Search strategy

The comprehensive search strategy was developed by an Information Specialist. EMBASE, MEDLINE, PubMED, the Cochrane CENTRAL Registry of Controlled Trials, and the Cochrane Database of Systematic Reviews were searched from 1950- January 26^th, 2015. The EMBASE electronic search strategy is outlined in S1 Table. Terms such as “emergency service,” “emergency department,” “health care,” and “health service” were combined using the Boolean operator “and” with terms such as “frequent,” “over-use*,” “overutiliz*,” and “super-user.” Results were limited to English language, and excluded comments, editorials and letters. No other limitations were used.

Selection of Literature

Studies were included if: they reported original data; had a control group (controlled trials or prospective comparative cohort studies); were set in an emergency department or acute care facility; focused on a general adult frequent ED user population; and, examined the impact of an intervention to reduce the ED utilization of frequent ED users. No fixed definition of frequent user was applied; any definition used within the included studies was accepted. Studies were excluded if they did not meet the criteria above, or if they only assessed a specific demographic or clinical group of frequent users (e.g. seniors, those with asthma, migraine suffers, homeless). References lists of previous systematic reviews were also hand-searched for relevant articles not identified in the search of the literature. All screening was completed in duplicate, with any disagreement between reviewers resolved through discussion and consensus, or through consultation with a third reviewer. A kappa statistic for agreement was calculated.

Data Extraction and Analysis

Data from the included studies was extracted in duplicate using standardized data extraction forms. Any discrepancies in data extraction were resolved through consensus and discussion. For all studies, study type, publication date, country, number of participants, health care setting, definition of “frequent user”, type of intervention and comparator(s), and the primary outcome of interest (e.g. the reduction in ED visits between groups or pre- and post-intervention) were extracted. For studies that did not report the change in ED visits, the difference in median or mean number of ED visits was calculated where possible. The secondary outcomes of cost, related to the implementation of the intervention and/or hospital charges before and after the intervention, were also extracted when available. For studies that did not report change or difference in costs, the difference in per-patient and/or total hospitalization costs before and after the intervention were also calculated, where possible. Costs were primarily reported in United States Dollars (USD; $) or Euros (EUR; €) based on the study setting; any exceptions were otherwise noted. Heterogeneity in the study populations, healthcare settings, intervention and comparator types, as well as reported outcomes, prohibited meta-analysis of study outcomes. All data analyses (i.e. calculation of kappa statistic, change in ED visits and cost outcomes) were conducted using STATA 13.1 IC (College Station, Texas, USA).

Quality Assessment

The Cochrane Risk of Bias Checklist was used to evaluate the quality of the included RCTs.[26] Broadly, seven areas of risk of bias are included in this checklist (random assignment generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and any additional potential sources of bias) with each area given either “low,” “high,” or “unclear” risk of bias.[26] Comparative cohort studies were assessed using the Downs and Black Checklist.[27] This checklist includes 27 criteria, widely covering areas reporting quality, external and internal validity, and power[27]. Studies are assigned a value of “1” if they meet the question criteria, and “0” if they do not or if it is not possible to determine; with one exception where one question may be given “2” points.[27]

Characteristics of Included Studies

One thousand and twenty-nine citations were identified using the search strategy outlined above. Of those, 952 were excluded based on abstract review. The remaining 77 studies proceeded to full-text review. Systematic reviews were also hand-searched and 4 additional relevant articles were identified. Sixty-four were then excluded following full-text review and 17 articles were included in the final analysis (Kappa = 0.905, 95% CI 0.775–1.00) (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA Flow Diagram.

A total of 1491 abstracts were identified from the electronic database search. After removal of duplicate records, 1029 abstracts were reviewed and 952 were excluded. Hand-searching of the references lists of relevant systematic reviews identified 4 additional full-text articles. Eighty-one articles in all were assessed in full-text, of which 64 were excluded and 17 studies (4 randomized controlled trials; 13 comparative cohort studies), within 3 intervention categories (care or case management; individualized care plans; information sharing), were included for final qualitative synthesis. Heterogeneity in the reported outcomes prohibited pooling of data for meta-analysis.

https://doi.org/10.1371/journal.pone.0123660.g001

The characteristics of included studies are summarized in Table 1. Included articles were published between 1997 and 2014. The majority of studies were in the United States (n = 8);[18–20,23,28–32] three were conducted in Australia[21,33] and New Zealand;[34] two were conducted in Sweden;[35,36] one was conducted in Canada;[22] one was conducted in Scotland;[24] and one was conducted in Taiwan.[37] Four of the studies were RCTs[18,28,35,36] and the remaining 13 were designed as non-randomized comparative cohort studies.[19–24,29–34,37] The study follow-up periods ranged from 6 to 12 months post-intervention and the primary outcome of interest was the change in mean ED visits either between groups or before and after the intervention. Sixteen of the included studies were conducted in a single acute care facility; often a public and/or academic hospital in either an urban and rural setting.[18–24,29–37] The patient population in the remaining study[28] was drawn from 7 different acute care hospitals. Sample sizes of the RCTs varied from 25 to 1799 patients and from 14 to 258 patients for comparative cohort studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of Included Studies.

https://doi.org/10.1371/journal.pone.0123660.t001

Quality of Included Studies

Each of the included RCTs had areas where the risk of bias was high, low, and unclear (Table 1 and S2 Table). Risk of bias was consistently high across all four studies in the areas of allocation concealment, and blinding. Since the interventions involved discussion with the participants, blinding and allocation concealment were not possible. Furthermore due to the nature of the examined interventions, treatment allocation was not concealed, nor could the participants, personnel, or outcome assessors be blinded.

The comparative cohort studies included were of low to moderate quality wherein quality assessment scores ranged from 14 to 20 points (Table 1 and S3 Table). Although typically scored out of 28, a modified Downs and Black Checklist[27] was used since some of the questions did not apply to the comparative cohort design (i.e. randomization); thereby reducing the denominator to 26. Studies received poor scores for reporting of adverse events and blinding.

Defining Frequent Users and Control Groups

In 16 of the included studies, frequent users were defined in the context of ED utilization (i.e. frequent ED users) with a threshold number of 3 to 10 ED visits within the 12 months prior to the study period. In the remaining comparative cohort study, frequent ED use was defined as 3 or more ED visits within a 72 hour period in the month prior to the study.[37] The control groups in the 4 RCTs were comprised of randomly assigned subjects not exposed to the intervention of interest. For the majority of the comparative cohort studies (n = 8) a pre- and post-intervention design was used, where the population exposed to the intervention served as their own historical control groups.[19–22,31–33,37] The control groups in the remaining studies (i.e. not exposed to the intervention) were either sampled from the same population as the intervention groups[23,24,30]—in one case utilizing the total sample population as the control group[29]—or sampled from the control group of previous work[34].

Types of interventions

Three types of interventions aimed at reducing the number of ED visits by frequent users were identified from the literature review: case management (also referred to as care management),[18–24,29,32,34,36,37] individualized care plans,[28,30,33] and information sharing.[31,35] Brief accounts of the specific interventions or combination of interventions as well as the composition of resources related to each intervention, as described by the included studies, are provided in Table 1.

Case Management

Broadly defined, case or care management (CM) is considered a comprehensive, interdisciplinary approach taken to assess, plan, personalize, and guide an individual’s health services to promote improved patient and health system outcomes.[19] A single point of contact (e.g. an individual described as either a case manager, care manager, or ED consultant) is assigned to a frequent ED user and is tasked with brokering access and guiding the patient through their customized care process, which may extend beyond the normal continuum of the ED and in-patient care, into the community.[5]

Among the 12 studies (2 RCTs; 10 comparative cohort studies) that examined CM as an intervention for frequent ED users (see Table 1), interdisciplinary teams consisting of case managers, primary care physicians, nurses, social workers, other allied health, and ED physicians and staff were identified to implement the proposed CM intervention.[18–24,29,32,34,36,37] The specific approaches employed for each CM intervention varied across studies and included a variety of health services such as crisis intervention, supportive therapy, linkage to primary care and other medical providers, referral to substance abusive services, regular CM meetings and needs assessments, and assistance with obtaining stable income and housing.

The mean or median number of ED visits before and after the CM intervention for the included studies are summarized in Table 2. Compared to the control groups, one RCT reported no change in the mean number of ED visits following CM,[18] whereas the second RCT reported a minor decrease in median ED visits among those in the intervention group.[36] Of the 10 comparative cohort studies evaluating a CM intervention, nine studies reported outcomes related to the change in ED visits: eight studies observed a decrease in the mean (between -0.66 and -37 ED visits) [19,37] or median number of ED visits (between -2.28 and -20 ED visits) compared to the controls[29] or before CM;[20,22–24,32] and 1 study reported an increase of 2.79 median ED visits post-intervention.[21]

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Reported Outcomes of Frequent User Interventions Among Included Studies.

https://doi.org/10.1371/journal.pone.0123660.t002

Of these 12 included studies, six (2 RCT and 4 comparative cohort studies) also assessed the change or difference in cost following the implementation of CM.[18,20,23,29,32,36] The results of these studies are summarized in Table 3. The two RCTs specifically assessed the costs of CM programs from a health system perspective. Shumway et al. reported increases in the cost of care for all participants in the 12-month follow-up period.[18] This increase in cost, however, was significantly less among those exposed to the CM intervention compared to those in the control group (CM: $3116 added costs per-patient vs. control: $6659 added costs per-patient; p<0.01).[18] The specific cost of the CM intervention was also reported as $606,711, or $3,633 per-patient.[18] The RCT conducted by Reinius et al. only reported the health system costs following a telephone-based CM intervention (i.e. direct cost of implementing the CM intervention was not reported).[36] The total costs per-patient were reported as €6,355 for the intervention group and €19,044 for the control group; this estimated 45% decrease in cost was found to be statistically significant (p = 0.004).[36]

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Health System Costs of Frequent ED Use and/or Interventions.

https://doi.org/10.1371/journal.pone.0123660.t003

Four comparative cohort studies also evaluated the cost of CM from a health system perspective (Table 3).[20,23,29,32] Broadly, all four studies reported reduced hospital costs (i.e. ED and in-patient charges) per patient in the 12-months following a CM intervention. The Shah et al. study reported a modest decrease of $671 in hospital charges per patient.[23] Whereas the greatest reduction in median per-patient hospital costs was $7,473, reported by the Okin et al. study; this resulted in a cost savings of $429,464 in the hospital charges for the entire intervention group (n = 53).[20] This study also reported a net cost savings (i.e. the cost of the intervention was subtracted from the savings due to the intervention) of $132,726.[20] Lastly, costs specifically related to the implementation of the CM programmes were reported by 4 studies (1 RCT; 3 comparative cohort studies) and ranged from $66,000 (or $1,833 per patient) to $606,711 ($3,633 per patient) for CM.[18,20,29,32]

Individualized Care Plan

One RCT and two comparative cohort studies examined the impact of individualized care plans on ED visitation rate amongst frequent ED users.[28,30,33] Similarly to CM, individualized care plans employ interdisciplinary strategies, including cross-departmental care plan meetings[28] and coordinated access to primary care resources.[30,33] However in contrast to the CM approach, individualized care plans were described as less comprehensive in their design, limited in the number of health services and, importantly, implemented without a designated case manager or equivalent.

The one RCT which evaluated implementation of individualized care plans found no difference in the median number of ED visits between intervention and control groups 12 months post-intervention (care plan group: -7 ED visits; control groups: -7 ED visits).[28] The two comparative cohort studies did not report outcome data related to the frequency of ED visitations following care plan implementation (Table 2).[30,33]

Of these 3 included studies, the two comparative cohort studies provided cost-related outcomes from a health system perspective (Table 3).[30,33] The first study reported a reduction in hospital charges of $742 per patient following the implementation of individualized care plans.[30] The second study did not provide the change in per patient hospital costs, yet reported the implementation costs of the individualized care plan program which was $63,434 Australia Dollars (AUD) or $1,714 AUD per patient.[33]

Information sharing

The term ‘information sharing’ was used to describe approaches related to the sharing of patient information (clinical and/or demographic information) amongst health care providers. Two studies (1 RCT and 1 comparative cohort study) evaluated the implementation of different information sharing strategies to frequent ED user populations.[31,35]

The RCT implemented an information sharing strategy wherein ED and primary care physicians shared access to printed case notes, through an electronic database system, of each frequent user patient.[35] Six months post-intervention, this form of information sharing did not result in a significant difference in the mean number of ED visits between treatment groups (information sharing group: -2.2; control group: -2.1) (Table 2). No cost-related outcomes were reported for this study.[35]

The comparative cohort study by Stokes-Buzzelli et al. utilized an electronic medical records (EMR) system to share individualized care plans between relevant ED physicians and allied health practitioners (e.g. medical and mental health social workers, psychologist, and nurse specialists).[31] This intervention reportedly led to increased consistency in identifying and managing frequent ED users and, ultimately, in a significant decrease in mean ED visits 12-months following the intervention (from 67.4 to 50.5 mean ED visits; p = 0.046) (Table 2). This study also reported a substantial reduction of $15,513 per patient in ED costs in the 12-months follow-up period.[31] The cost of the intervention, however, was not reported by this study (Table 3).