The somatosensory system processes sensory information received from the body and plays an important role in typical movements (Gaubert & Mockett, 2000). The prevalence of somatosensory impairments is high in people with stroke (Carey, 1995; Yekutiel, 2000). Somatosensory impairments are associated with disrupted motor learning (Scalha, Miyasaki, Lima, & Borges, 2011; Vidoni & Boyd, 2009; Welmer, Holmqvist, & Sommerfeld, 2008), influencing the abilities of the patient to perform everyday activities, such as personal care, and leading to a poor rehabilitation outcome (Carey, 1995; Kalra & Crome, 1993; Patel, Duncan, Lai, & Studenski, 2000; Sommerfeld & von Arbin, 2004; Tyson, Hanley, Chillala, Selley, & Tallis, 2008). Somatosensory capacity is thought to be a precursor to the recovery of movement and functional recovery and is possibly a predictor of rehabilitation outcomes (Peurala, Pitkänen, Sivenius, & Tarkka, 2002; Smania, Montagnana, Faccioli, Fiaschi, & Aglioti, 2003). Assessing somatosensory impairments after stroke is essential in helping to plan an effective intervention and in evaluating the effect of rehabilitation (Bentzel, 2002). Clinical administration of sensory evaluation in people after stroke should be a common and routine task.

Sensation may be classified into three modalities: superficial sensation, proprioceptive sensation, and cortical sensory functions (Bigley, 1990). An evaluation of all three modalities has been suggested in people with stroke for identification of somatosensory impairments (Bigley, 1990). Several clinical assessments of somatosensory impairments are available (Connell, Lincoln, & Radford, 2008; Connell & Tyson, 2012), such as the Semmes-Weinstein monofilament test (Bell-Krotoski, Fess, Figarola, & Hiltz, 1995), the Moving and Sustained Touch-Pressure test (Dannenbaum, Michaelsen, Desrosiers, & Levin, 2002), the Touch Perception Threshold test (Eek & Engardt, 2003), and the Sensory subscale of the Fugl-Meyer Assessment (FMA–S; Lin, Hsueh, Sheu, & Hsieh, 2004). These measurements report the presence of specific somatosensory impairments but cannot identify the comprehensive modalities of somatosensory impairments (Connell & Tyson, 2012; Dannenbaum et al., 2002; Winward, Halligan, & Wade, 2002). Moreover, the recovery of one somatosensory modality may be independent of that of other somatosensory modalities (Winward, Halligan, & Wade, 2007). Using a comprehensive assessment may provide more detailed information about various sensory modalities during rehabilitation and be more sensitive to treatment efficacy than using the measures of a specific modality (Endicott, Spitzer, Fleiss, & Cohen, 1976).

To date, no comprehensive assessment scales of somatosensory impairments with established reliability and validity have been identified except the Rivermead Assessment of Somatosensory Performance (RASP; Winward et al., 2002). The RASP, which requires commercially registered equipment (Winward et al., 2002), involves more modalities than the previously mentioned instruments but does not include stereognosis, which is the most important component of hand sensation in people with stroke (Gaubert & Mockett, 2000).

The Revised Nottingham Sensory Assessment (rNSA; Lincoln, Jackson, & Adams, 1998), a shorter version of the original Nottingham Sensory Assessment (Lincoln et al., 1991), contains all the sensory modalities, including stereognosis, based on those used in everyday clinical practice (Connell et al., 2008) and might be considered for clinicians to assess comprehensive somatosensory impairments in people with stroke. The rNSA has acceptable interrater reliability, but its validity and responsiveness have not been reported (Lincoln et al., 1998).

Valid and responsive measures are required for assessing improvement, making decisions, and justifying an intervention (Andresen, 2000; Wright & Young, 1997). Because somatosensory impairments not only correlate with sensorimotor recovery but also influence daily participation, the measurements applied to sensory, motor performance, and daily participation in people with stroke are suitable for the criterion measures. Because the validity and responsiveness of the rNSA have not been fully established specifically in people with stroke, more studies of this group are warranted. The purpose of this study was to examine the concurrent validity, predictive validity, and responsiveness of the rNSA during rehabilitation in people with stroke.

This study was a secondary analysis of data from previous randomized controlled trials of stroke motor rehabilitation (Lin, Chang, Wu, & Chen, 2009; Lin, Chen, Chen, Wu, & Chang, 2010; Wu, Chuang, Lin, Chen, & Tsay, 2011; Wu et al., 2012). It included 147 people with stroke recruited from the rehabilitation departments of six hospitals. The inclusion criteria used in the randomized controlled trials were clinical diagnosis of a first or recurrent unilateral stroke; no severe spasticity in the paretic arm (score >2 on the Modified Ashworth Scale; Bohannon & Smith, 1987); no serious cognitive deficits (score >24 on the Mini-Mental State Examination; Teng & Chui, 1987); and no other neurological, neuromuscular, or orthopedic disease. The institutional review board approved this study, and all participants provided informed consent before data collection.

Participants were randomly assigned to the treatment group (i.e., constraint-induced therapy, bilateral arm training, or robot-assisted therapy) or the active control group. Interventions in all groups lasted for 1.5–2 hr every weekday for 3–4 wk. Participants were evaluated before and immediately after the interventions. The evaluations were administered by six evaluators with training in clinical assessments and blinded to the group assignments.

The assessment used in this study was the rNSA, and the outcome measures were the FMA–S and the FMA Motor subscale (FMA–M). Participation in daily activities was assessed with the Nottingham Extended Activities of Daily Living (NEADL) scale (Nouri & Lincoln, 1987).

The rNSA, with acceptable intrarater and interrater reliability (Lincoln et al., 1998), examines somatosensory impairment using a 3-point scale (0 = absent, 1 = impaired, 2 = normal) and takes approximately 25–35 min to complete. It consists of 72 items grouped into three subscales measuring tactile sensation, proprioception, and stereognosis. The Tactile Sensation subscale includes light touch, temperature, pinprick, pressure, tactile localization, and bilateral simultaneous touch and is administered to the face, trunk, shoulder, elbow, wrist, hand, knee, ankle, and foot. Higher scores demonstrate better somatosensory function. We calculated the scores of these three subscales in this study.

The FMA–S, part of a widely used measurement in people with stroke, has good balanced usability and robustness (Connell & Tyson, 2012). It includes testing of light touch (four items: upper arms, palmar surface of the hands, thighs, and soles of feet) and proprioception (eight items: shoulder, elbow, wrist, thumb, hip, knee, ankle, toe). Three-point scoring is used (0 = absent, 1 = impaired, 2 = intact). The total score ranges from 0 to 24. High intrarater and interrater reliability have been demonstrated in chronically disabled people after stroke (Duncan, Propst, & Nelson, 1983) and in people less than 6 mo after stroke (Sanford, Moreland, Swanson, Stratford, & Gowland, 1993). The validity and responsiveness of the FMA–S have been established at different poststroke stages of recovery (Lin et al., 2004).

The FMA–M consists of 33 items measuring the movement and reflexes of the shoulder, elbow, forearm, wrist, and hand as well as coordination and speed. A 3-point ordinal scale is used (0 = cannot perform, 1 = can perform partially, 2 = can perform fully). A higher FMA–M score indicates less motor impairment (Fugl-Meyer, Jääskö, Leyman, Olsson, & Steglind, 1975). In the field of stroke rehabilitation, the FMA–M is recognized as a valid and reliable clinical measure of poststroke motor impairment severity (Sanford et al., 1993; Sullivan et al., 2011).

The NEADL scale has sound psychometric properties (Harwood & Ebrahim, 2002; Hsueh, Huang, Chen, Jush, & Hsieh, 2000) and has been recognized as an easily administered scale of daily participation. The NEADL items ask people with stroke to rate levels of difficulty in conducting 22 tasks relevant to motor function and kitchen, domestic, and leisure activities. The NEADL items are scored on a 4-point scale (0 = unable, 1 = with help, 2 = on my own with difficulty, and 3 = on my own), with higher scores indicative of higher independence.

The normality of data was checked by examining the value of skewness (±1) and visually verified by histograms. The collinearity among rNSA scores was examined by analyzing variance inflation factors (≤10), tolerance (≥0.1), and the condition index (>20). No collinearity problems were found with the data.

The Pearson correlation coefficient was used to examine the association between the rNSA and the outcome measures (i.e., the FMA–S, FMA–M, and NEADL scale) before and after treatment. Stepwise regression was used with pretreatment rNSA scores as predictors and with posttreatment scores of the outcome measures as outcome variables. The following criteria were used to interpret the magnitude of the correlation coefficients: <.25 is low; .25–.5, fair; .5–.75, moderate to good; and >.75, excellent (Portney & Watkins, 2009).

We used the standardized response mean (SRM) to calculate the responsiveness of the rNSA from before treatment to after treatment so that we could identify whether the rNSA responded to the intervention. The SRM, a variant of effect size, is the mean change in the score divided by the standard deviation of the changed scores. According to the Cohen criteria on effect size, 0.8 or more is large; 0.5–0.8, moderate; and 0.2–0.5, small (Wright & Young, 1997).

Table 1 shows participant characteristics and their FMA–S, FMA–M, and NEADL scale scores at pretreatment and posttreatment. Table 2 reports all rNSA subscale mean scores at pre- and posttreatment.

All the rNSA subscales were significantly correlated with the FMA–S at pretreatment and posttreatment (rs = .69–.95, p < .001; Table 3). The rNSA subscales were also significantly correlated with the FMA–M and the NEADL scale at pretreatment (p < .05) and posttreatment (p < .001), except for the correlation between rNSA temperature and the NEADL scale at posttreatment (r = .15, p = .10; Table 3).

Table 4 summarizes the results of stepwise regression analysis. The rNSA proprioception, stereognosis, and tactile-pinprick assessments for the proximal upper limb at pretreatment were significant predictors that explained 82.7% of variance in the FMA–S at posttreatment. The rNSA stereognosis assessment at pretreatment was a significant predictor that explained 11.7% of variance in the FMA–M at posttreatment. The rNSA tactile-pinprick assessment for the proximal upper limb at pretreatment was the only significant predictor and explained 17.3% of variance in the NEADL scale at posttreatment.

The results for all participants indicated that the responsiveness was small to moderate for the rNSA Tactile Sensation (SRM = .54), Proprioception (SRM = .34), and Stereognosis (SRM = .44) subscales. However, at pretreatment, 51.0% and 19.1% of the participants achieved maximum scores on the Proprioception and Stereognosis subscales, respectively; that is, many participants showed no improvements on these subscales. Calculating responsiveness for only the people with stroke who improved was recommended because responsiveness is defined as the ability to detect changes from treatment (Wright & Young, 1997). Therefore, we calculated the responsiveness of the rNSA scores for the 72 participants in our study whose pretreatment scores were below the maximum on any of the rNSA subscales. The results showed that responsiveness was moderate to large for rNSA Tactile Sensation (SRM = .83), Proprioception (SRM = .51), and Stereognosis (SRM = .55) subscales.

To the best of our knowledge, this is the first study to investigate the validity and responsiveness of the rNSA. Our findings may support the use of the rNSA by clinicians and researchers as a comprehensive measure to evaluate somatosensory impairments in people with stroke.

The results showed that the rNSA was significantly correlated to somatosensory performance, motor performance, and daily participation in people with stroke as measured by the FMA–S, FMA–M, and NEADL scale at pretreatment and posttreatment, except for the relationship between rNSA tactile temperature and the NEADL scale at posttreatment. The magnitude of the correlations between the rNSA and the FMA–S was excellent (rs = .75–.95), and the values between the rNSA and the FMA–M (rs = .22–.37) and between the rNSA and the NEADL scale (rs = .15–.33) were low to fair at pretreatment and posttreatment. This result showing that the correlation between the rNSA and the FMA–S was higher than the correlations between the rNSA and the FMA–M and the rNSA and the NEADL scale is not surprising because the rNSA and the FMA–S assess a very similar construct, that is, somatosensory function. The relationships between rNSA sensation and the FMA–M and between rNSA sensation and the NEADL scale were weak to moderate, consistent with previous studies (Lin et al., 2004; Scalha et al., 2011; Welmer et al., 2008; Winward et al., 2002).

Our findings suggested that somatosensory function in people with stroke is modestly associated with motor and activities of daily living (ADL) participation. However, people with chronic stroke might learn to use other compensatory mechanisms, such as vision sensation, for their impaired somatosensory function to perform motor tasks and participate in ADLs, leading to a weak to moderate effect of somatosensory impairments on motor and daily participation (Scalha et al., 2011; Welmer et al., 2008). The other explanation for the limited impact of somatosensory impairment on motor and daily participation is that because the integration of the somatosensory system and motor function is complex as a result of varied neurological pathways and brain regions, several other influencing factors must account for motor and ADL participation (Stein, Harvey, Winstein, Zorowitz, & Wittenberg, 2014).

The rNAS Proprioception subscale had significant predictive ability, accounting for 79.6% of the total variance of the FMA–S after treatment. This ability may be because many FMA–S items (67.0%) are related to proprioceptive abilities and, therefore, the rNSA Proprioception subscale was selected in the regression models and predicted larger amounts of variance in the FMA–S scores.

It is noteworthy and consistent with the findings of previous studies that the rNSA Stereognosis subscale might be the only significant predictor of motor performance. Stereognosis, the recognition of objects using the sense of touch, requires complex sensory processing (Cooper, Majnemer, Rosenblatt, & Birnbaum, 1995) and is suggested to be highly related to motor accuracy (Case-Smith, 1995). Stereognosis, as a higher order aspect of sensation, may thus be important to motor performance and considered a predictor for functional improvement (Carlson & Brooks, 2009). The finding that stereognosis is a significant predictor supports the hypothesis that using the rNSA Stereognosis subscale may enhance the comprehensive assessment of somatosensory function.

The rNSA tactile-pinprick for the proximal upper limb was the only significant predictor of participation in ADLs. Our results extend those of a previous study that showed a significant positive correlation between pinprick sensory function and Barthel Index (Mahoney & Barthel, 1965) scores, which were used to measure basic ADLs (Abdul-Mootal, Al-Nabi, & Al-Hakiem, 2001). Our study adds knowledge confirming that the pinprick sensation for the proximal upper limb might, to a certain degree, predict instrumental ADLs (IADLs), in this case, the complex daily tasks requiring arm–hand use measured by the NEADL scale. In addition, the pinprick, an important protective sensory modality, is necessary to ensure safety during IADLs (Kandel, Schwartz, Jessell, Siegelbaum, & Hudspeth, 2012; Rosanne & Joseph, 2010). On one hand, use of tools may be contraindicated in people with impaired pain sensation after stroke, which reduces practice, resulting in poor performance of IADLs. On the other hand, pain sensation, one of exteroceptive sensation, is conveyed by the spinothalamocortical pathway. Damage along the spinothalamocortical pathway would cause maladaptive sensory and emotional responses to stimuli, resulting in an energy drain that affects a person’s everyday life (Stein et al., 2014; Widar & Ahlström, 2002). Moreover, this finding lends support to findings that sensory feedback from the proximal part of the upper limb, such as the perception of a pinprick, may affect hand function (Hoffmann, Schmit, Kahn, & Kamper, 2011) and result in poor performance of IADLs.

In summary, we recommend the use of all sensory modalities of the rNSA to represent concurrent somatosensory function, motor performance, and daily participation. In contrast, use of specific sensory modalities of the rNSA is appropriate for separately predicting somatosensory functions, motor performance, and daily participation in people with stroke after intervention. In other words, to some extent, proprioception might predict somatosensory functions after an intervention, and stereognosis and tactile-pinprick for the proximal upper limb might predict motor performance and daily participation, respectively, after an intervention. These findings suggest that the rNSA might be valid to measure somatosensory function and that it significantly reflects motor abilities and daily participation in people with stroke. Because of the small variance explained in motor performance and daily participation, we suggest that future research combine information on brain lesions and cognitive function with stereognosis and tactile-pinprick for the proximal upper limb, respectively, to enhance the prediction of motor performance and daily participation after an intervention.

The results for participants who scored below the maximum scores at pretreatment indicated that the three rNSA subscales were moderately to largely responsive in detecting changes after intervention. These findings suggest that the rNSA may be a plausible assessment to detect changes in somatosensory dysfunction after an intervention. A close look at the data revealed that the Tactile Sensation subscale was more responsive to changes after the intervention compared with the Proprioception and Stereognosis subscales. Tactile sensation performance, therefore, might be a better indicator in the evaluation of intervention outcomes of stroke rehabilitation.

A few limitations of the study warrant consideration. First, the previous study indicated that the recovery of position sensation in people with stroke might reach a plateau after 3 mo poststroke (Lin et al., 2004). The time after stroke onset of most participants in our study was more than 3 mo, resulting in limited improvement of proprioception function. Future studies need to recruit participants with subacute stroke. Second, 85% of the participants in our study were people with chronic stroke. A person’s level of improvement after an intervention may be influenced by how long after the onset of stroke the intervention is used; therefore, future studies should evaluate people with subacute and chronic stroke separately. Third, many participants achieved maximum scores on the rNSA Proprioception subscale at pretreatment; therefore, the responsiveness of proprioception might have been overestimated. Future research should consider recruiting participants who score below the maximum scores on the Proprioception subscale at pretreatment.

The following findings from this study may be applied to occupational therapy practice:

• The use of all the rNSA sensory modalities could represent concurrent somatosensory function, motor performance, and participation in ADLs for people with stroke.

• Clinicians could use specific sensory modality of the rNSA to predict functional and participation conditions for people with stroke.

• The Tactile Sensation subscale of the rNSA could be used to detect change as a result of intervention for people with stroke.

This study examined the validity and responsiveness of the rNSA after stroke rehabilitation. The results support the concurrent validity, predictive validity, and responsiveness of the rNSA for people with stroke. We recommend the use of all sensory modalities of the rNSA to represent present somatosensory function, motor performance, and daily participation. In contrast, a different sensory modality could predict different functional outcome. For predicting somatosensory function, motor performance, and daily participation in people with stroke after intervention, the use of specific sensory modalities of the rNSA is appropriate. For example, proprioception could possibly predict somatosensory functions after an intervention, and stereognosis and tactile-pinprick for proximal upper limb could possibly predict motor performance and daily participation, respectively, after an intervention. Moreover, the rNSA Tactile Sensation subscale was more responsive in detecting a change after the rehabilitation intervention than the Proprioception and Stereognosis subscales. These findings may suggest that the rNSA is a practical and valid clinical tool for assessing somatosensory function, significantly reflecting motor abilities and daily participation in people with stroke, and for detecting sensation changes as a result of intervention. Future studies should include people at varying time points after the onset of stroke.

This study was partially supported by the National Health Research Institutes (NHRI-EX104-10403PI), Ministry of Science and Technology (102-2314-B-002-154-MY2, 102-2028-B-182-005-MY3, and 103-2314-B-182-004-MY3), Healthy Aging Research Center at Chang Gung University (EMRPD1E1711), and Chang Gung Memorial Hospital (CMRPD1C0402) in Taiwan.