Implementing a Case-management Intervention for Frequent Users of the Emergency Department in French-speaking Switzerland

Frequent users of the emergency departments (FUEDs) and other health care services are of much interest to clinicians, administrators and researchers. Emergency department (ED) overuse is linked to ED overcrowding, which in turn has a negative impact on patients and health system outcomes. There are over 1.4 million annual ED visits in Switzerland, with 84% of EDs reporting overcrowding. FUEDs made up 4.4% of the patient population and made 12.1% of all visits at the Lausanne University Hospital ED. Driving this high use of health care services is the fact that FUEDs often suffer from chronic medical diseases, including heart disease, pulmonary disease and cancer, at high rates, in addition to mental illness and substance use disorders. CM has been shown to be a promising intervention to reduce ED overuse by frequent users in research settings; however, there is limited knowledge about how to disseminate and implement a CM intervention for FUEDs on a large scale to diverse clinical settings, including community hospitals and non-academic centers.

The research project has the following three specific aims:

1. Develop and disseminate a practical CM intervention for FUEDs to several hospitals in the French-speaking region of Switzerland.
2. Study the process of implementation of the intervention.
3. Study FUEDs' trajectories on health outcomes (e.g., ED use, health care reorientation and quality of life) after receiving the CM intervention.

This research project is an observational study with a hybrid study design, measuring both implementation variables and clinical outcomes relating to the dissemination and implementation of the CM intervention. The implementation part of the study uses a mixed methods design (i.e., using both qualitative and quantitative analyses) to describe both qualitatively and quantitatively factors that may influence implementation process. The clinical part of the study uses a within-subject (pre-post intervention) design to evaluate participants' trajectories after receiving the CM intervention.

The whole research project will take place over five phases.

Phases 1-2. Development and exploration Procedures. The I-CaM research team will develop the CM Toolkit, the implementation program, the team member selection-support materials and the informational announcement to be disseminated to hospitals in the French-speaking region of Switzerland. Specifically, the research team will develop and send by email a survey aiming to gauge interest and needs regarding the CM intervention to all eligible hospitals (to key staff, Chief of Emergency Department). The I-CaM research team will follow up with more information about the CM intervention and the study procedures. Then, interested parties will participate in a one-day workshop at Lausanne University Hospital, during which key staff will receive training on the CM intervention, on the implementation science and on the study procedures in general. Following the workshop, all sites agreeing to participate will a) be included in the study and b) complete a questionnaire and participate in a semi-structured interview (assessing implementation outcomes). Further, the same assessment will be conducted with hospitals not interested in participating in the study, in face-to-face or by phone and mail depending upon possibilities (i.e., disinterest analysis).

Phase 3. Preparation Procedures. Included sites will then prepare for implementation of the CM intervention; sites will first identify CM intervention team members, including strategical and operational champions and clinicians. Whereas the strategical champion (i.e., leader) will promote and hold the implementation project, the operational champion will support its implementation and application and supervise clinicians who will be in charge of the CM intervention administration. Next, available resources will be established at each source and data collection and storage systems will be finalized to gather data on health service and implementation outcomes by the research team. The I-CaM research team will also conduct trainings for the CM team to relevant local staff. Finally, at the end of the preparation phase, champions and clinicians involved in the project on-site will complete a questionnaire.

Phase 4. Operation The operation phase. The CM intervention will be implemented at all sites included in the study. At each site, patients fulfilling the inclusion criteria will be contacted by the case managers. The number of participants included will depend upon each site resources. If possible (depending upon resources on each site), eligible patients having been treated in the ED in the 10 days prior the recruitment window will be contacted by phone and proposed to come back to the ED to participate in the study. Remaining FUEDs (i.e., the FUEDs not invited to participate in the study) will receive usual care. When first meeting the participant, the case managers will present the study and examine the exclusion criterion. If the patient does not have any exclusion criteria and is interested in participating, the case managers will conduct the informed consent process. After providing written consent, participants will receive the CM intervention. After each inclusion on site, the case managers will up-date the I-CaM research team in charge of the clinical assessment (i.e., baseline and follow-up assessments of the clinical variables to describe participants' trajectories). The I-CaM research team will contact participants and conduct the baseline assessment within 10 days following the inclusion. Participants will then complete follow-up assessments at 3, 6 and 12 months post-baseline. Other clinical variables will be directly extracted from medical records on-site by the I-CaM research team. At the end of the operation phase, the research team will conduct semi-structured interviews with case managers and champions. Both champions and case managers will also complete a questionnaire.

Phase 5. Sustainability Procedures. Following the implementation of the CM intervention, the sites will continue the CM programs at their discretion. The I-CaM research team will monitor the activity at each site and will be available for as-needed support to all sites. Clinical outcomes (i.e., ED use, quality of life, etc.) will not be measured over this finale phase. At the end of the sustainability phase, the research team will conduct a semi -structured interview with clinicians and champions. Both champions and clinicians will also complete a survey.

• The development and exploration phases (no patient recruited) started in August 2017 and will end in September 2018.
• The preparation phase (no patient recruited) started in March 2018 and will end in December 2018.
• The operation phase (patients' recruitment start; assessment of health-related outcomes) will start in October 2018 and end in April 2019.
• The sustainability phase (patients receive the CM intervention but health-related outcomes are not assessed) will start in April 2019 and end in November 2020.

Data Analysis Plan Implementation measures analysis. Descriptive statistics will be used to describe participants' characteristics and to report implementation outcomes. The investigators will also test implementation measure changes over time. Given the expected small sample size regarding implementation outcomes, these analyses will be triangulated with qualitative data. Specifically, interview contents will be transcribed and explored to identify participants' recurring codes and categories; the investigators will use conventional content analysis (i.e., a systematic process of coding and classification) [1] using a qualitative software (i.e., Atlas.ti or NVivo) to examine qualitative data.

Clinical outcomes. Data will be screened for missing cases, outliers, and normality of distributions using descriptive statistics and plots. The investigators will take appropriate steps to deal with missing data. First, they will conduct analyses to detect missingness patterns and test whether they may be considered "ignorable."[2] If more than 5% of outcome data are missing,[3] the sample will be divided into 2 groups (i.e., missing, not missing), and fully observed variables to predict missingness on the affected outcome will be used. If fully observed variables are not significant predictors, missingness may be considered as "observed at random" and fulfilling some criteria for "missing completely at random" (MCAR) assumptions. In that case, the investigators will use multiple imputations procedures for measured outcomes and direct maximum likelihood estimation for structural models. If data missingness is non ignorable (MNAR), pattern-mixture models with multiple imputation to model the missingness mechanism will be used.

Main analyses will comprise multilevel models (i.e., MLM; mixed effects model) [4] utilizing appropriate distributions for the outcome variables (e.g., Poisson, negative binomial, normal). MLM examine the effect of time (after receiving the CM intervention) on clinical outcomes (e.g., quality of life, self-efficacy). MLM is appropriate to handle nonindependence data. Specifically, data will be clustered by participants (i.e., repeated measures) and by hospitals. MLM does not assume independence of observations. Dependence is modeled through random effects (representing different sources of variability in the data). The investigators will include sources of random variability at the group level accounting for between-group differences and another random effects for the individual accounting for within-person differences in the repeated measures. MLM will be adjusted for demographic variables (i.e., age and gender), health-care utilization (including ED use), at-risk behaviors (e.g., alcohol use disorders) and health status. Descriptive statistics will be conducted on SPSS and MLM on STATA. The significance level will be set at p = .05.