Table 1

Overview of study and intervention characteristics and implementation outcomes associated with diversion implementation

Author ID
Study Design/methodological quality
Diversion intervention (I) featuresComparison (C) featuresGeographic setting and scopeStudy participantsImplementation
Sample sizeTarget age (sample mean or median)Target complaints% participants eligible% participants diverted *% participant adherence% participants refused diversionInitial ED attendance
Pre-hospital diversion
Dale (2003)18
Controlled clinical trial
High risk of bias
A nurse or paramedic assessed 999 calls as to whether or not to dispatch ambulance. If caller triaged as not requiring an ambulance, they were offered advice and asked whether they still wished an ambulance to attend. Ambulance was not cancelled without consent of callerUsual ambulance response with no telephone assessment/adviceEngland
Ambulance service sites covering the whole of Greater
London, Birmingham, Coventry, the Black Country and South
Staffordshire; total population of about 10 million
I: 635
C: 611
All ages
(I: mean 44 years, C: mean 49 years)
Non-serious concerns51.9% (n=330/635)36.1% (n=119/330)NR62.4% (n=206/330)RR†0.76 (95% CI 0.69 to 0.84)
Krumperman (2015)24
Observational cohort
Moderate quality
Low-acuity calls to a 911 call centre were diverted to a nurse call centre. A nurse used a protocol to provide advice and/or refer caller to their PCP or urgent care centreParamedics decided whether concern could be treated at the scene and/or referred to a PCP USA. Ambulance service sites covering a rural and urban area; exact setting and population not reportedI: 216
C: 374
All ages
Low-acuity concernsNRNR95% (n=205/216)NRNR
Mason (2007)11
Mason (2008)15

Dixon (2009)14
Cluster RCT
High risk of bias
Paramedic practitioner travelled with the ambulance and was trained to assess and treat low-acuity complaints at the sceneUsual ambulance response, including EMS crew assessment and transport to the EDEngland Ambulance service sites covering Sheffield.I: 1549
C: 1469
≥60 years (mean 83 years)Minor injury or illnessNRMason 200711
70.4% (n=1090/1549)
Mason (2012)19
Quasi-experimental trial
Low quality
Emergency care physicians working as a single responder to ambulance service 999 calls who assesses the patients and either discharges them at the scene or refers to the most appropriate care practitionerStandard paramedic/technician ambulance responding to ambulance service 999 callsEngland & Scotland
All NHS trusts employing emergency care physicians in England and Scotland were invited to participate. ’Control’ trust sites that did not employ emergency care physicians but were within the same or in a neighbouring county and offered the same service configurations as intervention trusts were selected to participate
I: 593
C: 514
(I: mean 69 years, C: mean 63 years)
Not specified. Emergency or urgent complaints that were eligible to be seen by the emergency care physiciansNR43.3% (n=257/593)NRNRNR
Ross (2013)25
Observational cohort
Moderate quality
Paramedic evaluation and transport to a detoxification facility with limited medical care on a 24-hour, 7-day per week basisTransport to the EDUSA El Paso County ambulance service agencies covering the greater Colorado Springs metropolitan area; total population of about 370,000I: 138
C: 580
>18 years
(I: median 46 years, C: median 43 years)‡
Alcohol intoxication without any significant acute illnesses or injuries19.2% (n=138/718)92% (n=127/138)NRNRNR
Snooks (2004)16
Cluster RCT
High risk of bias
Ambulance crews transported patients who met specific criteria to a minor injury unitTransport to the EDEngland Five ambulance stations in the London and Surrey ambulance servicesI: 409
C: 425
Minor injuries, but not illnessesNR10% (n=41/409)NRNRRR†0.96 (95% CI 0.89 to 1.04)
Snooks (2004)22
Observational cohort
Moderate quality
Ambulance crews used protocols to treat patients who fell within a list of dispatch criteria at home (treat and release)Transport to the EDEngland Two ambulance stations in West LondonI: 251
C: 537
(I: mean 54 years, C: mean 47 years)
Non-serious injuries or illnessesNR37.1% (n=93/251)40.2% (n=101/251)NRNR
Snooks (2014)13
Cluster RCT
High risk of bias
Computerised clinical decision support tool for paramedics to use to decide whether to take patients who had fallen to the ED or leave at home with referral to a community-based falls serviceTransport to the EDWales
Recruited patients from two UK study sites. Paramedics were eligible to participate in the trial if they worked at any of 13 ambulance stations with a falls referral pathway in place
I: 436
C: 343
>65 years
(I: median 83 years,
C: median 82 years)
FallsNR42% (n=183/436)NRNRRR†0.94 (95% CI 0.82 to 1.09)
Snooks (2017)12
Cluster RCT
High risk of bias
Clinical protocol used by paramedics for the care of older people who have fallen to assess and refer them to a community-based falls serviceTransport to the EDEngland and Wales
Three ambulance services in England and Wales in which a falls prevention service was available, but no services in place for paramedics to make referrals from the scene of emergency service call attendances
I: 2420
C: 2284
>65 years
(I: mean 82.54 years,
C: mean 82.14 years)
8.4% (n=204/2420)NRNRRR†1.04 (95% CI 1.00 to 1.09) aOR¶ 1.08 (95% CI 0.96 to 1.22)
ED-based diversion
Doran (2013)20
Quasi-experimental trial
Low quality
Research specialist and triage nurse identified eligible patients to be escorted from the ED waiting room to the primary care clinicUsual ED careUSA
Adult academic ED serving 100 000 patients treated yearly
I: 662
C: 191
≥23  years
(I: mean 47.3  years, control: mean 46.3  years)
Any25.7% (n=1404/5462)85% (n=563/662)93.4%
15% (n=99/662)NR
Ellbrant (2015)23
Observational cohort
Moderate quality
PED nurse determined whether patients would be sent home, referred for other ED care or admittedPED physician who determined whether patients would be sent home, referred for other ED care or admittedSweden
Academic hospital in Malmö, Sweden, serving approximately 400 000 urban people
I: 344
C: 713
0–17  years
Any32.5% (n=344/1057)84.3% (n=290/344)NRNRNR
Washington (2002)17
Unclear risk of bias
Triage nurses used standardised criteria to identify patients who would be safe for deferred care at a non-emergency setting at a later dateUsual ED careUSA
Academic level 1 hospital in Los Angeles County, California, with 91 000 visits annually
I: 75
C: 81
>18  years
(I: mean 41  years , C: mean 42  years)
Abdominal pain, musculoskeletal symptoms, or respiratory infection35.7% (n=421/1176)100% (n=75/75)96% (n=71/74)48% (n=143/299)NR
Washington (2000)21
Observational cohort
Moderate quality
Tertiary care medical centre in Los Angeles, California, with approximately 30 000 unscheduled walk-in visits
I: 226
C: 961
>18  years (mean 53  years of those patients screened n = 1187)19% (n=226/1187)68.1% (n=154/226)90.3% (n=139/154)NRNR
  • *Those participants assigned to receive diversion and were actually diverted from the ED.

  • †Unadjusted RR.

  • ‡Median age reported.

  • §The study did include paediatric patients <5 years of age, but an upper age limit was not specified.

  • ¶As well as indicators for group, site and their interaction, covariates adjusted for included age and its square, distance to ED, recruitment point, seasonality, biological sex and whether or not the index call was made out of GP hours.

  • a, adjusted; ED, emergency department; GP, general practitioner; NA, not available; NR, not reported; PCP, primary care physician; PED, paediatric emergency department; RCT, randomised controlled trial; RR, relative risk.