The Observational Scale of Level of Arousal: A brief tool for assessing and monitoring level of arousal in patients with delirium outside the ICU

Altered level of arousal, encompassing drowsiness and hypervigilance, affects at least 10% of acutely unwell patients. Existing scales provide limited coverage of milder changes in level of arousal. We devised the Observational Scale of Level of Arousal (OSLA) to enable more detailed arousal assessment. Here, we provide a preliminary case‐control study of performance of the OSLA in assessing abnormal level of arousal associated with delirium outside the ICU.


| INTRODUCTION
Delirium is a severe, acute neurocognitive disorder characterised by disturbances in attention, level of arousal and other mental functions.
It affects at least one in eight hospitalised older patients and is independently associated with multiple adverse outcomes. [1][2][3] Alterations in level of arousal are common in delirium, with many patients showing hypo-or hyperarousal. 4 The hypoactive subtype, characterised by drowsiness or somnolence, is the most common form of delirium. 5 The arousal component of delirium has been described variably in standard diagnostic criteria (Table 1). In DSM-5, severely reduced level of arousal precluding cognitive testing or interview but above the level of coma is considered to indicate severe inattention. Assessment of level of arousal is therefore a core part of the evaluation of the features of delirium, and arousal measurements appear to be useful in clinical practice as a strong indicator of delirium. 4,6,7 Outside the field of delirium, abnormal level of arousal is increasingly seen as a crucial marker of illness severity and predictor of mortality in hospitalised patients. 8 In United Kingdom hospitals, level of arousal is routinely assessed using the AVPU scale (A, alert; V, responds to voice; P, responds to pain; U, unresponsive) as one of six indicators as a National Early Warning Score. 9 Yet level of arousal in the specific context of delirium remains relatively understudied compared to its key cognitive symptom of attention deficits and other features. [10][11][12][13] In delirium, level of arousal is often reduced but there is a wide range of severities, from mild drowsiness to only being able to produce a basic motor response to a verbal stimulus. Conversely, patients may have heightened arousal and appear agitated and hyperalert. The Richmond Agitation-Sedation Scale (RASS), 14 which was originally developed to assess agitation or sedation levels in Intensive Care Unit (ICU) patients, has recently been modified for use as a delirium screen by including assessment of attention (mRASS). 7 The RASS is the most studied arousal scale in delirium. 4,15 However, a RASS score of +1 or −1 does not provide detailed information on the degree to which level of arousal is abnormal. More generally, an overall lack of granularity and operationalisation to capture this important feature of delirium in both arousal-specific and general delirium scales suggests that there would be value in having an instrument that provides a more detailed assessment.
To address this, we developed a new scale entitled the Observational Scale of Level of Arousal (OSLA; Table 2). It was designed for research use, to characterise the abnormalities of level of arousal associated with delirium, complementing assessments of attention and other features of delirium. Individual item scores characterise the profile of abnormalities while the single overall score provides an index of severity. The OSLA was used in a small study examining the relationship between altered arousal and inattention. 4 OSLA scores were strongly associated with delirium diagnosis. Another study reported good diagnostic accuracy of the

Key Points
• Assessment of level of arousal is a core part of the evaluation of delirium, and arousal measurements could be useful in clinical practice as a strong indicator of delirium. OSLA for detecting delirium on its own, and in combination with an attention task. 16 Here, we provide preliminary evaluation of the OSLA as a brief instrument to identify abnormal levels of arousal associated with delirium in patients with acute hip fracture. First, we assessed the psychometric characteristics of the OSLA, using exploratory factor analysis.
Second, we assessed the diagnostic performance of the OSLA for delirium detection, because acute onset altered level of arousal is considered a strong indicator of delirium. 4 Third, to explore the potential utility of the OSLA in detecting changes in delirium longitudinally, we assessed its ability to detect within-person fluctuations in delirium status and symptom severity over several test occasions.

| Design
This was a secondary analysis of data from a prospective cohort study in older adults with acute hip fracture with and without delirium. 17 Participants were assessed in the 24 hours prior to their surgery, repeatedly up to 14 days post-operatively, and at 3, 6, and 12 months post-operatively. Data up to day 14 are reported here. The study was approved by the Scotland A Research Ethics Committee and written consent from patients or legal proxies was obtained.

| Participants
A total of 108 community dwelling patients were recruited from orthopaedic wards at the Royal Infirmary of Edinburgh, Scotland.
Patients were eligible if they were aged over 60 years and had an acute hip fracture and spinal anaesthesia. Patients were not eligible if they were nursing home residents; had taken oral or inhaled steroids in the last 10 weeks; had significant Parkinson's disease or other comorbid diseases with a prognosis of less than 1 year; or had major communication difficulties such as aphasia or where English was not their first language.

| Measurements and procedures
The diagnosis of delirium was made by a geriatrician (RJH), aided by the use of the Confusion Assessment Method (CAM) 18 and Delirium Rating Scale-Revised-98(DRS-R98), 19 and supplemented with assessments of level of consciousness (Richmond Agitation-Sedation Scale (RASS) 14  The OSLA comprises four items, each assessing a different feature of arousal: eye opening, eye contact, posture, and movement ( Eye contact Score Description 0 Spontaneously makes and holds eye contact appropriately 1 Drowsy and makes eye contact to command but cannot hold it for very long 1 Alert but eyes wandering, some appropriate eye contact 2 Alert but eyes wandering, little or no appropriate eye contact 2 Drowsy but makes brief eye contact 3 Eyes will/are open but no eye contact Posture (NB take into account weakness due to stroke or neurological disease, etc.) Score Description 0 Sitting out in chair or up in bed, holding appropriate posture 1 Slumped in chair or bed but attempts to sit upright and sustain posture on request 2 Slumped in chair or bed and unable to sustain posture 3 Lying in bed and unable or no response to request to sustain posture Movement Score Description 0 Moves spontaneously and purposefully with no restless or agitated movements 1 Occasional or mild restless or fidgety movements, no aggressive or vigorous movements 1 Reduced frequency of movement, mildly slowed up 2 Frequent restless or fidgety movements, no aggressive or vigorous movements 2 Moderately reduced frequency and speed of movement, interfering with assessment or self-care 3 Aggressive or vigorous, recent pulling out of lines 4 Overtly combative, violent 4 Severely reduced frequency and speed of movement, few spontaneous movements Score (0-15) make regarding level of arousal in practice. The second stage involved editing the list of potential items to construct a scale which would be:

| Statistical analysis
Analysis was carried out using R version 3.0.1. 24 Cases were excluded pairwise where data were missing. A threshold of P < .05 was taken to denote statistical significance.
Demographics and test scores for each group are presented as medians (interquartile range [IQR]) unless otherwise specified. Comparisons of OSLA scores and all other data were made between groups with and without delirium using Mann-Whitney U tests with continuity correction, separately for each assessment and also pooled assessments. Estimates of effect size r were calculated by dividing zscores by the square-root of n. 25 Pearson's chi-squared tests were used for categorical data where appropriate. Correlations were calculated using Kendall's Tau due to frequent ties in the data. Holm corrections were applied to multiple comparisons.
We assessed the suitability of the data for exploratory factor analysis (EFA) using Bartlett's test of sphericity, the determinant of the correlation matrix and the Kaiser-Meyer-Olkin measure of sampling adequacy. Horn's parallel analysis (with 10 000 iterations) was used to empirically determine the number of factors to retain. Factors with eigenvalues greater >1 were assumed to be meaningful. We then conducted minimum residuals EFA without rotation-as parallel analysis suggested only a single factor-and used factor loadings of 0.40 or greater in the factor designation.
OSLA scores were compared between groups with and without delirium at each assessment. A receiver operating characteristic (ROC) analysis was conducted on OSLA scores with delirium diagnosis as a reference to assess the ability of the OSLA to detect delirium, for data collapsed across assessments and also separately for each assessment. The relationship between scores on the OSLA and the DRS-R98 severity scale was examined using Kendal-Tau correlations.
Exploratory linear mixed effects models were used to evaluate responsiveness of the OSLA to change in delirium status and severity over time, to provide additional information on its performance and also to explore the importance of level of arousal in contributing to a delirium diagnosis. These analyses fitted a within-person fluctuation model to test if changes in OSLA over time (of the order of days) were predicted by changes in each of the two time-varying predictors. Two models were fitted, one with delirium diagnosis as time-varying predictor (mean-centred) and all covariates (model 1), and one with delirium severity as time-varying predictor and all covariates (model 2) (R function lmer 24 ). The models included the following time-invariant covariates: age, sex, IQCODE, CCI, and APACHE II score. The dependent variable was OSLA score. accounted for 31% of the variance using minimum residuals factor analysis (Table S2).

| Association between OSLA and delirium status and severity
Scores on the OSLA were consistently higher in patients with delirium than those without delirium at each assessment, and also when data were collapsed across assessments (Table 4). A delirium diagnosis was always accompanied by an OSLA score greater than 0.
Further, scores on the OSLA correlated significantly with scores on the DRS-R98 severity sub-scale at each assessment point and when data were collapsed across assessments (

| DISCUSSION
The present study provides preliminary support for the utility of the OSLA as a brief, accurate instrument for measuring level of arousal in delirium. The OSLA showed gradations in scores and was responsive to a change in delirium diagnosis within patients over time. Factor analysis of the eye opening, eye contact, movement, and posture items confirmed that it is appropriate to group these four features under a single factor (ie, arousal).
A wide range of OSLA scores was seen in the present study (0-9) and a delirium diagnosis was always accompanied by an OSLA score greater than 0. The OSLA therefore appears to be sensitive to the gra- Interestingly, the OSLA in our study performed broadly comparably while assessing level of arousal alone, without any items specifically assessing attention. Of note, the present findings suggest a lower cut-

CONFLICT OF INTEREST
None declared.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available onrequest from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.