Comparing indices of responsiveness for the Coma Near‐Coma Scale with and without pain items: An Exploratory study

Abstract Introduction This study aimed to establish the indices of responsiveness for the Coma/Near‐Coma (CNC) scale without (8 items) and with (10 items) pain test stimuli. A secondary purpose was to examine whether the CNC 8 items and 10 items differ when detecting change in neurobehavioral function. Methods We analyzed CNC data from three studies of participants with disorders of consciousness: one observational study and two intervention studies. We generated Rasch person measures using the CNC 8 items and CNC 10 items for each participant at two time points 14 ± 2 days apart using Rasch Measurement Theory. We calculated the distribution‐based minimal clinically important difference (MCID) and minimal detectable change using 95% confidence intervals (MDC95). Results We used the Rasch transformed equal‐interval scale person measures in logits. For the CNC 8 items: Distribution‐based MCID 0.33 SD = 0.41 logits and MDC95 = 1.25 logits. For the CNC 10 items: Distribution‐based MCID 0.33 SD = 0.37 logits and MDC95 = 1.03 logits. Twelve and 13 participants made a change beyond measurement error (MDC95) using the CNC 8‐item and 10‐item scales, respectively. Conclusion Our preliminary evidence supports the clinical and research utility of the CNC 8‐item scale for measuring the responsiveness of neurobehavioral function, and that it demonstrates comparable responsiveness to the CNC 10‐item scale without administering the two pain items. The distribution‐based MCID can be used to evaluate group‐level changes while the MDC95 can support clinical, data‐driven decisions about an individual patient.

Clinicians and researchers rely on standardized assessment data to interpret whether patients in disorders of consciousness are making a change in neurobehavioral function (i.e., recovery of consciousness).
Evidence of psychometric properties for standardized, bedside observational assessments has grown in recent years (Pape et al., 2014;Weaver et al., 2022;Williams & Smith, 2017). A systematic review of assessments for disorders of consciousness identified there is a need for reliable and valid assessments to examine a critical gap in measurement, the assessment's ability to detect change (Seel et al., 2010). Only one neurobehavioral function assessment, the Disorders of Consciousness Scale-25, has quantified meaningful change by reporting indices of responsiveness (Mallinson et al., 2016).
Indices of neurobehavioral responsiveness can include the anchorbased and distribution-based minimal clinically important difference (MCID) and minimal detectable change based on a 95% confidence interval (MDC 95 ). Anchor-based MCIDs map patient change on an assessment to an external criterion that is considered important to family, care partners, patients, or clinicians (Crosby et al., 2003;Guyatt et al., 2002). To the best of our knowledge, there is currently no assessment of neurobehavioral function for which care partners' or clinicians' perceptions of change is the external criterion for an anchor-based MCID. The distribution-based MCID indicates whether the difference between scores from two time points are likely to represent a group-level change that exceeds a clinical standard. The distribution-based MCID represents the smallest amount of change in an outcome that might be considered important and provides an estimate of treatment effectiveness in research. The MDC 95 indicates whether the difference between scores from two time points are likely to represent an individual-level change that exceeds measurement error. Establishing indices of responsiveness, such as the MDC 95 , is important because clinicians must monitor neurobehavioral recovery of consciousness using validated measures to determine whether rehabilitation treatment is effective (Giacino et al., 2020). Furthermore, measuring a decline, that is beyond measurement error, in neurobehavioral recovery may signal the need to determine the presence of a new or previously undetected debilitating condition (e.g., subclinical seizures or hydrocephalus).
Assessments evaluating the neurobehavioral function for patients with disorders of consciousness often include items with painful stimuli (e.g., Coma/Near-Coma Scale, Coma Recovery Scale-Revised, and Disability Rating Scale) (Zasler et al., 2022). Previous work has identified the need to assess nociception in patients with disorders of consciousness (Schnakers et al., 2010). While localization to noxious stimulation is considered indicative of conscious perception (Giacino et al., 2002), painful stimuli has produced brain activation in the primary somatosensory cortex of patients in a vegetative state (Laureys et al., 2002). Thus, response to painful stimuli occurs when individuals are in the vegetative state, minimally conscious state, and emerged. Identification of nociception may support clinical decisions to adjust the individual's positioning or provide pain medication to increase the comfort of the individual but may not clearly distinguish different levels of consciousness. Prior research on the Coma/Near-Coma (CNC) scale found that responses to painful stimuli and neurobehavioral function are two dis-tinct concepts (Weaver et al., 2020). This earlier work indicates the CNC scale is a unidimensional assessment of neurobehavioral function only when painful test stimuli/items are not included. This finding suggests that when responses to painful stimuli change over time, the changes do not necessarily reflect changes in neurobehavioral function. To advance our understanding of this finding, the purpose of this study is to (i) compute the indices of responsiveness for the CNC scale without (8 items) and with (10 items) painful test stimuli, and (ii) determine whether the CNC 8 items and CNC 10 items differ according to detection of change in neurobehavioral function.

Participants
This retrospective cohort study includes 40 adults with severe brain injury. Data were included from three studies: (1)  (1) ≥ 18 years, (2) diagnosed with disorders of consciousness from brain injury, and (3) had two CNC scale assessments 14 ± 2 days apart. Participants were excluded from the study if brain injury was due to cancer, tumor, or encephalopathy.

Coma/Near-Coma Scale
The CNC scale is a short assessment (11 items) originally designed to capture neurobehavioral responses in persons in lower states of disorders of consciousness (Rappaport, 2005(Rappaport, , 1992. The assessment scoring form and training recommendations can be found on the Traumatic Brain Injury Model Systems website (Rappaport, 2000). The olfactory item was not administered in any of the three studies due to difficulty in controlling for consistency and shipping restrictions for ammonia; thus, the scale was administered with 10 items (Bender Pape et al., 2015;Pape, 2015;Pape et al., 2014;Zilliox et al., 2022). Each item is scored using a 3-point rating scale with response options of 0, 2, 4; a lower score indicates better neurobehavioral function (Rappaport, 2000). For example, for the Auditory item a "0" reflects that the patient responds to the bell ringing stimulus with behaviors such as eye TA B L E 1 Total raw score to Rasch person measure conversion for the 8-item and 10-item Coma Near Coma Scale from least to most neurobehavioral function. opening or orientation toward sound and that this occurs ≥ 3 times.
The study reported here compares the CNC scale comprised of the 10 items (excluding the olfactory item) and the CNC scale comprised of 8 items (excludes two pain items) (Weaver et al., 2020). For our analyses, the 3-point rating scale was rescored to 2, 1, 0, so that a

Data analyses
For the CNC 8 items and CNC 10 items, we confirmed the Wright's Person Separation Reliability using Winsteps software version 5.3.3.1, because our study includes eight participants that were not in the previous CNC scale psychometric analyses (Weaver et al., 2021).
We hypothesized the person separation reliability coefficients would remain the same because a strength of Rasch Measurement Theory is that the analyses are sample-free and item-free (Bond et al., 2020). We generated Rasch person measures for each participant at both timepoints for the CNC 8 items and CNC 10 items and used these data for responsiveness analyses.
We calculated the pooled standard deviation (SD pooled ), effect size, and the standardized response mean using MedCalc version 17.6. We interpreted the effect size and standardized response mean at 0.20, 0.50, and 0.80 as a small, moderate, and large responsiveness (Husted et al., 2000). The standard error of measurement (SEM) was calculated using the following formula SEM = SD pooled (

Indices of reliability and responsiveness
The CNC 8-item ordinal total raw scores ranged from 0 to 32 and were transformed to equal-interval Rasch person measures, ranging from   (Table 3). The small (0.20 SD), minimally important (0.33 SD), and medium (0.50 SD) distribution-based MCIDs are almost equivalent for the Coma Near-Coma-8 item and CNC 10-item versions ( Table 3).
The CNC 8 items has consistently larger distribution-based MCIDs than the CNC 10 items (Table 3). The mean difference between the baseline and follow-up scores after receiving an intervention (Table 2) was 0.37 and 0.35 for the CNC 8 items and CNC 10 items, respectively, which exceeds the respective small (0.20 SD) distribution-based MCIDs (  Notation: Data calculated using Rasch person measures (logits), a higher number indicates more neurobehavioral function.

TA B L E 3
Minimal detectable change and distribution-based minimal clinically important differences for the Coma Near-Coma Scale with and without painful test stimuli. ticipants and the CNC 10-item version identified 13 (33%) participants as making a change beyond measurement error or a "true" change. Of these patients making "true" change, the CNC 8-item version identified 9/12 (75%) participants as improved and 3 as declined (25%); the CNC 10-item version identified 10/13 (77%) participants as improved and 3

DISCUSSION
The preliminary evidence reported here, addresses the identified need for additional research on the measurement properties of the CNC scale, including responsiveness (Seel et al., 2010). The reported findings suggest that the CNC 8 items (without painful test stimuli) and CNC 10 items (with painful stimuli) have comparable reliability and precision for detecting change in neurobehavioral function.
As postulated, the Wright's Person Separation Reliability values did not change with the addition of 8 participants. The finding that the shorter CNC 8 items has comparable precision for detecting change to the longer CNC 10-item supports previous findings suggesting that pain is a distinct concept from neurobehavioral function (Weaver et al., 2020 (Altman & Bland, 2005

Study limitations
The indices of responsiveness we report should be interpreted cautiously because they were generated from a small sample of adults with disorders of consciousness. Future studies should substantiate our findings using larger samples. For patients with disorders of consciousness, establishing an anchor-based MCID reflecting family, care partner, or clinician perception of change is an important area of future inquiry.

CONCLUSION
Our preliminary evidence supports the clinical and research utility of the CNC 8 items, as it demonstrates adequate responsiveness to neurobehavioral change without administering the two pain items.
When evaluating change at the group level the CNC 8-item effect size and distribution-based MCIDs should be used. When evaluating change for an individual patient, the CNC 8-item MDC 95 should be used to determine when observed change is beyond measurement error.

DATA AVAILABILITY STATEMENT
Deidentified data from this study are available upon request.