Study design
This retrospective, single-center longitudinal cohort study at a University Hospital in Switzerland was part of a large Health Service Research project, which evaluated the prevention, screening and treatment of delirium in hospitalized patients. Results from the overall cohort including 10’906 hospitalized patients have been reported recently [12].
Setting
This University Hospital has approximately 39’000 admissions annually distributed across 43 departments and institutes. In the year 2014, a total of 4’002 patients were treated in one of the specialized ICUs for medical, abdominal and thoracic surgical, cardiovascular surgical, trauma surgical, and neurosurgical as well as burn patients.
In the year 2012, a concept for delirium management and Health Service Research project (Delir Path) was launched in all departments by a multi-disciplinary and multi-professional expert team. By covering all aspects from screening to pharmacological and non-pharmacological treatment its aim was the improvement of prevention, early recognition and treatment of delirium in hospitalized patients. Physicians and nurses received training via lectures, e-learning modules and bedside teaching, and had access to the developed algorithms available as pocket cards and on the hospital’s intranet. These algorithms comprised the screening of all patients with a Richmond Agitation Sedation Scale (RASS) [13] score of -3 to + 4 with the Intensive Care Delirium Screening Checklist (ICDSC) [14], performed by trained nurses once per shift. Positive delirium screening corresponding to an ICDSC score ≥ 4 was followed by appropriate pharmacological treatments and non-pharmacological measures. The drug of first choice was the neuroleptic pipamperon, which exists as tablets or as syrup and is strongly sedative while having a weak anti-psychotic action. If hallucinations occurred the neuroleptic drug haloperidol was added orally or intravenously. Vegetative symptoms were treated with intravenous clonidine or dexmedetomidine. Patients with nocturnal agitation, insomnia or increased risk of non-convulsive epileptic seizures received intravenous midazolam via a continuous infusion with doses between 0.05–0.1 mg/kg/h, which was interrupted daily at 6:00 a.m.
Participants
Data of all adult patients ≥ 18 years admitted to one of the six ICUs between 1st of January and 31st of December 2014 were included in the longitudinal cohort study. Patients from intermediate care units, patients with missing data and / or ICDSC score as well as patients without malignancy as principal diagnosis were excluded from the analysis.
Definitions of delirium and malignancy
Patients were considered delirious if the ICDSC score was ≥ 4. The ICDSC is one of the most widely used screening methods in the ICU setting and comprises eight criteria assessed over one entire nursing shift. Initial validation, meta-analysis and previous studies by our center showed that the chosen ICDSC cut-off score of ≥ 4 has good sensitivity (62%-99%) and specificity (57%-95%) as well as moderate to good reliability (κ 0.59–0.92) [14–17]. However, the different criteria of screening tests do not equally contribute to the test’s diagnostic performance [18, 19].
Patients were categorized as having malignancy when the principal diagnosis had been attributed to an ICD-10 code from the International Classification of Diseases (ICD) which belongs to the block C (“malignant neoplasms”) from chapter II (“neoplasms”).
Outcome variables
Outcome variables of interest were in-hospital mortality, ICU and hospital LOS in hours and days, respectively, duration of mechanical ventilation in hours, ICU nursing workload assessed with the Nine Equivalents Nursing Manpower Use Score (NEMS) [20], costs in the ICU and total costs per case in Swiss Francs, assessed by the hospital and provided to the Swiss Federal Statistical Office, as well as the rate of patients discharged home.
Potential confounders
The Charlson Comorbidity Index was calculated according to Quan et al. [21], with higher values signifying a higher comorbidity burden. The Simplified Acute Physiology Score II (SAPS II) indicating disease severity was computed with the worst values during the first 24 hours of the ICU stay [22]. Sepsis and shock were determined using the ICD-10 codes from principal and secondary diagnoses.
Data sources
All data are documented in the patient medical records. They refer to the Swiss Federal Statistical Office [23] medical and administrative database and the database Minimal Data Set – Intensive Care Unit (MDSi) [24]. Authorized administrative personnel extracted the data of interest and provided it to the investigators. The researchers had no possibility to identify patients from whom data were collected.
Statistical analyses
Characteristics of patients with malignancy were described for the entire population as well as for the subgroups of delirious and non-delirious patients. Values are depicted as numbers and percentages for categorical variables or median and interquartile range (IQR) for continuous variables. Groups were compared with Fisher’s exact test or Mann-Whitney U test depending on the variable. In unadjusted analyses comparing delirious to non-delirious patients, odds ratios (OR) and 95% confidence intervals (CI) were calculated for in-hospital mortality and for the rate of patients discharged home, and hazard ratios (HR) with 95% CI were computed for ICU and hospital LOS as well as duration of mechanical ventilation with univariate Cox regression. Regression coefficients and 95% CI were obtained from linear regression for ICU nursing workload as well as ICU and total costs. In multivariate analysis done with multivariate binary logistic regression, multivariate Cox regression and multiple linear regression, the OR, HR and regression coefficients and their 95% CI were adjusted for the following six covariates: presence of sepsis, presence of shock, emergency admission, age, Charlson Comorbidity Index and disease severity (SAPS II). However, in-hospital mortality was only adjusted for presence of sepsis and shock and SAPS II, because the number of events restricted the inclusion of more covariates. The null hypothesis was rejected with a two-sided p value < 0.05. All statistical analyses were performed with IBM SPSS Statistics, version 25, software (IBM, Armonk, NY, USA).