organisations can inform simulation design; a participatory Danish trial reached 73% implementation.
Guidance
What do regulators and standards bodies expect?
Practice and a debrief: OSHA writes that “Effective training programs should involve role-playing, simulations and drills”, and INACSL requires a planned debrief.
Written into standards and law
INACSL: “All simulation-based educational (SBE) activities must include a planned debriefing process.” California’s health care standard requires a practice debrief, and the Joint Commission requires de-escalation training at hire and annually.
Evidence
Which studies does this rest on?
The reviews, trials and datasets behind each figure.
| Study | Design | Sample | What was measured | Result |
|---|---|---|---|---|
| Tannenbaum and Cerasoli (2013), Human Factors | Meta-analysis | 2,136 people | Debrief against control | d = 0.67 |
| Keiser and Arthur (2021), J Appl Psychol | Meta-analysis | 61 studies | After-action review | d = 0.79 |
| Keiser and Arthur (2022), J Bus Psychol | Meta-analysis | 83 studies | Moderators | d = 0.92 |
| Cheng et al. (2014), Med Educ | Systematic review | 11,511 learners | Debrief features | Video 0.10 |
| Cook et al. (2013), Med Teach | Meta-analysis | 18,971 trainees | Design features | Difficulty 0.68 |
| Tong et al. (2022), Nurse Educ Today | Meta-analysis | 4,926 students | Prebriefing | 0.82 against 0.27 |
| Grundy et al. (2024), J Psychiatr Ment Health Nurs | Survey | 211 of 214 | Acceptability | Median 55 of 63 |
| BLS (2025), Table R8 | Employer survey | US private industry | Injury rates | 138.8 against 3.4 |
Decisions
What does this mean for a buyer?
Actions for the scenario brief.
- Set the opening scenarios where violence concentrates. Psychiatric hospitals record 138.8 cases per 10,000; 38.04% of ambulance staff report physical violence.
- Cast the other person as a patient, relative or visitor. Violence from the public ran 8.0 times the colleague rate in 2025.
- Build difficulty levels and spaced repeats before realism. They pooled 0.68 and 0.66; fidelity differed by 1% to 2%.
- Run a prebrief and a scripted debrief. Debrief effects run d = 0.67 to 0.92; a prebrief showed 0.82 against 0.27 on collaboration.
- Write scenarios with staff and people with lived experience. 211 of 214 trainees rated a co-designed course, median 55 of 63.
Board paper lines, ready to paste into a business case
Across 46 samples, debriefed groups outperformed controls by about 25% (d = 0.67; Tannenbaum and Cerasoli, 2013), and later reviews report d = 0.79 and 0.92. Across 289 comparative simulation studies, range of difficulty (0.68) and distributed practice (0.66) showed the largest pooled skills effects, while fidelity differed by 1% to 2% (Cook et al., 2013; Norman et al., 2012). US psychiatric hospitals recorded 138.8 violent-act injury cases per 10,000 full-time workers in 2023-24 against 3.4 in ambulatory care (BLS Table R8), so our scenarios start in inpatient mental health and residential care.
Limits and method
What these figures cannot show, and how they were counted.
- The pooled figures come from different outcomes and fields, including non-health teams.
- Cook 2013 pooled studies with large heterogeneity (I2 over 50%); mastery learning rests on 3 studies, repetitive practice on 7.
- Most trials measure performance inside a simulation, not at work.
- No measured comparison of branching against linear scenarios was found.
- The ratio sources (a 2010 chapter, a paywalled 2010 article) and the 2025 INACSL debriefing text were not read; the 2021 wording is quoted.
- BLS rates are US private industry, 2023-24 annualised, OIICS version 3, not continuous with 2018.
- NHS figures are self-report, for England, with corrected 2023 data; staff group values are 100 minus the share answering never.
- The MedEdPORTAL title screen is PTR’s own; other libraries could not be read. EDITION results are uncontrolled associations.
- Studies whose printed figures disagree internally are left out.
- Every source was read on 26 September 2026.
Where PTR fits
PTR builds de-escalation practice in a headset, on screen and with AI characters, with scenarios written alongside staff. See de-escalation training, VR de-escalation training, training for healthcare, AI roleplay practice, clinical educators and how a pilot runs.
- Evaluating a de-escalation pilot
- What VR de-escalation research measured
- Branching scenarios and AI characters
- Workplace violence claims data
Cite this research
Free to quote and cite with a link to this page. Name the version you read, so readers can see what has changed since.
APA 7
Loza, K. (2026, September 27). The debrief carries the effect: what the evidence says about scenario and debrief design (Version 2.0). People Tech Revolution. https://peopletechrevolution.com/insights/design-de-escalation-scenario-debrief
Data: Meta-analyses and reviews of debriefing, simulation design, prebriefing, psychological safety and co-design (2006 to 2026); BLS Survey of Occupational Injuries and Illnesses, Table R8, 2023-24; NHS Staff Survey 2025 briefing and detailed spreadsheets; INACSL, AHRQ and OSHA guidance; PubMed counts of MedEdPORTAL records. Read 26 September 2026.
Version history
| Version | Date | What changed |
|---|---|---|
| Version 2.0 | Rebuilt on meta-analyses and national data: debrief and design effect sizes, BLS setting rates, NHS Staff Survey series, prebrief, co-design and guidance; 6 charts, evidence table, decisions and limits. | |
| Version 1.0 | First published. |
Sources
Each study, dataset and standard at its publisher.
