Study design and setting
This is a retrospective, observational, single-center cohort study. It includes critically ill older adults, recruited at the ED at the Northern Älvsborg-Uddevalla (NU) Hospital Group, Region Västra Götaland, Sweden, between February 2013 and February 2014. This county hospital has an uptake population of approximately 280 000 inhabitants.
The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice Guidelines, and was approved by the regional ethical review board at Sahlgrenska University Hospital in Gothenburg, Sweden (D.no. 962–13). The study was registered at the Swedish National Database of Research and Development; identifier 142,071 (http://www.researchweb.org/is/vgr/project/142071; February 5, 2014) as Medical Emergency Care (MEC)—an observational study of the emergency care of the critically ill medical patient. Before a secondary data collection regarding long-term mortality was performed, complementary ethical approval was given by the Swedish ethical review authority (D.no. 2020–04,407), waiving the need for a renewal of the informed consent. Due to the expected high mortality rate it would not have been possible to collect a second informed consent.
Data collection and participants
The primary data collection has been described previously in Bergh et al. [20] All adult internal medicine patients treated in the ED and classified as critically ill in accordance with the Rapid Emergency Treatment Triage System (RETTS) [21] were included consecutively. RETTS, developed for risk assessment in EDs, has been used in order to perform a sensitive identification of critically ill patients [22, 23]. It relies on the following vital signs (VS): airway obstruction/stridor; oxygen saturation < 90%; respiratory rate < 8 or > 30 per minute; regular heart rhythm > 130 or irregular heart rhythm > 150 beats per minute; systolic blood pressure < 90 mmHg; unconsciousness, defined as Reaction Level Scale (RLS) > 3 or Glasgow Coma Scale (GCS) < 8; ongoing seizure [20, 22, 23]. Simultaneously the symptoms that caused the contact with health care is to be considered (the Emergency Signs and Symptoms code [ESS code]). The combination of VS and ESS gives the patient a colour of either red, orange, yellow, green or blue in order of severity of the condition and reflecting the time required to assessment by a physician. In this study we included patients given the colour red, reflecting urgent requirement of a physician assessment, i.e. critically ill patients.
The exclusion criteria were lack of written informed consent, if a patient was wrongly registered, and if the patient was treated for cardiac arrest, need for acute percutaneous coronary intervention (PCI) or included in the acute stroke fast track [20]. Patients with trauma or other surgical conditions were excluded. Information was collected retrospectively from the ambulance records and the hospital medical records.
A secondary data collection was performed regarding mortality until December 31, 2020. This information was extracted from the State’s Personal Address Register (SPAR) at the Swedish Tax Agency. This is a comprehensive state agency register, which includes all persons who are registered as residents in Sweden. The data in SPAR are updated every day with data from the Swedish Population Register [24], and are a reliable source of information regarding death and survival confirmation.
Approximately 7% of all internal medicine patients ≥ 70 years of age admitted to the ED were critically ill. Of 832 patients correctly classified as critically ill, written informed consent was given by 610 patients [20]. For the analyses described here, patients aged ≥ 70 years were selected.
Methods and measurements
Clinical and demographic characteristics were primarily collected at index admission to the ED from the patient ambulance records and subsequent medical records from the ED and the hospital medical wards. The following variables were recorded: age, sex, date and time of arrival at the ED, main symptoms and VS in the ambulance, working diagnosis in the ED and medical history including the Charlson Comorbidity Index (CCI) components. At discharge from hospital, the type of department to which the patient was admitted, care in the intensive care unit (ICU) or cardiac intensive care unit (cICU), length of stay (LOS) in hospital, discharge diagnosis and in-hospital mortality were recorded.
Data regarding post-discharge outcomes up to 12 months were collected from medical records. These included information on mortality, re-hospitalizations and total LOS. A secondary data collection was performed regarding mortality, in which all patients were followed-up for 6.5–7.5 years post-discharge, as described in the data collection section. The cases refer to unique individual patients, and re-hospitalizations were registered as an outcome.
The Charlson Comorbidity Index
The patient’s total burden of morbidity was measured by the CCI [25, 26]. It contains 19 categories of comorbidity and predicts mortality for a patient in a general medical context. Each comorbidity is assigned a score of 1, 2, 3, or 6, depending on the risk of death associated with this condition.
The CCI score was dichotomized as 0–2 (mild grade) versus > 2 (moderate or severe grade), a commonly applied stratification [27,28,29].
Outcomes
The primary outcome was all-cause post-discharge death until December 31, 2020.
Secondary outcomes were death within one month after admission to the ED; and total time spent in hospital and numbers of re-hospitalizations up to 12 months post-discharge.
Statistical analysis
Descriptive statistics are presented as number and percentage, mean ± standard deviation or median with 25th, 75th percentiles. A Cox proportional hazards model was used to calculate hazard ratios (HR) and corresponding 95% confidence intervals regarding 30-day and long-term mortality, in both univariable and multivariable analyses.
To identify independent predictors of mortality, we first used stepwise backward selection, starting with a model including age and all the other candidate variables with an un adjusted p-value below 0.30 and using p < 0.05 as the limit for staying in the model. After this selection procedure was finished, we included all the remaining variables with an age adjusted p < 0.30 separately, one at a time, to see whether they contributed significantly to the model (using differences in -2 log likelihood). The above was performed separately for 30-day and long-term mortality respectively.
The Kaplan–Meier (KM) method was used to calculate cumulative mortality curves, using 100—KM survival estimate as an assessment of cumulative incidence. This method was also used for calculation of rehospitalization rate during the first 12 months for patients discharged alive after index, where those non-rehospitalized who died were censored at time of death and comparisons between CCI groups were performed using the log rank test.
Numbers of re-hospitalizations and total days rehospitalized during 12 months after index discharge among those alive at 12 months were compared between CCI groups using the Mann–Whitney U test.
All tests were two-sided and p-values below 0.05 were considered statistically significant. All analyses were performed using SAS for Windows version 9.4.