Table 2 Summary table of studies describing interventions based in acute care settings

Storrow et al., 2005, USA [27]

Patients ≥18 undergoing evaluation for suspected heart failure (HF) exacerbation. Only those classified as low-to-moderate-risk eligible for inclusion.

Observational sequential cohort study (pilot study).

Observation unit established in ED.

Observation unit available to treating physician to use in treatment (n = 28).

Heart failure standard care without observation unit available to treating physicians (n = 36).

Repeat visits to ED and readmission with primary complaint of HF all within 30 days.

No significant difference in hospital readmission rates (p = 0.538).

Potential for enrolment bias by treating physician. Observational study.

Foo et al., 2012, Singapore [33]

Patients ≥ 65 in the emergency department observation unit (EDOU). Thirteen conditions were accepted into the EDOU. Patients excluded if had poor premorbid condition, nursing home resident or those admitted to inpatient ward from EDOU.

Before/after prospective study.

Emergency department observation unit.

Geriatric assessment and intervention in the EDOU prior to discharge by emergency nurse trained in geriatric care; exploring the patient’s medical, social and functional status (with referral to physiotherapist, appropriate community or social care services or GP if required) (n = 315).

Historical controls received usual EDOU care (n = 172).

Unscheduled ED re-attendance and hospitalisation at 3, 6, 9 and 12 months.

Significant reduction in ED re-attendance at 3, 6, 9 and 12 months: overall reduction of 41% (adjusted IRR 0.59, 95% CI 0.48–0. 71) at 12 months. Hospitalisation rates significantly reduced at 3, 6, 9 and 12 months: overall reduction of 36% (adjusted IRR 0.64, 95% CI 0.51–0.79)

Before/after design. Possible that recruitment

process favoured a positive outcome.

Schull et al., 2012, Canada [31]

Data suggest adult patients attending ED.

Retrospective analysis of the difference in median ED LOS and admission rates among all ED visits after versus before CDU implementation at pilot-CDU and control sites. First 18 months of CDU operation compared with 1 year baseline period prior to CDU.

Pilot-CDU sites (7). All CDUs within or next to ED.

7 Pilot CDU sites. Staffing models varied by site. Variation in CDU protocols. Number of beds varied by CDU site (n = 455, 942).

9 EDs without a CDU. ED had been unsuccessful in applying for pilot-CDU funding

(n = 1,172,305).

Admission rates, ED revisit rates (after 48 hours, 72 hours, 7 days and 30 days) and ED length of stay.

Small decrease in hospital admission rate high-acuity patients: −0.8% (−1.5% to −0.03%) and moderate-acuity patients: −0.6% (−1.1% to −0.2%). No changes in ED revisit rates. 4% of ED patients admitted to CDUs.

Only 4% of ED patients admitted to CDUs. No mention of target population. Difficult to see efficiency gains. Pilot study. Missing retrospective data. Different sites had different protocols, staffing etc.

Conroy et al., 2014, UK [15]

Patients presenting to ED ≥16 years.

Pre-post cohort before and after establishment of Emergency Frailty Unit (EFU).

Emergency Frailty Unit (EFU).

Comprehensive geriatric assessment in the EFU. Unit included input of acute medical consultant and later full geriatrician coverage (08 h00 – 18 h00, 7 days a week). Intervention moved to geriatrician integrated assessment with focus on patients identified for discharge and improvement of pathways to community. (n = 110, 517).

Usual care (model of care using Emergency Decision Unit without specialist geriatric input) situated in ED (n = 109, 994).

Primary: ED conversion rate (admission avoidance).

Secondary: readmissions following attendance at 7,30 and 90 days.

Admissions (ED conversion rate) for patients >85 years fell from 69.6% (control) to 61.2% (intervention) (95% CI: 66.0– 73.1%) in the control period, p < 0.001. RR 0.88 (95% CI 0.81-0.95)

Readmission rates fell across all age groups comparing intervention and control groups. Readmission risk ratio for those 85+ 0.77 (95% CI: 0.63–0.93) for 90 day readmissions.

ED attendance increased in older people (65+) over the study period. ED attendance decreased for 16–64 year olds over study period.

No concurrent control group therefore causal effect difficult to establish.

Li et al., 2010, Australia [11]

All general medical patients presenting to ED.

Retrospective before and after study. Before and after the establishment of an AAU.

Acute assessment unit at a University teaching hospital

Establishment of an acute assessment unit. Remit to receive adult patients who were not clinically appropriate for sub-speciality medical unit or for a surgical service (n = 3992).

ED patients requiring admission either referred to subspecialty service or to an ‘on-take’ medical team of the day (n = 2652).

Rate of unplanned readmissions within 7 and 28 days.

No change in the rates of unplanned readmissions within 7 and 28 days. At 7 days 3.8% (pre AAU) vs 3.7% (post AAU). At 28 days 8.7% (pre AAU) and 8% (post AAU) (p = 0.80).

Observational, uncontrolled study. May be affected by unknown bias and confounders.

Roberts et al., 2010, UK (Northern Ireland) [19]

Patients ≥16 with probable medical conditions, likely to be admitted through processes of standard ED care, but may potentially have been managed by a GP or as an out-patient following senior review.

Retrospective cohort. CDU cohort compared to three age-stratified, historical cohorts from same clinical centre.

Clinical decision unit (CDU) located within ED. Pilot CDU (3 beds). Staffed by middle-grade physician and experienced nurses.

All patients who participated in the pilot CDU were included in the study cohort (n = 854). Most patients in the CDU group sourced from the ‘Major’ area in the ED.

Three comparison cohorts chosen from the preceding 3 years −2003, 2004 & 2005. These patients identified as those classified as ‘Medical’ by triage nurse a group most likely to have been diverted to the ‘Major area’. These were selected on an age-stratified basis, using the study cohort as the template (n = 854 for each cohort).

30-day unplanned re-attendance rate for those not hospitalized, and monthly medical admission figures.

Significant difference found in admission patterns of the different cohorts. Approximately 511 (59.8%, 95% CI: 56.5-63.1%) to 560 (65.6%, 95% CI: 62.3-68.7%) admitted in comparison group vs 186 (21.8%, 95% CI: 19.1-24.7%) in CDU (intervention) group P < 0.05. A greater proportion of patients from CDU had unplanned re-attendances 11.8% (95% CI: 9.5-14.5%) compared with between 4.4% (95% I 2.6-7.4%) and 7.5% (95% CI: 5.1-11%). P > 0.05 NOT SIGNIFICANT for all cohorts. Modestly significant compared to 2003 and 2004 cohorts.

Historical cohorts can’t exclude residual confounding.

Rogers et al., 2011, UK [20]

Adults (≥18 years). All GP referrals with a view to medical admission, but that are possibly avoidable, included either in MAU and/or by the GP support unit (GPSU).

Before and after study. Observational analysis. Analysis of number of patients referred and admitted to an MAU during a 6 month intervention period compared to control period.

Emergency MAU in one acute hospital.

Team of GPs working near emergency MAU (GP support unit). All GP emergency medical referrals made between 10:00–19:00 on weekdays discussed with GPSU rather than MAU.

6 months prior to GPSU in situ.

Number of patients referred and admitted on week days by different modes (A&E, GP and GP via A&E). Total number of referrals and admissions.

Mean number of GP referrals to MAU per day decreased by 1.55 (−2.45 to −0.51). Non-significant decrease in mean number admitted to hospital per day from MAU 0.48 (−1.39 to 0.44). GP admissions not targeted through GPSU increased by 3.99 per day (2.64 to 5.33). Modest reduction in GP admissions to MAU but no reduction in number of GP admissions to hospital wards.

Before and after design.

Ong et al., 2012, Australia [13]

Patients ≥65 years. Diagnosis groups: falls and gait disorder, COPD, other major respiratory diseases, cellulitis. Target patients those requiring a short stay admission with potential discharge within 48 hr and sub-acute patients with multiple-comorbidities.

Retrospective case–control. Medical files of patients reviewed.

MAU and general medical ward. MAU “Assess and manage undifferentiated patients for 36-48 h before transfer to medical ward or discharge home.”

Patients admitted to Medical Assessment Unit (MAU) before ED assessment completed and allied health review initiated when required (n = 47).

Patients admitted to General medical wards through standard ED assessment and management (n = 42).

Hospital readmissions in 1 month

No significant difference in readmission rate. Readmissions within 1 month similar in both groups (4.2% MAU) and (4.8% non-MAU group). MAU group shorter ED LOS (4.9 + − 3 h vs 6.5 + − 2.8 h, p = 0.012).

Small sample size and short duration of study. Retrospective design. Confounding.

Koehler et al., 2009, USA [25]

High-risk elderly medical in-patients. ≥70 years, use of ≥ 5 medications regularly, ≥ 3 chronic comorbid conditions, require assistance with ≥1 ADL (predisposed to unplanned readmission or ED re-attendance). Patients enrolled within 72 hours of admission and likely to be able to be discharged home.

RCT – pilot.

Medical in-patients. 2 medical units.

Intensive patient-centred educational program (by ‘highly experienced’ research staff) starting no later than 24 hours after enrolment. Medication counselling/reconciliation, condition specific education/enhanced discharge planning by a care coordinator, and phone follow-up (n = 20).

Usual care (n = 21).

Unplanned hospital readmission or ED visitation at 30 and 60 days post discharge.

0-30 day post discharge readmission/ED visit rates lower in intervention group (n = 2 vs 8) p = 0.03. No difference in 31–60 day readmission/ED visits. Longer time to first visit event in intervention vs usual care group (36.2 versus 15.7 days p = 0.05).

Small sample size. Incomplete blinding. Pilot study.

Lisby et al., 2010, Denmark [34]

Patients ≥70 years, in acute internal medicine ward and taking at least one drug daily with expected admission >24 hr.

RCT, non-blinded.

Acute Internal medicine ward.

Clinical pharmacist conducted medication reviews and drug counselling after usual medication review in the ward. Medication history conferred to pharmacologist and medication changes recommended (n = 50). Intervention conducted within 24 hr of admission or by first-coming day of week.

Usual medication review in ward (n = 49).

Usual medication review on admission (junior physician) and within 24 hr of admission by senior physician. Ward physicians not obliged to follow recommendations of routine medication review.

Number of emergency department visits. Readmissions.

No difference in ED visits Mean (95% CI) Intervention 0.1 (0.0-0.2) and control 0.1 (0.0 to 0.2). No significant difference in readmissions intervention 0.4 (0.3-0.6) and control 0.5 (0.3-0.7).

Possible contamination bias. Trial in one clinical setting and contamination bias could have optimized drug prescriptions in the control arm.

Insufficient statistical power to detect a significant difference.

Bowles et al., 2014, USA [23]

Hospitalized patients aged ≥55 years.

Study data collected within 24–48 hours of hospital admission.

Quasi-experimental study at one medical centre.

4 medical units at one urban hospital, “Primary practice setting”.

The Discharge Decision Support System (D2S2) used to assess patients within 24–48 hrs of admission. Results shared with case managers to alert them of patient’s risk status and to arrange referral for post-acute care where necessary (high-risk – refer and low-risk –do not refer) (n = 252).

Usual care. D2S2 completed but information not shared with case managers (n = 281).

Readmission outcomes at 30 and 60 days.

Percentage of high-risk patients readmitted by 30 and 60 days decreased by 6% and 9% respectively. Showing a 26% relative reduction in readmission of high-risk patients in pre and post intervention phases.

Two-phase study: additional interventions may have resulted in the changes seen. Limited to a single hospital - lacks generalizability.

Goldman et al.,2014 USA [24]

Hospitalized adults ≥55 years with anticipated discharge into community. Patients enrolled who had been admitted in the previous 24 hours.

RCT

Safety-net hospital (provide care for patients at high risk of readmission.) Hospitalized adults (internal or family medicine, cardiology or neurology departments)

In-hospital, one-on-one, self-management disease-specific education by nurse within 24 hours of discharge (in preferred language). Telephone follow-up after discharge (on days 1 to 3 and 6 to 10). Patients had access to telephone support line – calls returned within 24 hours. On discharge patients received ‘After Hospital Care Plan’ booklet (n = 347).

Usual discharge care (n = 353).

ED visits or readmissions at 30, 90 and 180 days after discharge.

No statistically significant differences in ED visits or readmissions between intervention and control groups. HR (30 days) 1.26 95% CI; 0.89 to 1.78 (p = 0.19). HR (90 days) 1.21 95% CI 0.91 to 1.62 (p = 0.19). HR (180 days) 1.11 95% CI 0.86 to 1.43 (p = 0.44).

ED VISITS (not hospitalised) 30 days HR 1.41 95% CI 0.81-2.44 (p = 0.22). 90 days HR 1.41 (0.88-2.24) (p = 0.15). 180 days HR 1.41 (0.97-2.06) (p = 0.07).

Intervention group had greater proportion of patients with 2–5 ED visits.

Study lacked power due to lower than expected rates of readmission. Possible enhanced care given to’usual care patients’. Single centre study.

Greening et al., 2014, UK [17]

Patients aged ≥40 admitted to hospital with an exacerbation of chronic respiratory disease. Patients randomised within 48 hours of hospital admission.

RCT.

An acute cardiorespiratory unit and an acute medical unit.

Early rehabilitation intervention started within 48 hours of admission and delivered by physiotherapists and nurses. Education and self-management package also part of intervention. Intervention lasted 6 weeks. Post discharge unsupervised home based program with telephone support at 48 hrs, two weeks and four weeks (n = 196).

Standard care from in-patient physiotherapist, dietician referral if necessary. Out-patient pulmonary rehabilitation offered three months after discharge (n = 193).

Readmission rate at 12 months. Readmissions for respiratory and other causes.

No significant difference in readmission rates between intervention and control groups (HR 1.1, 95% CI 0.86 to 1.43, p = 0.4).

Excluded patients with more than 5 admissions in the preceding 12 months.

Kampan, 2006, Thailand [35]

Type 2 Diabetic patients hospitalized with hypoglycaemia.

RCT

One hospital

Counselling and clinical pathway for treatment of hypoglycaemia. Assessment and treatment within the first 3 consecutive days of hospitalization (n = 33).

Conventional treatment for hypoglycaemia (n = 32).

Readmissions with recurrent hypoglycaemia at 1 and 3 months.

Significant decrease in readmissions with hypoglycaemia at 1 and 3 months in intervention compared to control group (6.06% intervention vs 34.38% control group; p = 0.036).

Insufficient evidence regarding randomisation. Study staff aware of treatment allocation. Likely not blind to intervention.