Design, setting, and population
We prospectively screened patients admitted to the ED of a tertiary university hospital between September 30, 2019, and March 17, 2020. Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo is a 2,200-bed hospital located in Sao Paulo, Brazil, dedicated to the care of high-complexity medical and surgical patients. This report is published in accordance with the STROBE guideline and recommendations.13
Eligible patients were 65 years or older and hospitalized for less than 24 hours. We excluded candidates according to the following criteria: (a) previous hospitalization in the 30 days preceding admission; (b) hospitalization for end-of-life care; and (c) expected hospital discharge in 48 hours or less.
The study was approved by the local institutional review board (Comissão de Ética para Análise de Projetos de Pesquisa do HCFMUSP [CAPPESq], no. 77169716.2.0000.0068, October 11, 2017) and registered in Brazilian Clinical Trials Registry [(ReBEC), no. RBR-233bct)]. We obtained written informed consent from all participants or their legal representatives and used REDCap® (Research Electronic Data Capture) resources to secure and manage all study-related data.14
Baseline characteristics
Trained investigators completed the study interviews and assessments using standardized REDCap forms. We collected baseline sociodemographic and clinical data including age, sex, literacy level, medical history, Charlson comorbidity index (Charlson),15 frailty status using the FRAIL scale,16 polypharmacy (chronic use of five or more medications), and admission diagnoses. We performed functional and cognitive assessments using the activities of daily living (ADL) and the 10-point Cognitive Screener (10-CS) scales, respectively.17
Delirium assessments
We completed the Confusion Assessment Method (CAM) algorithm5 twice daily to detect delirium throughout the hospital stay. Our standardized interview protocol incorporated a brief neuropsychiatric anamnesis, cognitive screening (10-CS), attention testing (days of the week backwards and vigilance A test),18 level of consciousness assessment (Richmond Agitation and Sedation Scale [RASS]),19 and electronic medical record revisions.20 Delirium episodes were considered resolved if the patient was non-delirious for two consecutive evaluations.
Our raters attended training sessions before the study initiation, which included simulations and bedside evaluations, and we achieved high interrater reliability levels for CAM-based delirium diagnosis (> 95%). Even so, whenever our raters were uncertain regarding the presence of delirium, two experienced physicians (FBG and JCGA) repeated or reviewed the assessments to adjudicate the final diagnosis.
Blood samples
We collected the following laboratory tests upon study inclusion: blood count, C-reactive protein, platelets, creatinine, blood urea nitrogen, bilirubin, inflammatory biomarkers (IL-1B, IL-4, IL-10, INF-g and TNF-a and neuronal injury biomarkers (S100B, Neuron Specific Enolase and Tau protein).
Three registered nurses performed the sampling while patients were in the ED, which consisted of 30ml of blood collected by venipuncture. Blood samples used for brain injury biomarkers analysis were immediately centrifuged for 10 minutes, and plasma was preserved at -20ºC for up to 48 hours before being transferred to a -80ºC freezer for long-term storage and further processing.
We measured cytokine plasma levels using the magnetic bead immunoassay Milliplex® and the MAGPIX® System (Merck Millipore, USA).
The sampling procedures were performed on inclusion (S1) and repeated 72 hours after inclusion (S2).21 Participants who were discharged or died within 72 hours of admission, or refused to provide additional samples, were not punctured again. We obtained a third sample (S3) from participants who converted either from a negative to positive CAM (incident delirium) or from a positive to negative CAM (delirium resolution) after S2 (Fig. 1).
Statistical analyses
Our primary outcome was the overall occurrence of delirium based on the CAM criteria. Secondary outcomes were association between S100B, NSE and Tau protein and delirium diagnosis and patients’ outcomes (admissions to intensive care, length of hospital stay, and in-hospital mortality).
We used a convenience sample, which limited the total number of enrolled patients. Despite our ED providing medical care to 800 elderly patients monthly and 30% of them being eligible for hospitalization, most of these patients are transferred from other less complex hospitals and had been hospitalized for more than 24 hours at the time of recruitment. Therefore, 200 patients were expected to be enrolled for 6 months, with the expectation of 7–10 eligible patients per week. We finished our recruitment a month ahead of schedule because of the beginning of the COVID-19 pandemic.
All enrolled patients were included in the analysis of primary and secondary outcomes on an intention-to-treat basis. We initiated the analysis using the Shapiro-Wilk test to determine which variables were normally and non-normally distributed, especially results of S100B. Then, we performed analysis according to delirium occurrence using unpaired t-tests or Kruskal-Wallis for normal and non-normal variables respectively, and categorical variables were analyzed using Pearson’s χ2 test.
All analyses were performed with Stata software, version 10.