Diagnosis of Delirium and Assessment of Sleep Quality in Patients in the Anesthesia and Intensive Care Unit

Delirium is a condition involving functional losses in the brain due to both structural and vascular abnormalities caused dehydration and hypoperfusion, which prevent the transport of oxygen and glucose to neurons, thereby leading to metabolic dysfunction in the brain. Moreover, delirium, which is characterized by disorientation, inattention, and changes in thoughts and/or behaviors, is a serious and widespread neuropsychiatric condition that frequently goes unnoticed, although it is preventable and reversible following appropriate intervention. Numerous risk factors have been found to affect the development of delirium, including prolonged hospital stays, medication use, coping mechanisms during illness, personality traits, and treatment compliance. It is believed that intensive care patients face a higher risk of developing delirium. In this regard, invasive procedures, regular nursing care, environmental noise, continuous light exposure, pre-existing sleep disorders, pain, anxiety, fear, and loneliness represent physiological and psychological factors that can lead to sleep deprivation in intensive care patients. Indeed, places such as intensive care units (ICUs), where patients are exposed to constant stimuli and may experience severe pain, can trigger stress responses that lead to agitation, sleep disorders, and delirium.

Previous studies have indicated that medications used to control delirium can negatively affect sleep quality, while worsening sleep quality can exacerbate delirium. The main goal in preventing delirium is to ensure a regular sleep-wake cycle, support adequate fluid and nutrient intake, and provide both preventive and therapeutic treatments through a systematic, multidisciplinary team-based approach to reduce the frequency and duration of delirium. Artificial intelligence and machine learning approaches enable machines to simulate complex processes such as thinking and consciousness, thereby allowing computers to learn the relationships between inputs and outputs from large datasets, which helps them to make optimal decisions, analyses, and predictions. In the healthcare field, such technology is used in many areas, including the early detection and prediction of disease, disease classification, treatment, prevention of adverse outcomes, rapid analysis of clinical data, cost-saving initiatives, provision of effective and quality care, minimization of human errors, and clinical decision-making .

In the present study, the relationship between delirium and sleep quality in patients in the anesthesia and ICU was examined and a predictive analysis using a machine learning approach was performed.

The data required for this study were collected using the Patient Information Form, Nursing Delirium Screening Scale (Nu-DESC), Richards-Campbell Sleep Scale (RCSS), Richmond Agitation-Sedation Scale (RASS), and GCS. The collected data were analyzed using the SPSS statistical program (SPSS-26), and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were followed when reporting the results. The sample size was determined using the known population sample calculation method. Both frequency and percentage analyses were used to describe the distribution of the participants' demographic characteristics, while the mean and standard deviation were used to determine the participants' levels on the utilized scales. Cronbach's alpha reliability analysis was conducted to assess the scale reliability, and the post hoc power analysis showed that the study had medium power at a 95% confidence level with 99% significance . The skewness-kurtosis values were checked to assess the normality of the distribution. An independent samples t-test was used to assess the differences between groups. For comparisons among more than two groups, a one-way analysis of variance (ANOVA) was applied, while the prediction and receiver operating characteristic (ROC) curve analyses were conducted using R version 4.1.3.