Background and Purpose: Considerable disagreement exists among researchers with regard to the prevalence, pathophysiology, and treatment of “pusher behavior” (PB), partly because of different testing procedures. This study was primarily aimed at establishing cutoff scores for and the construct validity of the Scale for Contraversive Pushing (SCP). The prevalence of PB in people with right- and left-brain lesions also was investigated.
Subjects and Methods: The study subjects were 105 consecutive patients with recent stroke. Two methods were used to diagnose PB: clinical examination and SCP score with 3 different cutoff points—an SCP total score of greater than 0 (Crit_1), subscores in each section of the scale of greater than 0 (Crit_2), and subscores in each section of the scale of ≥1 (Crit_3). Clinical and SCP diagnoses were independently made by different examiners. The Cohen κ coefficient was used to determine the agreement between clinical and SCP diagnoses. The construct validity of the SCP was estimated by calculation of Spearman rank correlation coefficients for SCP and balance, mobility, and functional scores.
Results: The agreement between clinical and SCP diagnoses was low (κ=.212) when Crit_1 was used. Crit_2 led to the highest agreement with the clinical diagnosis (κ=.933). However, only Crit_3, although globally less accurate (κ=.754), ensured no false-positive results. The construct validity of the SCP was demonstrated by significant (P<.001) moderate to high correlations with mobility (rho=.595), functional (rho=.632), and balance (rho=.666) scores. The prevalence of PB was not influenced by the side of the lesion. A limitation of the study was that the reliability of the clinical examination method was not investigated.
Discussion and Conclusion: The results support the validity of the SCP and suggest the need to choose different SCP cutoff criteria (Crit_2 or Crit_3) according to the aim of the evaluation.
“Pusher behavior” (PB) is a poorly understood disorder exhibited by some people with stroke. This behavior has 3 typical features: (1) contralesional tilted posture with severe imbalance, (2) tendency to push strongly toward the paretic side with the nonaffected limbs, and (3) resistance to external correction of the tilted posture.1 Researchers have reported different findings with regard to the prevalence of PB in the acute phase after stroke. Some authors found that 5% to 10% of people receiving rehabilitation for stroke exhibited the behavior,2–4 but others reported notably higher values, from 40% to 50%5 and 63%.6 Most likely, this large discrepancy is mainly attributable to the use of different criteria for diagnosis, although it also may be a function of the interval between stroke onset and examination. Other terms that occasionally have been used to indicate postural control problems similar to PB in people who have had a stroke are “listing phenomenon”3,7 and “lateropulsion,”8 but these terms have not been used with a unique meaning. Recently, Karnath9 made a useful terminological distinction by indicating precisely that active pushing and resistance to correction of posture are the main features that distinguish PB from other postural disorders in stroke.
Although few data are available, PB may hinder or at least delay functional recovery, especially balance and gait functions.2 Therefore, in recent years, this behavior has received increasing interest. There is growing agreement that PB reflects some misrepresentation of verticality,10,11 but the exact mechanism is still unclear. The orientation of the body relative to the gravito-inertial force requires an implicit representation of verticality that is critical for balance control. This representation is based on visual, somesthetic, and vestibular information.11 Because people who have had a stroke may show a biased perception of uprightness,12 it has been suggested that PB may emerge as an extreme form of this misrepresentation; that is, people with PB may tend to align their body posture with a contralesionally tilted perceived vertical.11 Indeed, Perennou et al11 measured the segmental body orientation in people who had PB and who were seated on a laterally unstable, rocking platform and were asked to keep their bodies aligned to the vertical. The authors found that these people were able to align the head correctly to the vertical but showed a remarkable tilting of the pelvis toward the contralesional side. In contrast to these results, Karnath et al10 found that people with PB, in the absence of vision, felt that their bodies were oriented upright when they were tilted an average of 18 degrees to the ipsilesional, rather than to the contralesional, side; moreover, the perceived vertical of these people was within normal limits when they were allowed to view the structured surroundings. Therefore, the authors hypothesized that the disorder may be the consequence of a conflict between 2 reference systems, that is, a normal visual vertical and an ipsilesionally tilted postural vertical. Not surprisingly, the rehabilitation approaches suggested on the basis of these pathophysiological findings are quite different.10,13
Valid and reliable assessment instruments are necessary to identify the presence and the severity of the behavior, to study the underlying mechanisms, to compare the effectiveness of different treatment strategies, and to define the prevalence and the prognosis of the disorder.
Karnath et al10 developed an assessment tool, the Scale for Contraversive Pushing (SCP), aimed at diagnosing the presence of PB and quantifying its severity. The SCP comprises 3 sections, each assessing a particular feature of the syndrome, namely, the symmetry of spontaneous body posture (section A), the use of nonaffected extremities (leg or arm) to push by abduction and extension thrust (section B), and the resistance to passive correction of the tilted posture (section C). For each item, the score ranges from 0 to 1. Because each feature is examined in both sitting and standing positions, the maximum score in each section is 2. Recently, the authors completed and integrated instructions and task definitions for SCP administration and scoring.14 The SCP is shown in the Appendix. We evaluated some psychometric characteristics of the SCP, namely, interrater reliability, internal consistency, and diagnostic accuracy, in a sample of 26 people who had recently had a stroke and who were selected on the basis of the presence of postural impairment.15 We found that the interobserver reliability of the SCP was very high for both subscores (intraclass correlation coefficients= .944, .929, and .939 for sections A, B, and C, respectively) and total score (intraclass correlation coefficient= .971). The scale also showed excellent internal consistency (Cronbach alpha=.919).
With regard to the validity of the SCP as a diagnostic tool for PB, we addressed the issue of the appropriate cutoff scores for the scale. Cutoff scores for the SCP have not been definitively established, and controversy currently exists regarding the most appropriate way to interpret SCP scores. On the basis of their clinical experience, in their first article on the topic, Karnath et al10 suggested a score of ≥1 in each section of the scale as the cutoff criterion for making the diagnosis of PB. In their study on the natural history of PB after stroke, however, Danells et al6 used a much less conservative criterion and assigned a diagnosis of PB to all people who had an SCP total score of greater than 0, but they did not assess the validity of this cutoff score. We investigated the validity of the original criteria of Karnath et al10 by calculating the agreement between the clinical diagnosis, according to the recommendations of Davies,1 and the SCP diagnosis. We found that the criteria of Karnath et al10 led to many false-negative results when applied to a sample of people with stroke and postural impairment. Consequently, we proposed the use of a different criterion, that is, a score of ≥0 in each section of the SCP; this criterion improved substantially the validity of the scale as a diagnostic tool for PB in that selected sample of people with stroke.15 This criterion, however, as well as the other suggested cutoff points for the SCP, needs to be tested in an unselected, more heterogeneous, and larger sample of people with hemiplegia, in whom the rate of true-negative results (people without a diagnosis of PB) could be expected to be very high.
The construct validity of the SCP has not been formally investigated yet. Because PB is a distinctive disorder of postural control in which the perception of body orientation is severely disturbed,10,11 it may be expected that the disorder will impair activities that require managing gravitational force when moving the body to an upright posture and balancing. In the acute stage of recovery from stroke, balance impairment is a strong predictor of function in terms of activities of daily living and mobility,16 and the severity of balance impairment correlates with the severity of disability.17 On the basis of these findings, we believed that the construct validity of the SCP would be supported by a moderate correlation between SCP scores and balance, mobility, and disability scores.
The present study was primarily aimed at investigating the validity of the suggested cutoff scores for the SCP and at evaluating the construct validity of the scale. A further objective was to estimate the prevalence of PB and its relationship to the side of the lesion in people receiving neurorehabilitation for stroke.
All consecutive patients admitted for stroke to 2 inpatient rehabilitation hospitals (Villa Fiorita Hospital, Prato, Italy, and Piero Palagi Hospital, Florence, Italy) from January 2006 to December 2006 were included in the study, provided that they met the following 4 criteria: first stroke; monolateral stroke, as indicated by computed tomography scans or magnetic resonance imaging; recent stroke (≤30 days from onset); and presence of at least minimal unilateral motor impairment at upper or lower limbs or presence of balance deficit (or presence of both), as indicated by impaired performance on the motricity and balance sections of the Fugl-Meyer Assessment Scale (FM).18 The characteristics of the selected study participants are shown in Table 1. Participants gave their written informed consent for collection, storage, and use of personal data.
At admission to the rehabilitation unit, each participant was assessed by a staff physical therapist, who administered the FM, the mobility section of the motor assessment chart developed by Lindmark and Hamrin (LIND-MOB),19 and the Barthel Index (BI).20 These tests are routinely administered to all people admitted to the Villa Fiorita and Piero Palagi hospitals for stroke rehabilitation. The FM is a 226-point multiple-item scale that evaluates balance and motricity, sensation, range of joint motion, and pain in the paretic arms. All items are scored on a 3-point (0–2) ordinal scale, in which 0 indicates no function. The excellent psychometric characteristics of the FM are well established.21 The LIND-MOB evaluates an individual's abilities in rolling to the affected and nonaffected sides, sitting up from a lying-down position and the reverse, leaning forward from a seated position to touch the feet, standing up, and walking. Items are scored according to the need for assistance in order to complete the task on a 4-point (0–3) ordinal scale, in which 0 indicates the inability to perform the task. In a sample of 21 people with acute stroke, the LIND-MOB showed good intrarater reliability (kappa values ranging from .68 to .84) and interrater reliability (kappa values ranging from .68 to .76).22 The BI, a 10-item performance-based scale that evaluates independence in basic activities of daily living, has been shown to have excellent reliability and validity.23 Because the presence and the severity of PB are expected to interfere with balance, mobility, and functional level, these evaluations were used to estimate the construct validity of the SCP.
Four examiners assessed the presence and severity of PB within 3 days of admission. One examiner was a geriatrician with experience in stroke rehabilitation. The others were physical therapists with many years of experience with people who have had a stroke. At each rehabilitation unit, people were independently evaluated by 2 examiners within a few hours of each other. One of the examiners (MP and CB at Villa Fiorita and Piero Palagi hospitals, respectively) selected participants and made the clinical diagnosis of PB according to the recommendations of Davies.1 The clinical examination in the 2 hospitals was based on the same criteria and focused on careful observation of people while lying down, sitting, standing, weight transferring, and walking. The examiners watched for characteristic signs of the behavior, such as whole-body incorrect alignment manifested as typical lengthening or shortening of the affected or unaffected side of the trunk and active extension in the unaffected extremities with resistance to be moved toward the unaffected side. However, the examiners also watched for other signs, such as difficulty in placing the unaffected leg in the supine position, flexion of the paretic leg while weight transferring toward the contralateral side during standing, and adduction of the paretic leg in the swing phase of gait. People were diagnosed as having PB when at least 2 of the above-mentioned signs were present, with one of them judged as severe.
The second examiner (LN or MB) administered the SCP and diagnosed the participant as having PB or not having PB on the basis of 3 cutoff criteria—a total SCP score of greater than 0 (ie, the criterion used by Danells et al,6 defined as Crit_1); a score of greater than 0 in each SCP section, leading to a total SCP score of ≥1.75 (ie, the criterion suggested by Baccini et al,15 defined as Crit_2); and a score of ≥1 in each SCP section, leading to a total SCP score of ≥3 (ie, the original criterion suggested by Karnath et al,10 defined as Crit_3). The second examiner was unaware of all earlier assessments as well as the clinical diagnosis of PB.
The agreement between the SCP diagnosis resulting from the 3 different cutoff criteria and the clinical diagnosis was estimated by calculation of the Cohen kappa coefficient (κ). From SCP scores and clinical evaluations, sensitivity, specificity, predictive value of a positive test, predictive value of a negative test, positive and negative likelihood ratios, and global diagnostic accuracy of the scale were calculated for each criterion, along with 95% confidence intervals. The construct validity of the SCP was estimated by calculation of the Spearman rank correlation coefficients for the SCP total score and subscores and the balance subscore of the FM (FM-BAL), the LIND-MOB subscore, and the BI score.
Comparisons of the prevalence of a clinical diagnosis of PB in people with right- and left-brain lesions and in people with stroke who were referred to the 2 rehabilitation units were carried out with the chi-square test.
The level of statistical significance was set at .05. Data analyses were performed with the statistical package SPSS Version 12.0 for Windows.* Validity indexes and 95% confidence intervals were calculated with Microsoft Excel 2003.†
Validity of Different Cutoff Criteria
The Figure shows the flow diagram for the study. In the population studied, 17 of 105 participants (16.2%) were diagnosed as having PB on the basis of the clinical examination. The core features of PB in these participants are summarized in Table 2. On the basis of the SCP scores, however, a diagnosis of PB was made for 65, 19, and 11 participants when Crit_1, Crit_2, and Crit_3, respectively, were used. Consequently, when different cutoff criteria were used, the clinical and SCP diagnoses differed (Tab. 3), as did the validity indexes of the scale (Tab. 4). When Crit_1 was used, the diagnostic agreement between the SCP and the clinical examination was poor: agreement was found for only 54.3% of participants (57/105; Cohen κ=.212, SE=.052). Forty were diagnosed as not having PB and 17 were diagnosed as having PB by both methods. With this cutoff criterion, 48 participants were incorrectly diagnosed by the SCP as having PB. Consequently, despite perfect sensitivity, the scale had very low specificity, and the predictive value of a positive test was very low as well.
Both Crit_2 and Crit_3 led to much better agreement between SCP and clinical diagnoses. The best SCP diagnostic accuracy was found with Crit_2, which correctly diagnosed 98.1% of participants (103/105; Cohen κ=.933, SE=.047). With this criterion, the specificity of the scale was greatly enhanced, without any decrease in sensitivity or the predictive value of a negative test. However, Crit_2 led to false-positive results in 2 participants. Crit_3 led to correct diagnoses in 94.3% of participants (99/105; Cohen κ=.754, SE=.094). No false-positive results were obtained with the scale when this more conservative criterion was used, but 6 participants were classified as not having PB, contrary to the clinical diagnosis. Therefore, when Crit_3 was used, the scale was highly specific, but its sensitivity was relatively low.
SCP Construct Validity
Table 5 shows the Spearman rank correlation coefficients for SCP scores and functional (BI), mobility (LIND-MOB), and balance (FM-BAL) examinations. For the SCP total score, the correlations were moderate to high, and all were significant at the .001 level. The values are negative because, for the SCP, lower values indicate better function, whereas for the other instruments, the opposite is true. Section A scores showed distinctly higher correlations with the other measures than section B and C scores.
Prevalence of PB in Right- and Left-Brain Lesions
According to the clinical examination, the overall prevalence of PB in the population studied was 16.2%. Prevalence did not differ significantly between the 2 rehabilitation units (16.1% and 16.3% at Villa Fiorita Hospital and at Piero Palagi Hospital, respectively; P=.972). With regard to the side of the lesion, no significant differences were found between participants with right- and left-brain lesions (18.4% and 14.3%, respectively; P=.571).
The main purpose of the present study was to determine the validity of 3 cutoff criteria suggested for using the SCP as a diagnostic tool for PB. The results confirmed that the cutoff score used by Danells et al,6 which assigned a diagnosis of PB to all people with an SCP score of greater than 0, is absolutely inadequate, leading to a large number of false-positive diagnoses. This result is not surprising, because most people who have had a stroke show an asymmetric posture, with less weight bearing by the affected leg than by the unaffected leg.24,25 A lateral tilting in itself, however, does not indicate the presence of clear PB. As indicated in our earlier article,15 the presence of all of the typical features of the syndrome seems to be necessary for correctly diagnosing PB in people after stroke. The findings in the present study strongly support this assumption.
The results also indicated that the criterion we suggested in previous research15 seems to be the most suitable for making a correct diagnosis, even when applied in an unselected, more representative sample of people with hemiplegia. However, although correctly classifying 98.1% of people, this cutoff score misdiagnosed 2 participants as having PB. Therefore, when this criterion is used, the presence of a minimal number of false-positive diagnoses cannot be excluded. Conversely, when the original cutoff score suggested by Karnath et al10 was used, no false-positive diagnoses emerged. This more conservative criterion proved to be less accurate than the former because it failed to detect the presence of the behavior in 6 of 17 people (35%). However, in this larger and heterogeneous sample of people with stroke, it was significantly more suitable than we found it to be in our previous research.15 This result also was expected because in the general population of people with stroke and receiving rehabilitation, the rate of true-negative results is expected to be much higher than that in a sample of people with stroke and selected for the presence of postural disorders.
On the basis of these findings, we suggest that SCP cutoff criteria be selected according to the aim of the evaluation. For epidemiological purposes, we believe that the cutoff criterion that proved to be the most accurate, that is, a score of greater than 0 in each section of the scale (Crit_2), is the most suitable. This criterion may be useful for studies aimed at investigating the prevalence of the syndrome or its association with other features, such as the side of a cerebral lesion or the presence of neglect. There is considerable disagreement among researchers about these topics,2,4–6 and it may be argued that conflicting findings may be partially attributable to different criteria for diagnosing PB.
On the other hand, in research about the pathophysiology of PB, a correct diagnosis is a prerequisite to drawing correct inferences. For example, there is much debate about a hypothetical bias in the perceived postural or visual vertical and whether the bias is contralesional or ipsilesional.10,11,24,25 We suggest that, in studies aimed at investigating these issues, the use of both Crit_2 and Crit_3 for diagnosing the syndrome would be advisable. The more conservative Crit_3 should be used for selecting people who definitely exhibit PB, whereas Crit_2 should be used to include people without PB. In this way, people with an uncertain diagnosis would be excluded, and the risk of an incorrect allocation of people to the PB group or the non-PB group would be minimized.
The preceding discussion is based on the assumption that the clinical diagnoses were correct. A limitation of the present study was that we did not formally assess the reliability of the method used to make the clinical diagnoses. Therefore, possible bias related to slightly different criteria used in the clinical examinations cannot be excluded.
Our findings also provide evidence regarding SCP construct validity, because SCP scores correlated significantly with functional, mobility, and balance evaluations. Not surprisingly, the strongest correlation was found with balance scores, because balance is directly affected by PB. People with PB, in fact, use their nonaffected extremities to push toward the contralateral side and eventually fall.9 Differences in the strength of the correlations found for the 3 sections of the SCP most likely reflected the particular aspects of the disorder being investigated. As mentioned above, active pushing and resistance to passive correction of posture (assessed in sections B and C) are distinguishing features of PB,9 whereas laterally tilted posture and falling (assessed in section A) are not.15 Indeed, symmetry of posture and ability to sit and stand upright were impaired in a significant proportion of people without PB. These people performed poorly in the balance, mobility, and functional tests but usually scored 0 in sections B and C of the SCP. Overall, it may be speculated that PB influences performance insofar as it affects the ability to balance. However, a direct effect of active pushing on performance, such as the ability to roll and rise from a lying-down position or to transfer, may be present as well.
Recently, a modified form of the SCP was published and validated.26 In a small sample of people with hemiplegia (N=19), the authors found low to moderate correlations between the revised SCP scores and balance and functional scores at both admission (r=.52 and r=.43, respectively) and discharge (r=.49 and r=.45, respectively). These values are distinctly lower than the values that we found for the original scale. Indeed, the modified version of the SCP is different enough from the original scale to be rather a completely new assessment tool. Our results indicate that the original version of the SCP seems to be preferable because it correlates better with functional and postural abilities. Moreover, the cutoff score for diagnosing PB with the modified scale has not been empirically investigated.
The validity of the SCP also seems to be slightly higher than the validity of a lateropulsion scale (LS) developed by D'Aquila et al.8 In a convenience sample of 85 people with stroke evaluated at admission to a rehabilitation unit, the authors reported r=.57 for LS and FM-BAL scores and r=.56 for LS and mobility (Functional Independence Measure) scores. Overall, the 13 motor or mobility items of the Functional Independence Measure that assess self-care, sphincter control, transfer, and locomotion are quite similar to the items of the BI. Both of these values are lower than the values found in the present study. In addition to reliability and validity, however, evaluation instruments should demonstrate good responsiveness, that is, the ability to detect changes over time.27,28 Further research should be aimed at comparing the SCP and the LS with regard to this important psychometric feature.
Finally, the similar prevalences of PB found in people with right- and left-brain lesions are consistent with other findings. The mean interval from stroke onset in the sample studied was 13.7 (SD=6.2) days, and other research has shown that in this subacute phase, the syndrome affects people with right hemiplegia and left hemiplegia equally.2,5 The clinical impression that PB is more frequently exhibited by people with right-brain injury is most likely attributable to the fact that the behavior may persist longer in people with right-brain lesions than in those with left-brain lesions. When assessed several weeks from stroke onset, in fact, PB was found to be significantly more frequent in people with left hemiparesis.5
The data presented here indicate that the SCP cutoff score suggested in our previous study15 is superior to other criteria for correctly classifying nearly all people with stroke as having PB or not having PB. However, the use of the more conservative original criterion of Karnath et al10 is advisable when the absolute certainty of no false-positive diagnoses is requested. The construct validity of the SCP seems to be slightly higher than that of other published scales for PB. The sensitivity to change of the SCP and of the other scales still needs to be addressed.
Mr Baccini, Mr Paci, and Mr Rinaldi provided concept/idea/research design. Mr Baccini provided writing and data analysis. Mr Baccini, Mr Paci, Dr Nannetti, and Ms Biricolti provided data collection. Mr Baccini and Mr Rinaldi provided project management. Mr Rinaldi provided facilities/equipment. All authors provided consultation (including review of manuscript before submission).
↵* SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.
↵† Microsoft Corp, One Microsoft Way, Redmond, WA 98052-6399.
- Received June 21, 2007.
- Accepted April 17, 2008.
- American Physical Therapy Association