Our objective is to create and validate predictive models of poor outcome in people over 65 years of age seen in emergency departments. To this end, models will be developed and validated internally, adapting the reference values of vital signs to this population segment and also taking into account their most frequent reasons for consultation. We will use the EDEN cohort of 18374 patients recruited in one week in 2019 in 40 Spanish hospitals. Subsequently, they will be externally validated in a cohort of 18000 patients prospectively recruited in 7 hospitals in 2024. In addition, the predictive capacity of the FIM frailty scale will be measured. lnformation will be collected from electronic medical records regarding socio-demographic data, functional status, reasons for consultation, clinical and ED history including comorbidities, previous and prescribed ED treatments and vital signs. The main dependent variable is poor outcome defined as ED exitus, admission to ICU/ICU or conventional admission, as well as mortality, revisits and readmissions 30 days after hospital discharge. Updated poor outcome models will be provided that can serve as a basis for creating triage models adapted to the over-65s in the future.

HYPOTHESIS

Current emergency triage systems (MAT, SET, others), as well as early warning systems (NEWS-2), are unable to detect all geriatric patients at risk of poor outcomes, partly due to:

1. the reference values for vital signs are not the same as in the population under 65 years of age
2. atypical presentations of diseases occur more frequently than in the population under 65 years of age.

3.- It is possible to develop and validate predictive models of poor outcomes in the emergency department (death in the emergency department, admission to the ICU or conventional admission, as well as mortality, revisits and readmissions within 30 days of hospital discharge) that incorporate adapted vital signs and symptoms that are more frequent in this population.

4.- Frailty is related to poor outcomes in the emergency department and will add predictive power to the above models.

Objectives:

1. Establish reference values for vital signs in people over 65 years of age.
2. Identify the most common presentations in emergency departments among this age group.
3. Develop and validate predictive models of poor outcomes (e.g. death in the emergency department, admission to the ICU/UCRI, conventional admission, mortality, revisits, and readmissions within 30 days of hospital discharge) in the emergency department. Incorporate the ranges of adapted vital signs and the most frequent symptoms in this population into these models.
4. Evaluate the predictive capacity of the FIM frailty scale to detect poor outcomes, comparing a model incorporating this scale with those created in Objective 3.

DESIGN: Ambispective cohort study, to be developed in two phases: 1. Retrospective cohort study based on data from the 2019 EDEN study for objectives 1, 2, and 3. 2. Prospective cohort study of patients over 65 years of age treated in the emergency departments of participating hospitals in 2024 for objectives 3 and 4.

SCOPE t.-EDEN (Emergency Department and Elders in Need) cohort study conducted on patients aged 65 years or older who attended Spanish hospital emergency departments belonging to the SIESTA network for care for any condition in 2019. Objectives 1, 2 and the development and internal validation of models of poor outcomes in emergency departments.

2.-External validation cohort: to be created, based on the EHRs of the Galdakao-Usansolo, Basurto, and Araba university hospitals (Basque Country) San Carlos Hospital (Madrid), Hospital del Mar (Catalonia), University Hospital of the Canary Islands (Canary Islands), Miguel Servet Hospital (Aragon), in which patients aged 65 or over who are treated in 2024 will be recruited. Objective 3, external validation, and 4, compare predictive capacity when the frailty scale is added.

ELIGIBILITY CRITERIA:

1.-Inclusion criteria: patients over 65 years of age who attend the emergency department for any reason during the study period. 2.-Exclusion criteria: arrival at the emergency department in critical condition, pre-mortem situations, absence of a carer in situations of severe dependency, or Severe mental illness or the presence of mental pathology and refusal to participate in the study.

3.-Losses: Changes of address that make the patient untraceable; in the prospective cohort: language difficulties, voluntary withdrawal from the study during the one-month follow-up period.

SAMPLE SIZE CALCULATION: Predictive model development studies establish that it is necessary to have at least 10 events of the dependent variable of primary interest (in our case, we calculated this for in-hospital mortality, which is the least frequent event of all those that make up the primary outcome variable) for each independent variable included in the multivariate logistic regression model. (Steyerberg, 2009) Given that our intention is to include a limited but comprehensive number of variables in the multivariate model (predictably, no fewer than 10), we estimate that at least 100 events of the dependent variable will be necessary in the derivation sample to ensure that the regression model converges adequately. In the EDEN cohort, in the period 2019, 0.55% died in the emergency department. This means that in 18,181 patients, it is estimated that there will be 100 events of the dependent variable death. At least 100 events are needed to perform external validation (Collins GS, 2016). That is why we would need another 18,181 patients to be divided among the participating centres, 8 per day, 13,102/2023, 2020 per centre.

SAMPLING: Consecutive convenience sampling, during one week per season of the year.

EXPOSURE VARIABLES:

1. Sociodemographic characteristics: age, sex, manner of arrival at the emergency department, previous medical consultation and with whom, situation of accompaniment at home, and whether social assistance is available.
2. Related to the patient's baseline functional status: dependence according to the Barthel Index -18 -, comorbidity according to the Charlson Comorbidity Index -CCI -, whether they have had falls in the previous 6 months and whether there are previous diagnoses of dementia and depression).

Chronic treatments. e) Emergency episode: reason for consultation, vital signs, MAT/SET triage level, additional tests requested, laboratory test results, length of stay in the emergency department (time of discharge-time of arrival) e) Resource consumption: 6 relating to diagnostic tests, 13 relating to therapeutic actions and 5 relating to the use of structural resources: need for observation in the emergency department, hospitalisation, intensive care during hospitalisation and total length of stay in the emergency department (for the entire cohort) and in hospital (only for hospitalised patients).

d) Destination upon discharge e) FIM scale that allows the subject to be classified into four categories after viewing 4 vignettes.

OUTCOME VARIABLES:

Primary: Composite variable: emergency department deaths, transfer to critical care unit from emergency department (ICU/UCRI), hospital admission, in-hospital mortality.

Secondary: Follow-up: return visits/readmission/death up to one month after discharge, functional deterioration at one month (by telephone) in survivors discharged.

DATA COLLECTION METHODS. 1.-EDEN cohort: A multipurpose registry was created that included all patients who visited the emergency department, regardless of the reason for their visit, in 2019 and 2020. Two different periods were established, each comprising a full week. Information was extracted from 228 primary variables corresponding to sociodemographic data, comorbidity, functional status and baseline treatments, clinical aspects, consumption of diagnostic and therapeutic resources, and final diagnosis in the emergency department, patient disposition after emergency care, hospitalisation (if any) and follow-up after discharge from the process (whether directly from the emergency department or after hospitalisation). Follow-up was carried out electronically by consulting the patient's medical records. b) Prospective cohort of patients from the nine participating centres aged over 65 who attend the emergency services in 2024, from which the same information will be extracted as in the retrospective study. In addition, the FIM frailty scale will be administered.LIMITATIONS: We are not taking into account situations of deterioration that do not reach CPR, as we do not have clearly identifiable variables in the large databases to record them. Situations with a poor prognosis in which the patient will not benefit from a stay in the ICU, as well as expected CRPs and deaths, are not recorded in the databases. This represents a non-differential selection bias, as there is no way to differentiate them in the sampling through the electronic medical record, which will underestimate the effect of the intervention.

ETHICAL CONSIDERATIONS: The EDEN project was approved by the Clinical Research Ethics Committee of the Hospital Clínico San Carlos in Madrid (protocol HCSC/22/005-E). Due to the characteristics of the study and the time periods for which data collection was planned, exemption from written informed consent by patients was accepted. The database was used with coded patients to preserve anonymity. The creation of the EDEN and EDEN-Covid cohorts and the work arising from them have at all times followed the ethical principles of the Declaration of Helsinki. The IPs of the EDEN project will submit an addendum to this committee. This project will be approved by the CEIM -Euskadi.

Inclusion Criteria:

• patients over 65 years of age who attend the emergency department for any reason during the study period

Exclusion Criteria:

• arrival at the emergency department in critical condition in pre-mortem situations,
• absence of a carer in situations of severe dependency
• presence of mental illness, and refusal to participate in the study.

patients over 65 years of age who attend the emergency department for any reason during the study period

We will share the information in zenodo

https://zenodo.org/records/15729462

https://zenodo.org/records/15729462