Study aim
The aim of this study is to compare hospital ED readmission rates within 30 days and disability scores at 3 and 6 months in older adults discharged from the ED of hospitals where GMT provides transitional care or standard in-hospital management only.
This study also intends to describe the patterns and results of ED-based and GMT-led transitional care interventions in intervention group.
Study design
This is a national multicentre, prospective, controlled, quasi-experimental study. All participating centres have an ED and a GMT, some of them providing transitional care. Participants recruited from hospitals where GMT provides transitional care form the “intervention group”, and participants recruited from hospitals where GMT provide standard in-hospital management compose the “control group”.
Settings
The study will take place in twelve hospitals, seven providing transitional care (intervention centres) for older adults discharged from the ED and five providing standard care (control centres) (Table 1).
All 12 centres meet the same criteria for standardized in-hospital management during the ED visit, including (i) identification of patients at risk by the ED team, based on clinical characteristics or screening tools with a procedure for reporting to the in-hospital GMT; (ii) a multidisciplinary GMT working in the ED and providing a standardized geriatric assessment; and (iii) a discharge procedure with at least a medical report and referral to the general practitioner (GP).
-
Intervention centres: seven centres, including three university hospitals and four regional hospitals where GMTs are organised to provide transitional care interventions. We defined transitional interventions as (a) necessarily multidisciplinary, (b) always initiated by a phone call to the patient, relatives or carers within a week following ED discharge, (c) with an incremental follow-up and the possibility of home visits, and (d) based on a community-hospital collaboration. Community-hospital collaboration matches at least one of the following criteria: joint clinical meetings and/or joint home visits and/or a shared professional and/or a shared information system. Each GMT intervention has specific components that are taken into account in the description of each team.
-
Control centres: five centres not implementing transitional care with two university hospitals and three regional hospitals. After identifying the patient at risk, the in-hospital GMT provides standard management and recommendations at the time of the ED visit, without home-based intervention or coordination with community actors.
The principal investigator and the scientific committee will check once a year that centres meet the criteria of the group to which they are allocated. If the criteria of the centre change during the study, there will be a 6-month washout period with no inclusion before allocation to the new group.
Participants
Participants will be included just before they are discharged from the ED. The inclusion criteria are age ≥ 75 years, return to home after an ED visit, and a Triage Risk Screening Tool (TRST) score ≥ 2 indicating a high risk for early readmission and/or loss of independence after discharge [12]. The exclusion criteria are living in a nursing home, being under legal guardianship or incapable of providing consent. All participants provide written consent after receiving oral and written information. Participants with language barriers or severe cognitive or psychiatric disorders may be included if a relative is physically present at the time of the ED visit and consents to the study.
Geriatric mobile team interventions
Following inclusion, the participants will benefit from the usual care specific to the inclusion centre they are admitted to.
In both groups, the standardized intervention consists, in the ED, of medical ED care, risk identification by the TRST, and a comprehensive geriatric assessment (CGA) with tailored recommendations reported during discharge planning. The minimal CGA assesses at least comorbidity, medication, social status, and functional, cognitive and nutritional status.
In the intervention group, the initial in-hospital GMT intervention is followed by a systematic telephone call to the patient and/or his/her caregiver and/or GP between day 1 and day 6 after discharge. Since then, the intervention of the GMT and the out-of-hospital management strategy is deliberately not standardized to assess the specific transitional care pattern provided by each GMT via a pragmatic approach. The characteristics of these interventions will be described.
Endpoints and measurements
Data describing the centres will be collected and updated once a year. These data include the number and qualifications of GMT professionals, the annual number of patients supported in the ED and in transitional care, the number of ED visits for adults of all ages and for adults aged 75 or older, the home discharge-to-hospitalization ratio, the readmission rate, and GMT intervention components, as described in Table 1.
Clinical data will be collected at baseline and during follow-up at three (M3) and six (M6) months, the participants’ health trajectories in the 6 months following the ED visit will be described (Fig. 1).
Baseline data
The data are collected at the time of the ED visit (D0) by the GMT. Prior to inclusion, a TRST is performed by the ED team as part of care to confirm that the patient is at high risk for readmission. Clinical assessment is based on the patient’s anamnesis and medical records. If necessary, the GMT may also rely on information transmitted by a relative or the GP to crosscheck the information. The baseline data are:
-
Sociodemographic data including age, sex, level of independence in activities of daily living (ADL) and instrumental activities of daily living (IADL) the week prior to the ED visit [13, 14], and identification of relatives, family and professional caregivers (healthcare and personal care) before the ED visit, with quantification of weekly time allocated to caring.
-
Medical data including comorbidity assessed by the Charlson index [15], number of drugs in the usual treatment, nutritional risk assessed using the Mini Nutritional Assessment – Short Form (MNA-SF) with measurement of calf circumference [16], and screening for cognitive disorders with the Abbreviated Mental Test 4 (AMT4) scale [17] and grading into mild, moderate or severe stages according to the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-V) criteria.
-
ED data: visit duration and main diagnosis.
-
Healthcare pathway data including the following: history of regular visits to the GP (at least twice a year), history of ED visits or hospitalizations in the previous month, and history of previous geriatric management (in a geriatric ward, a geriatric outpatient clinic or appointments with a geriatrician).
Primary endpoint
The primary endpoint of the study is the incidence of readmission to any French ED between seven (D7) and thirty (D30) days after discharge from the initial ED visit, regardless of the reason for readmission. Very early readmissions (from D1 to D6) will not be taken into account in the primary endpoint, as the time to initiate the first intervention is a maximum of six days in the intervention group. Moreover, these very early readmissions are most often motivated by a rapid deterioration of the initial pathology, a diagnostic error or a possible side effect of a therapy [18]. Hospital-community transitional care is not intended to avoid this type of very early readmission.
These data will be collected from the National Health Data System (SNDS), allowing the identification of admissions to an ED on a national level and not exclusively in the centres involved in the study. Readmission to any French ED is localized in different tables of SNDS. ED visits, which are followed by a hospitalisation, will be found in table T_MCOaaB, where aa represents for the last two digits of the year of event. ED visits, which are not followed by a hospitalisation, will be found in table T_MCOaaFBSTC if patients are admitted to public centres, or table ER_PRS_F if the ED admission is in private centres. Data from SNDS will be merged with data collected in eCRF using determined pairing method, which was authorized by the French Data Protection Authority, CNIL (CNIL n°DR-2023-080).
Secondary endpoints
These data aim to describe the evolution of the subjects’ health trajectories in the 6 months following the ED visit. Quantitative data will be collected in the NHDS, and qualitative data will be collected by telephone interview by an independent investigator at M3 and M6. If a participant cannot be reached by phone or e-mail after 3 attempts at 3 or 6 months, these qualitative data will be considered as missing. However data collected in the NHDS will still be considered until the end of the 6-months follow-up.
Healthcare pathway
Additional data concerning healthcare utilization in the 6 months following ED discharge will be collected from the NHDS: ED visits between D1 and D6 and between D7 and M6, delay to the first ED readmission, the number of ED visits, consultations and outpatient clinic visits, scheduled and unscheduled hospitalizations, cumulative duration of hospitalization until M6, and the number of primary care medical visits. All causes of death will also be collected based on the NHDS.
Evolution of autonomy
An independent investigator will perform follow-up telephone interviews with the patient, caregiver or GP in both groups at M3 and M6. These interviews will assess the place of residence (home, nursing home), independence in ADLs and IADLs, number of professional and informal caregivers and weekly time allocated to caring.
Transitional care description
GMT actions will be identified at M6 after discharge from the ED based on the medical records. These data are the number of telephone calls or email exchanges, home visits, multidisciplinary staff, interactions with community professionals (such as joint home visits, medical appointments, rehabilitation care, requests for social workers), the total duration of follow-up and the reasons for interruption of follow-up (refusal, death, lost to follow-up).
Data collection
Data available in the participating ED and GMT will be collected using electronic case report form (CleanWeb) accessible via two-factor authentication, while data specific to health system consumption will be retrieved via the NHDS, as described elsewhere. Data-management of the eCRF data will be handled by the Clinical Research Unit – Paris Secteur Ouest. The NHDS data will be processed by the CNAM team (Caisse Nationale de l’Assurance Maladie) and migrated to a dedicated server, which ensures the security of information. The eCRF will generate, a table of correspondence between the participants’ inclusion number and a random hook identifier, which will be used to allow data juncture with NHDS data (CNIL n°DR-2023-080).
Statistical analysis
Sample size
Previous observational studies have shown that the proportion of ED readmissions in non-intervention units is 20%, and 14% in intervention units [1]. To detect an expected difference of 6% with at least 80% power and a type 1 error 0.05, a sample of 1228 patients is required. We assumed that 7% of included participants might be early readmitted (between D1 and D6), therefore excluded from principal analysis, it is necessary to include 1322 participants, i.e., 661 per group, to have 1228 assessable subjects. To compare the proportions of ED readmissions between D7 and M1 between the two groups, with a two-sided test (type I error α = 0.05), 80% power,.e., an expected difference between the two groups of 6% [7], it is necessary to have observe 1228 evaluable subjects, i.e., 614 participants per group. Assuming that 7% of the participants included will be readmitted to the emergency room between D1 and D6 [2] and therefore excluded from the main analysis, it is necessary to include 1322 participants, i.e., 661 per group, to have 1228 assessable subjects. The calculation was performed with R software (R Foundation for Statistical Computing, Vienna, Austria. ) (version 4.0.3). The number of subjects required is defined on the assumption of balanced and comparable groups. In case of noncomparable groups, a propensity score analysis will be applied. Propensity score is estimated by fitting a multivariable logistic regression model. Thus, the suggested sample size is sufficient to match the condition of a minimum of 10 observed events per covariate included in the propensity score model.
Statistical analysis
Statistical analyses will be carried out using SAS software (version 9.4 or later) or R software (R Foundation for Statistical Computing, Vienna, Austria. ) version 4.0.3 or later) at the Clinical Research Unit – Paris Nord Val de Seine Secteur Ouest. The study population is defined as all subjects who meet the eligibility criteria and who are not readmitted to the ED between D1 and D6. Sensitivity analyses will be carried out by excluding from the ‘intervention’ group the participants included in the ‘intervention’ group but for whom transitional care is not deployed (for example, participants who are unreachable).
Propensity score model
Comparability of groups at inclusion will be verified. Indeed, a potential selection bias may be observed due to the process of assignment to the intervention being dependent on the inclusion centre.
In the event of an imbalance between groups, selection bias will be limited by the use of the propensity score method. Propensity score i.e. the probability of a patient being in intervention or control group, will be estimated by a function of the baseline characteristics through a multi-covariate logistic regression model. Sociodemographic characteristics, variables having an impact on the probability of belonging to the intervention group, and variables potentially associated with readmission to the ED between D7 and M1 will be considered for the construction of the propensity score, taking into account possible interactions between variables. Once the propensity score balance has been approved, the area of common support in propensity score will be defined using the “minimum-maximum” method or the “trimming” distribution comparison method [19]. The area of common support selected will maximize the number of subjects observed after truncation of the distributions, with also a critical study of balance and overlap.
Primary analysis
The primary analysis is to assess the impact of the GMT intervention on hospital ED readmission between D7 and M1.
If the groups are comparable, a chi-square test of comparison of proportions will be used, at the significance threshold of alpha = 0.05.
In case of imbalance between two groups and the implementation of propensity score method, the primary analysis will then be carried out on the validated sample using the inverse probability weighting method (IPTW). The treatment effect will then be assessed using a logistic regression model at the α = 0.05 significance level. A sensitivity analysis will also be carried out using the nearest neighbour matching method, and if necessary, strata will be applied or calibrated (maximum distance between individuals) to ensure balance between groups.
Secondary analysis
Secondary analyses will be carried out on the same population as that defined for the primary objective.
The time to first readmission to the ED will be compared between the two groups using survival analysis with the log-rank test. Subjects whose death occurred before M6 will be censored from their date of death. The cumulative length of hospital stays within 6 months, loss of independence measured by ADLs and IADLs, the evolution of help system and the place of residence (personal home or nursing home) will be compared between groups. Student’s t test or Wilcoxon test of comparison of means is used to interpret the effect of the intervention on continuous variables, and chi-squared tests or Fisher’s exact tests is performed for categorical variables.
The patient pathway in each group will be defined by a series of observable events, potentially repeated during follow-up. These events, specified among the secondary evaluation criteria, will be described in absolute and relative frequencies according to the group. The delays between events and/or intervention determinants (as described in the intervention planning) will also be described by their mean and standard deviation. Potential factors associated with the performance of a home visit, the cumulative duration of the intervention of the GMT and a scheduled day hospital between D0 and M6 will be studied using linear regression models. Eligible variables for the multiple regression model will be selected at the p < 0.20 threshold. The final regression model for each of these determinants will be determined by stepwise selection of variables, at the α = 0.05 significance level.
The determinants of transitional care in intervention group will be described by absolute and relative frequencies for categorical variables, by median and interquartile interval for numeric variables. The association of observable baseline factors with each determinant will be assessed by a bivariate regression model. Eligible variables for the multiple regression model will be selected at the p < 0.20 threshold. The final regression model for each of these determinants will be determined by stepwise selection of variables, at the α = 0.05 significance level. Logistic regressions will be used for binary endpoints, and linear regressions will be used for continuous numerical endpoints.
All secondary analyses aim at comparing the two groups (as the evolution of autonomy), the same strategy presented for the primary analyses, especially with propensity score analysis and IPTW approach, will be used.
Missing data
Regarding the management of missing data for the main evaluation criterion, readmission to an ED between D7 and M1 will be collected from data available in the NHDS. This data source guarantees the availability of this information until M1 and M6 (at the end of the study). A patient who dies between D7 and M1 will be considered a failure of the strategy.
Trial status
Start of inclusion is planned on second trimester of 2025.
Ethics and dissemination
Sponsorship has been agreed by Assistance Publique—Hôpitaux de Paris (AP-HP, Clinical Research and Innovation Department) for this research protocol. The study protocol was approved by a national ethics committee (notice n° IDRCB 2021-A02657-34) and the French Data Protection Authority (CNIL, notice n°DR-2023-080). The study protocol was registered in the Clinical Trial (ID: NCT05814328; Date: 2023-04-14).
In accordance with Article L.1122-1-1 of the French Public Health Code, no research will be carried out without patient free and informed consent, obtained in writing after the person has been given the information specified in Article L.1122–1 of said Code. Written informed consent will be obtained from all patients, their next of kin, as appropriate.
link