Introduction

Sleep is vital for physical and mental health. Nowadays, more attention is paid to sleep disturbances in intensive care unit (ICU) patients, as they may contribute to the development of delirium. Studies have shown numerous similarities in the clinical and physiological profiles of patients with delirium and sleep disturbances (1). A study of 29 ICU patients found an association between delirium and severe sleep reduction (2). There is an electrophysiological relationship between sleep architecture changes and delirium, with delirium occurring in patients with rapid eye movement sleep loss and those with clinically confirmed atypical sleep, characterised by electroencephalography findings suggesting wakefulness (2-4). A meta-analysis confirmed that preexisting sleep disturbances are likely associated with higher rates of postoperative delirium [odds ratio (OR): 5.24; 95% confidence interval (CI): 3.61-7.60; p<0.001] (5). Even though the link between sleep disturbances and delirium was studied and analysed by many authors (1, 6, 7), the available literature suggests that there may be a close relationship between delirium, sleep, circadian rhythm, and critical illness. However, no causal pathway has yet been clearly described, and the directionality of the relationship is not understood. The attempts to reduce the incidence of delirium are based on identifying and modifying risk factors. Sleep disturbances are thus one of the potentially modifiable risk factors. Professionals’ increasing interest in the recent Society of Critical Care Medicine (SCCM) guidelines on sedation and delirium is reflected in the recognition that professionals’ increasing interest in the recent SCCM guidelines on sedation and delirium, therefore, the sleep promotion strategy is a fundamental and integral part of delirium prevention and management (8).

Therefore, the present study aimed to investigate the relationship between subjective sleep quality assessed with the numeric rating score (NRS) and the presence of delirium identified with both the confusion assessment method for the intensive care unit (CAM-ICU) and the intensive care delirium screening checklist (ICDSC). The second endpoint was to analyse the effect of other selected predictors on the occurrence of delirium.

Materials and Methods

Design

A Prospective Observational Study

Patients: Data for the study were collected in the Department of Anaesthesiology and Intensive Care Medicine ICU (5 beds) and multidisciplinary ICU (10 beds) of AGEL Hospital between February 2020 and August 2020. Adult conscious patients who consented to participate and were staying in the ICU for more than 24 hours were included in the study. The following demographic data were collected: Age, sex, smoking status, and alcohol consumption. The following were recorded from the clinical data: operation, length of stay in ICU, overall mortality, type of admission, pain visual analogue scale (VAS), sedation richmond agitation-sedation scale (RASS), therapeutic intervention scoring system (TISS) score, history of mechanical ventilation, restraints, and medication (opioids, benzodiazepines, antipsychotics). The exclusion criteria were a terminal illness; a diagnosis of dementia; and an altered consciousness: Glasgow coma scale (GCS) score ≤12 or deep sedation (RASS score ≤ -4).

Assessment instruments: Two instruments for diagnosing delirium were used in the study. The ICDSC includes the following eight items: Altered level of consciousness, inattention, disorientation, hallucination-delusion, agitation or retardation, inappropriate speech or mood, sleep-wake cycle disturbance and symptom fluctuation. Each positive item scores one point. If the total score is ≥4, delirium is diagnosed. Scores of 1-3 indicate subsyndromal delirium (9).

When using the CAM-ICU to diagnose delirium, the first step is to assess the level of sedation with the RASS (in deeply sedated patients, not responding to stimulation, RASS score ≤-4, the presence of delirium cannot be established). The second step is an assessment of four key features of delirium: Acute change or fluctuating course of mental status (Feature 1), inattention (Feature 2), altered level of consciousness (Feature 3), and disorganised thinking (Feature 4). Delirium is considered positive when Feature 1 and Feature 2, and either Feature 3 or Feature 4 are present. If not, delirium is excluded (CAM-ICU negative). RASS scores ranging from 0 to -3 are associated with hypoactive delirium. A RASS score of +1 to +4 suggests hyperactive delirium. Mixed delirium occurs when a patient fluctuates between the two forms of delirium (10).

Sleep quality was assessed with the NRS. Patients used this 10-point analogue scale to rate their subjective quality of sleep. All assessments were performed in the morning, between 8.00 AM and noon. Nurses asked patients the following question: Could you rank your sleep of last night on a scale between 0 (a worst night’s sleep) and 10 (a best night’s sleep)?

Good vs. bad sleep definition: In the study, patients’ sleep was classified as either good (NRS >5) or bad (NRS ≤5), and the sample was divided accordingly. The cut-off was arbitrarily determined based on literature data (16) showing good statistical results, namely a sensitivity of 83%, a specificity of 79%, an area under the receiver operating characteristic curve of 0.81 (95% CI: 0.74-0.87).

Process of translation: The instrument was translated and linguistically validated according to the guidelines and standards, for the translation and cultural adaptation of patient-reported outcome measures (11).

Data Collection

Two assessment instruments (CAM-ICU and ICDSC) were used to detect delirium. Sleep quality was subjectively evaluated with the NRS. Nurses performed delirium screening twice a day, and sleep quality was assessed once a day. On average, the forms took approximately 5 minutes to complete. In total (126 patients), 1299 paired questionnaires and 278 NRS records were obtained.

Ethical Aspects

The study, conducted in accordance with the Declaration of Helsinki, was approved by the Ethics Committee of Vzdělávací a výzkumný institut AGEL (no: INT 2019003, date: 08.12.2019). Respondents’ participation was voluntary and anonymous. The author approved using the Czech version of the CAM-ICU. The ICDSC was translated with the author’s permission. The NRS was used as published by Rood et al. (12).

Statistical Analysis

Relationships between pairs of metrics, ordinal or binary variables, were tested using Kendall’s t coefficient. The relationships between a set of explanatory variables-differentiators and predictors-on one side, and the predicted (explained, dependent) binary or metric variables on the other, were evaluated by multivariate regression with a reduction of dimensionality known as optimized potentials for liquid simulations (OPLS). This test can cope with the problem of severe multicollinearity (high intercorrelations) in the matrix of explanatory variables, while ordinary multiple regression fails to evaluate such data correctly. The multicollinearity in OPLS is favourable, as it enhances the predictive power of the model. In the OPLS models with binary predicted variables, the logarithm of the ratio of the probability of positive outcome to the probability of adverse outcome (logarithm of the likelihood ratio) was chosen as a single dependent variable, ensuring that the predicted probability ranged between 0 and 1. The statistical software SIMCA-P v.12.0 from Umetrics AB (Umeå, Sweden), which was used for OPLS analysis, enabled the identification of the number of relevant components, the detection of multivariate non-homogeneities, and the testing of multivariate normal distribution and homoscedasticity (constant variance).

Results

The study comprised 126 consecutively admitted patients (76 males/50 females; 60.3/39.7%) with a median age of 71 (60, 77). Twenty-seven patients (21.4%) had a positive history of mechanical ventilation, and 38 respondents (30.2%) underwent surgery. Acute admissions prevailed (81%). The admission diagnoses varied, with the most frequent being the following international classification of diseases (ICD) categories (in descending order): Diseases of the respiratory system (ICD J) 17.5%, diseases of the circulatory (ICD I) and digestive (ICD K) systems 16.7% each. 18.3% of admissions were classified as abnormal clinical findings (ICD R), including frequent ICU syndromes (shock, hypovolemia, sepsis, etc.) without further specification. The most frequently administered drugs related to analgesia, sedation and delirium treatment were opioids (53 patients; 42.1%), antipsychotics (38 patients, 30.2%) and benzodiazepines (27 patients, 21.4%). The median length of stay in the ICU and hospital was six days (from 4 to 9) and 15.5 days (from 9 to 20), respectively. During their stay in the ICU, ten patients (7.9%) died. The number of deaths throughout the entire hospital stay until discharge (including ICU deaths) was 18 (14.3%). The median TISS score measuring nursing workload was 557, suggesting that the sample primarily included conscious patients who were not critically ill.

From the 126 patients, 1299 paired records assessing delirium and 278 records evaluating subjective sleep quality were obtained. According to CAM-ICU assessment, 37 patients were classified as delirium-positive (326 records; 29.4%) and 89 delirium-negative (973 records; 70.6%). Combining delirium-positivity with RASS, 18 patients showed hyperactive delirium (total of 152 records, 14.3%), 12 hypoactive delirium (94 records; 9.5%) and seven mixed forms (80 records, 5.6%). According to ICDSC, delirium (a score of 4-8) was diagnosed in 40 patients (total of 346 records; 31.7%), subsyndromal delirium (a score of 1-3) in 32 patients (381 records; 25.4%) and 54 patients (572 records; 42.9%) were delirium-negative. Thirty-three patients (total of 75 records; 26.2%) reported good sleep (NRS >5), and 93 patients (203 records; 73.8%) had lousy sleep (NRS ≤5). Based on this rating, the studied population was divided into two subgroups. (Table 1).

Kendall’s t values (using 95% CI), which were used to express the power of the relationships, were interpreted as follows: higher values indicated stronger relationships. In contrast, positive or negative values indicated direct or indirect causality (13). Almost all the following parameters were shown to be statistically significant regarding sleep disturbance (p<0.001), excluding alcohol, age, RASS, gender, operation, type of admission, some diagnoses, and hospital mortality. The results obtained (ranked by the absolute strength of the first three in the relationship and given with CI) were GCS -0.383 (-0.413 - -0.352), physical restraints 0.243 (0.209-0.276), VAS 0.196 (0.161-0.23) (Tables 2,3). 

The association between poor sleep quality (bad sleep, NRS ≤5) and delirium assessment (CAM-ICU, ICDSC) scores was studied. The results showed a significant relationship (p<0.001) between sleep disturbances and delirium assessment methods. Kendall’s t was 0.391 (CI: 0.36-0.421) for CAM-ICU positivity and 0.463 (0.435-0.491) for ICDSC positivity, respectively. An important point was that these positive associations (delirium positivity and bad sleep) were rated moderate (Table 4) (13).

Advanced statistics were used to select a set of predictors (risk factors) evaluated in the OPLS model to assess variances in the presence of delirium (for each diagnostic tool). In the OPLS model for multivariate regression, the risk factor with the highest statistical confidence for the CAM-ICU positivity and ICDSC positivity was the first three predictors (according to component loading): (1) GCS followed by (2) physical restraints and (3) VAS. The association of these three predictors were assessed as moderate to strong (14), and prediction is recommended. The rest of the variables and the degree of influence of the monitored variables were evaluated as weak, and thus, they are not suitable for predicting disorders (Tables 5,6).

Discussion

In this study, we have identified critical findings. Firstly, although screening questionnaires can help diagnose delirium quickly (within 2 to 5 minutes), different questionnaires may detect delirium in varying ways. Unfortunately, the patient’s ability to answer the questionnaire is limited in the ICU environment. Secondly, we found that patients who reported poor sleep quality had a higher incidence of delirium: 93 (73.8%) compared to 33 (26.1%). While several validated methods exist for screening, monitoring, and diagnosing sleep in the ICU, each technique has limitations and cannot be used for all patients. This is also one of the reasons why the effects of poor sleep quality and delirium development on patient outcomes are not immediately apparent. Finally, to prevent the growth of delirium, predicting its occurrence based on various indicators is trending; however, many of these indicators are not modifiable (e.g., age, TISS, gender).

The incidence of delirium varies considerably depending on the population of patients examined and diagnostic methods. Delirium has been reported in 16-89% of ICU patients, and its incidence appears to be highest (up to 80%) in mechanically ventilated patients (14, 15). Our reported incidence (29.4% when assessed with the CAM-ICU and 31.7% with ICDSC, respectively) lies within the lower part of the range, which could be explained by patients’ characteristics (the majority were not very sick and were not mechanically ventilated). Delirium includes three motor subtypes-hyperactive, hypoactive, and mixed-which may be associated with different prognoses. In the present study, 14.3% of cases were hyperactive, 9.5% hypoactive, and 5.6% mixed. A meta-analysis of 18 studies showed different incidences: Hypoactive (11%), followed by mixed (7%) and hyperactive (4%) (16). Another methodological pitfall of assessing delirium with certain diagnostic instruments is influenced by sedative drugs, which may affect the results, potentially leading to overrated positivity in cases where the RASS is not 0. A possible solution is to assess consciousness only after pharmacological sedation wears off. Therefore, to assess the persistence of delirium, many ICUs use routine daily sedation disruptions, (spontaneous awakening trials) as a part of standardised protocols for the need for further sedation (8). The ICDSC diagnosed subsyndromal delirium (10) in 25.4% of cases. Subsyndromal delirium could be viewed as a pre-delirium-a transition between delirium and normal mental status. It is common in ICU patients, but its true incidence and effect on the outcomes of critically ill patients remain unclear. In a meta-analysis of 6 studies, subsyndromal delirium was found in one-third of critically ill patients, with a limited impact on their outcomes (17). One of the study’s primary goals was to assess the impact of sleep disturbances (for our purposes, classified subjectively as bad sleep, NRS ≤5) and their association with studied parameters. The study presumes that sleep disturbances may be a risk factor for delirium and prolonged mechanical ventilation, independently associated with other parameters (ICU deaths, ICU length of stay, and hospital length of stay). Our findings are consistent with these hypotheses and are similar to data reported by other authors (18, 19). Even though our results are based on subjective assessments, which is a substantial limitation, the relationship between delirium and sleep disorders has been confirmed. On the other hand, contrary data exist. The study by Kamdar et al. (20) has shown no difference between subjectively perceived sleep quality assessed with the Richards-Campbell sleep questionnaire (RCSQ) in patients with and without delirium (mean RCSQ 57 vs 58) and there is no relation between perceived sleep quality and transition to delirium (adjusted OR: 1; 95% CI: 0.99-1.00). Interventional studies, however, suggest the opposite. According to Patel et al. (21), the sleep efficiency index has the potential to predict the development of delirium, with patients reporting high sleep efficiency index scores demonstrating a reduced risk of delirium (OR: 0.9; 95% CI: 0.84-0.97). Similarly, Van Rompey et al. (22) revealed, using Cox regression, that earplugs lowered the risk of delirium or mild confusion in the ICU by 53% (hazard ratio: 0.47; 95% CI: 0.27-0.82), with more patients reporting better subjectively assessed sleep quality. Previous studies that have addressed the problem are far from providing unambiguous results.

Another issue regarding the sleep-delirium study is the selection of adequate assessment instruments. Many authors have mentioned problems finding suitable techniques for assessing delirium and detecting sleep disorders simultaneously. It seems reasonable to combine an objective instrument with a subjective assessment (23). A possible approach (suitable mainly for non-ICU patients) is an objective assessment of sleep by actigraphy in combination with another subjective method, a monitoring technique based on alterations in motor activity (23). In ICU patients with altered consciousness, such as those with lower GCS or under sedation, polysomnography, together with a validated subjective questionnaire filled out by nurses, is considered the gold standard (24).

According to reported results, patients with perceived poor sleep quality more often received a sedative medications (benzodiazepines, opioids, and antipsychotics). Thus, the optimal approach to analgesia and sedation in ICU patients seems to be an important consideration. Good clinical practice is well-established, involving using drugs with short half-lives, implementing nurse-driven sedation protocols, including daily awakening trials, limiting deep sedation, minimising the use of muscle relaxants, and monitoring the depth of sedation if necessary (12). Maintenance of normal circadian rhythm, promotion of physiological (good quality) sleep, and prevention of sleep deprivation and disorders are crucial parts of ICU nursing care and are closely related to sedation strategy, affecting numerous clinical outcome parameters, including delirium incidence. Recently, the main principle of delirium management has been shifting from treatment to prevention, requiring knowledge of the associated risk factors. According to Ely et al. (25), patients staying in the ICU have ten or more risk factors for delirium onset. A meta-analysis by Zaal et al. (26) identified 11 risk factors for delirium supported by solid or moderate levels of evidence. Similarly, Van Rompaey et al. (27) grouped the most important risk factors into four domains, with 13 risk factors being identified as significant. Our findings agree with the previously mentioned studies, and add more statistical significance to relationships between delirium and its predictors by applying an OPLS model with consistent results. All the findings above related to sleep and delirium are generalisable and applicable to everyday clinical practice in the form of the ABCDE bundle of proper analgesia, sedation, and delirium management. It has been shown that such a bundle of care, including appropriate pain management, light sedation, avoidance of benzodiazepines, early awakening and weaning from mechanical ventilation, routine delirium monitoring and early mobilisation, improves patient outcomes and decreases delirium incidence by one-third (14).

Study Limitations

The study’s primary limitations are the size of the sample, the number of patients, including its unicentric design and the selection of subjective sleep quality instruments. For a complex and comprehensive evaluation, valid, consistent, and objective methods for sleep measurement, such as actigraphy and polysomnography must be combined with subjective assessment instruments that are completed by patients or nurses. The high-quality multicenter randomised trial could overcome these limitations and increase knowledge of the relationship between sleep disturbances and delirium in ICU patients.

Conclusion

Even though the relationship between sleep disturbances and delirium has not been fully elucidated, many authors assume a bidirectional causal relationship, suggesting that sleep disorders are a risk factor for the development of delirium. The results of the presented study are consistent with this hypothesis. Early detection of delirium is fundamental, and choosing appropriate diagnostic tools remains a concern. Modern trends in intensive care reflect this two-way relation between sleep and delirium by respecting sleep-promoting (primarily non-pharmacological) strategies, delirium prevention, and early therapy as the standard of nursing care. More detailed analysis of this sleep-delirium association is needed for better and more personalised care in the future, minimising the incidence of delirium and need for sedation while maximising ICU patients’ sleep quality.