Table 2 Intervention studies meeting EPOC criteria for study design (n = 7)

Demaerschalk 2010 [38], USA

STRokE DOC AZ

RCT

Regional (spoke) and Academic Metropolitan (hub) hospitals

Hospital staff

Dec. 2007–Oct. 2008

Unit of analysis: patient

Concealed allocation: yes

Blinded: no

Allocation to condition: permuted block randomization of patients stratified by site

Patient: >18 years

tPA window: onset <3 h.

Patient: n = 54

Hospital: n = 3

Response rate, 68.4 %.

Representativeness: no demographic differences between groups

Myocardial infarction higher in int. group (p < 0.02).

Int-1: audio and video contact with a certified stroke team at a hub site, who had access to medical history, performed NIHSS, and reviewed test results and CT images

Int-2: a hub stroke consultant queried history, physical exam (including NIHSS), test results, CT report

tPA rate: denominator = acute stroke with <3 h onset.

Service delivery:

1. Evaluation times (e.g. door-ED)

2. Correct treatment decision

Patient outcomes:

1. Barthel Index (score 95–100)

2. mRS (score ≤2).

Cochran-Mantel-Haenszel test: comparison of correct decision rate between groups

Fisher’s exact test: rate of tPA, rate of intracranial haemorrhage, mortality, 90 day mRS

Wilcoxon rank sum test: 90-day Barthel Index and time comparisons

tPA rate: Int-1, 30 %; Int-2, 30 %

Service delivery:

1. NS

2. NS

Patient outcome:

1. NS

2. NS

Note: insufficient power to assess difference in tPA rates between groups.

Dirks, 2011 [41], The Netherlands.

PRACTISE

Cluster RCT

Hospitals

Hospital staff, including stroke neurologist and stroke nurse

May 2005–Jan. 2008

Unit of analysis: hospital

Concealed allocation: no

Blinded: no

Allocation to condition: hospitals randomised after pairwise matching on hospital type, tPA rate, stroke patients/year

Patient: >18 years

Hospital: 100–500 stroke admissions/year

tPA window: <4 h of onset

Patient: n = 1657.

Hospital: n = 12.

Response rate: Not reported.

Representativeness: patients: mean age, sex distribution and mean NIHSS at admission were similar between groups

Int: 5 × half day (across 2 years) meetings based on Breakthrough Series model. Teams of stroke neurologist and stroke nurse were created, who noted barriers to tPA use, set goals and plan actions

C: usual practices.

tPA rate: denominator = ischemic stroke, <4 h onset

Service delivery:

1. Onset-to-door time (min)

2. Door-to-needle time (min)

Patient outcome:

1. mRS <3 (at 3 months)

2. Quality of life—EuroQoL (at 3 months)

3. Mortality

Intention to treat

Multilevel logistic and linear regressions: comparison of tPA use, mRS, QoL and mortality between intervention groups.

Service delivery time analysis was adjusted for size, type and previous tPA rates, age, sex.

tPA rate: Int, 44 %; C, 39 % (unadjusted OR = 1.24 [1.02-1.51]).

Service delivery:

1. NS

2. NS

Patient outcome:

1. Poorer in C group

2. NS

3. NS

Meyer 2008 [39], USA

STRokE DOC

RCT

Remote “spoke” hospitals

Hospital staff

Jan. 2004–Aug. 2007

Unit of analysis: patient

Concealed allocation: no

Blinded: no

Allocation to condition: patients randomised within permuted blocks stratified by site

Patient: >18 years and ability to sign consent

tPA window: <3 h for treatment, but no time limit on eligibility for trial

Patient: n = 222 (111 vs 111)

Hospital: n = 4

Response rate: Patients: Not reported.

Representativeness: No demographic differences between groups. Int-1 had higher NIHSS score at presentation than Int-2 (p < 0.005).

Int-1: telemedicine (including video) consultation with patient by hub consultant including CT imaging

Int-2: telephone consultations for spoke sites with hub consultants

Hub provided treatment recommendations for both groups

tPA rate: denominator = acute stroke.

Service delivery:

1. Correct treatment decisions

2. Stroke onset to each point of care pathway (min)

Patient outcome:

1. Barthel Index (score 95–100).

2. mRS (score ≤2).

Fisher’s exact test: difference in tPA rate, functional outcomes

tPA rate: Int-1, 28 %; Int-2, 23 % (OR = 1.3 [0.7–2.5], p = NS).

Service delivery:

1. Greater in Int-1 compared to Int-2 (98 vs 82 %, OR = 10.9 [2.7–44.6], p < 0.001).

2. Few differences in service delivery times.

Patient outcome:

1. No difference between groups

2. No difference between groups

Morgenstern et al. 2003 [42], USA

TTL Temple Foundation Stroke Project

CBA

Hospitals in two communities

Community members and hospital staff

Feb. 1998–Sept. 2000

Unit of analysis: patient

Concealed allocation: no

Blinded: no

Allocation to condition: comparison community selected to match chosen intervention community

Patient: >21 years and county resident

tPA window:<3 h

Patient:

Phase 1: n = 277 (136 vs 141)

Phase 2: n = 499 (266 vs 233)

Phase 3: n = 150 (80 vs 70)

Hospital: n = 10

Response rate: Patients: N/A

Hospitals: not reported

Representativeness: hospital characteristics reported

Int: community mass media, hospital-based systems change via multi-disciplinary team development of ED protocols, problem solving, medical education, feedback.

C: not specified.

tPA rate: denominator = ischemic stroke

Service delivery:

1. Delay time to hospital

2. Staff-reported barriers to treatment

Patient outcome: none assessed

Fisher’s exact test: rate of tPA

ANOVA: delay in times

tPA rate: Int (phases 1–3): 2.2, 8.6, 11.2 % (p < 0.007); C (phases 1–3): 0.7, 0.9 %, (p = NS)

Service delivery:

1. No difference in either group

2. Reduction for Int group only (no statistical test)

Schwamm et al. 2009 [32], USA

ITS

Academic and community hospitals

Hospitals

April 2003-July 2007

Unit of analysis: hospital

Concealed allocation: N/A

Blinded: N/A

Allocation to condition: N/A (ITS design)

Patient: Principal diagnosis of stroke or TIA, arrival <2 h from onset, ICD-9. Retrospective chart review to confirm stroke/TIA

Hospital: >30 patients

Patient: n = 322,847 (ischemic = 73.2 %; TIA = 26.8 %)

Hospital: n = 790

Response rate: Unclear. Staggered recruitment over 4 years. By Jan. 2007, 8.35 % hospitals had dropped out (n = 66)

Representativeness: hospital characteristics provided

Int: quality improvement (Get With The Guidelines [GWTG]) programme, with organisational meetings, tool kits, collaborative workshops, hospital recognition, decision support information, performance feedback.

tPA rate: denominator = stroke or TIA, and arrival <2 h of onset

Service delivery: none assessed

Patient outcome:

1. Symptomatic intracranial haemorrhage within 36 h of tPA

Cochran-Mantel-Haenszel test: mean score for changes in rate of tPA and intracranial haemorrhage over time

tPA rate: significant increase from baseline (42.1 %) to year 5 (72.8 %; p < 0.0001).

Patient outcome:

1. NS over time

Greatest improvement (composite performance/program year in GWTG) in hospitals with more beds (p < 0.0001), larger annual stroke volume (p < 0.0001) and teaching status (p < 0.0001)

Scott et al. 2013 [43], USA

INSTINCT

Cluster RCT

Community hospitals

Physicians, pharmacists, nurses, EMS, admin teams

Jan.–Dec. 2007

Unit of analysis: hospital

Concealed allocation: no

Blinded: no

Allocation to condition: within pairs, hospitals were randomised to intervention or control groups. Randomisation reversed for three pairs to achieve greater urban/rural balance

Hospitals: discharging ≥100 stroke patient/year, <100 000 ED visits/year and non-academic stroke centres

tPA window: not specified

Hospitals: n = 24

Response rate: 83 %

Representativeness: not reported

Int: clinical practice guideline promotion, development of local stroke champions, continuing education, telephone support for treatment decision, academic detailing, audit and feedback

C: usual practices

tPA rate: denominator = ischemic stroke

Service delivery:

1. Adherence to tPA guidelines

Patient outcome:

1. Safety data from proportion of patients (2.2 %), with reported haemorrhage

Intention-to-treat (ITT) and target population (without one pair that was excluded after randomisation)

Generalised linear mixed model: assumed intra-hospital correlation between tPA rates at pre- and post- intervention periods

tPA rate: ITT: Int (pre and post), 1.25 and 2.79 %; C (n = 1; pre and post), 1.25 and 2.10 %. Int vs C, p = NS.

Target analysis: Int (pre and post), 1.0 and 2.62 %; C (pre and post),1.09 and 1.72 %. Int vs C, RR = 1.68 [1.09–2.57], p = 0.02

Service delivery:

1. NS difference between groups

Patient outcome:

1. NS difference between groups

Theiss et al. 2013 [40], Germany

CBA

Comprehensive stroke centres, and primary care hospitals

Hospitals

2006–2009

Unit of analysis: hospital

Concealed allocation: no

Blinded: not reported

Allocation to condition: hospitals matched on beds, distance from closest hub site and departments of internal medicine

Hospitals: not reported

No study hospitals had specialised stroke care prior to study start

Hospitals: n = 15

Response rate: not reported.

Representativeness: not reported

Int: tele-consultation service. Consisted of hub (n = 5) and spoke (n = 5) sites

C: usual practices

tPA rate denominator: all stroke

Service delivery: none assessed

Patient outcome:

1. Intracerebral haemorrhage

2. Mortality

Mean and SEM: for descriptive data

Student t and Fisher exact tests: longitudinal and pairwise comparisons, pooled ischemic stroke mortality

tPA rate: Hub sites: (pooled) increased 4.2 to 7.7 % (p < 0.0001); Spoke sites: (pooled) increased 1.1 to 5.9 % (p < 0.0001); C: (one hospital only) increased 0.8 to 5.7 % (p = 0 . 03).

Patient outcome:

1. NS

2. Significant decreases in spoke site only (10.3 to 7.3 %, p = 0.03)