Methodological quality of studies

The quality of the included studies was appraised using the COSMIN taxonomy of measurement properties and definitions for health-related patient reported outcomes [9–11] and the COSMIN Risk of Bias checklist [18] for systematic reviews of patient-reported outcome measures. The methodological quality of each study was individually assessed to evaluate whether it met the standards for measurement tool development, content validity, structural validity, internal consistency, cross-cultural validity/measurement invariance, reliability, measurement error, criterion validity, hypothesis testing for construct validity and responsiveness. The Risk of Bias checklist rated each measurement property as either “very good”, “adequate”, “doubtful” or “inadequate”. As there is no accepted “gold standard” measure of upper limb activity, criterion validity was not evaluated, and construct validity and responsiveness properties were appraised within the hypothesis testing criteria of COSMIN. Where a priori hypotheses were not stated, studies were assigned an appropriate generic hypothesis from the list developed by the COSMIN group [18]. Information regarding the interpretability and generalizability were collected.

Quality of measurement properties

The results of individual studies reporting on the psychometric properties were then evaluated using Terwee’s quality criteria for measurement properties [9], see S1 File. Results were rated as sufficient ‘+’, indeterminant ‘?’or insufficient ‘-’.

Sample size of studies

Sample size was only assessed within individual studies evaluating the measurement properties of content validity, structural validity and cross-cultural validity as per COSMIN guidelines. Sample sizes of individual studies evaluating the remaining measurement properties were not assessed via the Risk of Bias Checklist, and sample sizes per those measurement properties were instead pooled at the synthesis stage [9].

Synthesis of best evidence

All identified evidence and results were then pooled and the modified COSMIN GRADE approach used to determine the overall quality of the evidence [9]. The modified COSMIN GRADE approach considers and downgrades the level of evidence and consequently trustworthiness of results depending on the risk of bias (methodological quality), inconsistency of results, imprecision (based on total sample size) and indirectness (evidence from different populations than the population of interest) [9, p1151]; indirectness was not applicable in this review as studies conducted in samples other than those specified in the inclusion and exclusion criteria were excluded. The synthesis determines either “high”, “moderate” “low” or “very low” quality levels of ‘sufficient’, ‘insufficient’, ‘inconsistent’ or ‘indeterminant’.

Action Research Arm Test

The Action Research Arm Test (ARAT) [173] is an obervational performance test that evaluates a person’s ability to use their upper limb to handle objects using grasp, grip, pinch and gross motor movements. Twelve studies evaluated the psychometric properties of the ARAT [38, 45, 57, 63, 75, 113, 114, 120, 129, 140, 142, 172], four of those studies specifically identified participants with upper limb spasticity [38, 45, 75, 114]. The majority of studies included participants post-stroke with a single study including a mixed sample, post-stroke and TBI [120].

Content validity. The Upper Extremity Function Test (UEFT) [174] was modified by Lyle [173] to produce the ARAT. No further content validity studies were identified. The ARAT was found to have sufficient relevance, but indeterminant ratings for comprehensiveness and comprehensibility and no participants were interviewed regarding those properties.

Results for whole sample. Research supports hierarchical ordering of items [173] and reliability within (ICC = 0.99) and between raters (ICC 0.99) [172]. The ARAT was found to correlate highly with other like-tests of activity and dexterity (r = 0.65–0.95) [57, 63, 129, 140, 142, 172] and weak to moderately with the Functional Independence Measure (FIM), a more global measure of function (r = 0.47) [140]. ARAT scores were not, however, a predictor of overall quality of life [129]. The ARAT was found to be responsive over time in acute as well as chronic stroke and TBI samples [38, 57, 114, 140]. ARAT was found to be equally sensitive to change as like measures when used with participants less than 6 months post-stroke [57, 140]. Mixed results have been reported with respect to ceiling effect in stroke populations [63, 75] and there is one study which has reported a minimal, clinically important change of 12 points (dominant) and 17 (non-dominant) [113].

Results pertaining to sample with upper limb spasticity. The ARAT correlated strongly with like measures of activity and dexterity (r = 0.69–0.95) [38] and less with a global measure of function (Functional Independence Measure (FIM) r = 0.2–0.6) [114] and impairments, including grip and pinch strength, spasticity and AROM (r = - 0.28–0.86) [38, 45, 114]. The ARAT was moderate to highly responsive to capture change in participants less than 6 months post-stroke (ES = 0.55–1.018) [38, 114], being as equally responsive as like measures (NHPT and Jebsen-Taylor test of hand function), more responsive than measures of impairment (pinch and grip strength), but less responsive than the SIS-Hand (ES = 0.55–1.018) [38]. Neither a floor nor ceiling effects were found in a sample of participants greater than 6 months post-stroke [75].

Arm Activity measure

The Arm Activity measure (ArmA) is a 20-item self-report tool which includes 7 passive and 13 active items to capture real arm activity in neurological populations [33]. Five studies [30–34] evaluated the psychometric properties of the ArmA, the majority of studies included a mixed sample including participants post-stroke, TBI and MS. All included studies specifically identified participants with upper limb spasticity.

Content validity. The ArmA was developed based on goal analysis, systematic literature review and a modified Delphi survey which demonstrated relevance, comprehensiveness and comprehensibility [30, 33].

Results pertaining to sample with upper limb spasticity. The ArmA subscales demonstrated internal consistency (passive subscale α = 0.85, active subscale α = 0.96) and retest reliability (quadratic weight kappa 0.90 (CI 0.68–1.12), active subscale 0.93 (CI 0.71–1.15)) in a sample with upper limb spasticity [34]. The ArmA demonstrated convergent and divergent validity with passive and active items of the Leeds Adult Spasticity Scale (LASIS) and Disabilities of Arm Shoulder and Hand (DASH) (convergent: Rho 0.48; p = 0.01 to 0.63; p = 0.01; divergent: Rho 0.02; p = 0.9 to 0.23; p = 0.078) [34] and was found to be responsive [32, 34]. Preliminary analysis suggests clinically meaningful change is indicated by 2.5 or 3 point improvement (passive subscale) and 1.1 or 2.5 point improvement (active subscale) [34]. The ArmA active function subscale suffered a ceiling effect (37%), however no floor effect was observed for either subscale [34].

Assessment of Quality of Life

The Assessment of Quality of Life (AQoL) is a generic HRQoL measure that assesses independent living, social relationships, physical senses, psychological wellbeing and illness [90]. Three studies evaluated the psychometric properties of the AQoL, one included participants greater than 6 months post TBI [89] and two less than 6 months post-stroke [90, 154]. Neither study specifically identified participants with upper limb spasticity.

Content validity. Development research underpinning the AQoL [90] demonstrated sufficient relevance, but indeterminant ratings for comprehensiveness and comprehensibility. No other content validity studies conducted in a neurological sample were identified.

Results for whole sample. The AQoL discriminated between participants with and without TBI (effect size (ES) = 0.80), with participants post TBI scoring 2.0 utilities lower than participants without [89]. The AQoL correlated more strongly with measures of handicap (London Handicap Scale (LHS) r = 0.83) than disability (Barthel Index (BI) r = 0.77) or impairment (National Institute of Health Stroke Scale (NIHSS) r = -0.69) in the first 6 months post-stroke and was a significant predictors of death or institutionalization at 12 months [155]. No floor or ceiling effects (1–2%) were found in a stroke population [155].

Barthel Index

The Barthel Index (BI) was initially developed to score the abilities of participants to care for themselves [123]. The BI evaluates 10 activity areas, with a maximum score of 100 indicating independence in all included areas. Six studies evaluated the psychometric properties of the BI [28, 72, 111, 123, 164, 166]. Five studies were completed with participants post-stroke, 4 included participants less than 6 months post-stroke [28, 72, 111, 164], 1 greater than 6 months post-stroke [166] and 1 discussed tool development with a non-specific sample [123]. No included studies specifically identified participants with upper limb spasticity.

Content validity. No research on the development of the BI was located.

Results for whole sample. The BI correlated moderately with measures of upper limb function (Fugl-Meyer Rho = 0.60) (Functional Test for the Hemiplegic/Paretic Upper Limb Rho = 0.61) [72] and global measures function (FIM r_s = 0.95, p<0.0001; Modified Rankin Scale (MRS) r_s = 0.89, p<0.0001; Office of Population Censuses and Surveys (OPCS) disability instrument r = 0.73, p<0.001) [111, 166]. The BI was equally responsive to change within the first three months post-stroke as like global measures (FIM) [164] and a measure of motor function (Fugl-Meyer Test) [72], however determined responsiveness was low. Evidence of a ceiling effect was found in a sample greater than 6 months post-stroke [166].

Barthel Index (Collin & Wade)

The Barthel Activities of Daily Living Index (BI C&W) [49] is a modification of the original BI measurement tool, with all 10 areas of activity included but is scored in increments of 1 rather than 5 as per the original BI [123]. Nine studies evaluated the psychometric properties of the BI(C&W) [35, 49, 56, 83, 97, 149, 159, 163, 167], 6 studies included participants post-stroke [35, 56, 83, 149, 163, 167] and 3 included mixed samples (stroke, MS, TBI) [49, 97, 159]. No studies specifically identified participants with upper limb spasticity.

Content validity. No information presenting the methodology used to revise the original BI was found, only justification from revised test authors who felt the original five-point incremental scoring was misleading in accuracy [49].

Results for whole sample. Research supports use of a summed BI(C&W) score due to a single factor (68% of variance) underlying the scale [163]. While the hierarchical nature of the BI(C&W) was supported by Wade and Hewer [163], Barer and Murphy [35] reported a failure to meet Guttman scaling criteria. Test-retest reliability results appear mixed, with high agreement (75%) between scores but variations in kappa (-0.99 to 0.81) [83]. Inter-rater reliability between self-report, family, nursing staff and skilled observers was acceptable (agreement within 2 points or less for 72% of participants) [49]. The BI(C&W) was strongly associated with measures of upper limb activity (r = 0.729–0.826) (Motricity Index Upper Limb (MI UL) and Motricity Index (MI) total, Frenchay Activity Index (FAI)), complex daily activities (r ≥ 0.80), and disability (r_s = 0.726–0.80) (London Handicap Scale, Modified Rankin Scale (MRS)), and less with measures of psychological wellbeing and impairments (depression, anxiety, pain) (r = 0.2–0.423) [56, 149, 163, 167]. Research suggests that BI(C&W) is at least equally responsive to FIM [97, 159]. However, BI(C&W) suffered from floor and ceiling effects across the acute through to community continuum in a mixed neurorehabilitation sample [97, 149, 159, 167].

Chedoke-McMaster Stroke Assessment

The Chedoke-McMaster Stroke assessment (CMSA) is comprised of two parts; the impairment inventory and the activity inventory (formerly known as the disability inventory) [81]. The CMSA impairment inventory classifies participants into subgroups based on the stages of motor recovery, while the CMSA activity inventory provides a measure of activity performance. Four studies evaluated the psychometric properties of the CMSA, two included participants less than 6 months post-stroke[54, 81], two did not report on the length of time post-stroke for participants [80, 128] and no study specifically identified participants with upper limb spasticity.

Content validity. Evidence located for the development of the CMSA [80, 128], did not indicate participants were consulted on the comprehensiveness or comprehensibility of included items. Relevance of items for the intended purpose of assessment of stroke clients within rehabilitation setting was sufficient, however further content validity studies were not identified.

Results for whole sample. Evidence supports the reliability of the CMSA; inter-rater (ICC 0.88 (95% CI 0.76–0.94) to 0.99 (95%CI 0.98–1.00)), intra-rater (ICC 0.93 (95% CI 0.85–0.96) to 0.98 (95% CI 0.95–0.99)), test retest (ICC 0.98 (95% CI 0.95–0.99)) [81]. Consistent with the definition of the CMSA, strong correlations with both subscales and total scores for like measures of upper limb activity performance (Fugl-Meyer r = 0.95, p<0.001) and global measures of function (FIM r = 0.79, p<0.05) were demonstrated [81]. The predictive validity through use of the Gowland’s predictive equations, however, were not supported due to large error associated with the predicted value [54]. The CMSA was found to be more responsive than the FIM when used with participants less than 6 months post-stroke [81].

Disability Assessment Scale

The Disability Assessment Scale (DAS) is a brief measure of functional disability [42]. Two studies were included, both identified participants with upper limb spasticity [42, 58].

Content validity. Brashear and colleagues [42] reported the development of the DAS to fill the identified gap within the evaluation of functional impairment commonly seen in participants with post-stroke upper limb spasticity (i.e. dressing, hygiene, limb position, pain). No additional research underpinning measurement tool development was reported.

Results pertaining to sample with upper limb spasticity identified. Good to excellent intra-rater reliability (78% of evaluations weighted kappa ≥ .4) and good inter-rater reliability (Kendall W 0.49 (95% CI 0.30–1.00, p < .001) to 0.77 (95% CI 0.37–1.00, p < .001) was reported when used by professionals (neurologists, physiatrists, occupational therapists and physical therapists) with a mean of 6 years clinical experience [42]. Greater DAS scores were found to be associated with Stroke-Adapted Version of the Sickness Impact Scale (SA-SIP) scores (P < .05), reduced quality of life and caregiver burden (P < .05) [58, 175].

EuroQol-5 dimension

The EuroQol-5 dimension (EQ-5D) is a generic measure of health-related quality of life [73, 78, 169]. Nineteen studies evaluated the psychometric properties of the EQ-5D, including participants with MS (n = 6), [73, 106–108, 127, 131] a mixed neurological sample (n = 1) [27] and post-stroke (n = 12) [36, 37, 58, 60–62, 78, 135, 136, 148, 171]. Two studies specifically identified participants with upper limb spasticity [58, 78].

Content validity. During the development of the EQ-5D there is no evidence that participants were consulted on the comprehensiveness or comprehensibility of included items. Relevance of items for the intended purpose was sufficient [108]. The EQ-5D contains 6 of 9 recommended dimensions for patient-based, health related quality of life measures and is less comprehensive than the Stroke Impact Scale (SIS) [148].

Results for whole sample. Test-retest reliability of the patient-reported EQ-5D was moderate to good for VAS and the mobility domain (ICC ≥0.70) [61, 73], test-retest reliability was lower in proxy-reported scores [61]. The EQ-5D correlated moderately with global measures of function such as the EDSS (r = -0.66) [73], but was less sensitive than disease-specific quality of life scales and the generic SF-36 when used with participants with MS [131]. A single study found a moderate inverse relationship between the EQ-5D and the Nine Hole Peg Test, a specific measure of upper limb use (r = -0.56) [73]. When used with participants post-stroke, the EQ-5D correlated with global measures of function including the SF-6D, a classification for describing health from a selection of SF-36 items (r = 0.77) [37] and the SF-36 (r = 0.57–0.63) [60]. Evidence of the discriminant ability was found between participants post-stroke and those who had not suffered a stroke [36, 171], between stroke type and severity [62], and between participants with and without spasticity [78]. The EQ-5D Index had the greatest change score when compared to like generic HRQoL measures less than 6 months post-stroke [136], was more responsive to changes in disability (MRS r = -0.36) and daily activities (BI r = 0.57) in comparison to the EQ-5D VAS [136]. Contrarily, neither the EQ-5D Index or VAS was responsive to change over a one year period post-stroke despite 23.8% of participants reporting improvement and 23.2% deterioration [135]. The EQ-5D did not demonstrate either floor and ceiling effects when used with acute participants post-stroke [136].

Results pertaining to sample with upper limb spasticity identified. The EQ-5D index scores were found to correlate with measures of disability (p < .002) and carer burden (p < .05) [58] and to distinguish between participants with and without upper limb spasticity post-stroke, with mean differences (-0.07, 95% CI -0.12 to -0.33) equivalent to the MCID established for the EQ-5D for other health conditions (MCID is yet to be established for post-stroke populations) [78].

Modified Frenchay Arm Test

The modified Frenchay Arm Test (mFAT), reduces the 25 clinical tests to 5 so as to measure arm function after stroke [94]. Two studies evaluated the psychometric properties of the mFAT [94]; no studies specifically identified participants with upper limb spasticity.

Content validity. No studies were identified providing information targeting measurement tool development and/or content validity.

Results for whole sample. There was evidence for the reliability of the mFAT (inter-rater (Rho = 0.75–0.99), test-retest (Rho = 0.68–0.90 and 0.83–0.99)) when administered to participants 18 months post-stroke [94]. The mFAT was found to be less sensitive than the NHPT in participants less than 6 months post-stroke with mild impairments [94]. Floor effects (30%) and ceiling effects (34%) were evident within acute stroke [94].

Functional independence measure

A total of 20 studies evaluated the psychometric properties, in participants post-stroke (n = 9) [44, 70, 82, 87, 92, 93, 111, 132, 134], TBI (n = 5) [50, 52, 53, 86, 91], MS (n = 2) [141, 151] and a mixed neurological sample (n = 3) [97, 153, 159]. One study specifically identified participants with upper limb spasticity in a sample with MS [141].

Content validity. The FIM was found to have sufficient relevance, but indeterminant ratings for comprehensiveness and comprehensibility during development, as nil information was located to determine if participants were interviewed regarding those properties [105].

Results pertaining to whole sample. A two factor structure was identified for the FIM by a number of researchers, with separate motor and cognitive domains accounting for 89.4 to 97.9% of variance [86, 92, 93, 151]. Evidence for internal consistency has been reported across a number of sample populations (complete FIM α = 0.94–0.98, FIM motor α = 0.93–0.97 and FIM cognitive α = 0.93–0.94 for stroke, MS, traumatic and non-traumatic samples [151, 153]). And between-rater reliability has been demonstrated for both the motor and cognitive domains of the FIM in acute stroke (ICC 0.96, 0.91) respectively [87] and with participants with MS (FIM total inter-rater ICC = 0.99, FIM total intra-rater ICC = 0.94) [151]. Predictive associations between FIM scores and length of stay, discharge destination, minutes of assistance and supervision required on discharge and return to driving were identified [44, 50, 52, 82, 91, 132, 134]. When used with participants with MS, FIM was found to be a valid measure of disability [141], strongly correlating with like global measures (BI r = 0.88), activity measures (Ambulation Index r = - 0.73) and moderate to strongly with specific activity measures including housework (r = 0.64, p<0.001), work (r = -0.59 p<0.001), independence (r = -0.44, p = 0.001), and disability r = -0.96, p< 0.001) [151]. The FIM total score was at best only moderately responsive to change in a neurorehabilitation sample (ES 0.52–0.72), but the FIM cognitive was not (ES = 0.35–0.43) [97]. In comparison to other measures, the FIM was found to be less responsive than the original BI, equally responsive to BI(C&W) in stroke and more responsive than EDSS in MS, yet still only weak to moderately responsive to change (FIM ES = 0.46, FIM SRM 0.53, EDSS 0.15) [141, 151, 159]. Evidence of floor and ceiling effects for FIM were also found [44, 151, 159].

Results pertaining to sample with upper limb spasticity identified. FIM scores correlated with a measures of disability (Kurtkze Expanded Disability Status Scale (EDSS) r_s = -0.69) [141] and was found to be responsive when capturing change in participants with MS (SRM = 0.53) [141].

Goal Attainment Scaling

Goal Attainment Scaling (GAS) was first introduced by Kirusek and Sherman [176] and provides a structured approach to defining and measuring individualized patient centered and/or program based goals. A total of 9 studies evaluated the psychometric properties, in post-stroke (n = 2) [43, 156], MS (n = 1) [104], TBI (n = 3) [59, 120, 125]and mixed ABI (n = 3) samples [41, 115, 124]. Only one study met inclusion criteria that specifically identified participants with upper limb spasticity (in a sample greater than 6 months post-stroke) [156].

Content validity. Not assessed, as GAS identifies goal content particular to individual participants and programs (i.e. high face validity).

Results for whole sample. There were conflicting results in inter-rater reliability within a mixed neurological sample, while Joyce, Rockwood and Mate-Kole [102] report high reliability (r = 0.92, r = 0.94) between an individual rater familiar with GAS and the treating team, Bovend’Eerdt, Dawes, Izadi and Wade [41] found a fair level (ICC_A,k 0.478) and low agreement (LOA -1.52 ± 25.54) between a therapist and masked assessor. When used with participants with MS, GAS change score correlated weakly with the BI (r_s = -0.25) and FIM (r_s = -0.6) [104]. In a sample of participants with ABI secondary to trauma and stroke, GAS also correlated strongly with global clinical impressions (r = 0.81) [104], weak to strongly with measures of daily activity, participation, disability, vocational outcome and quality of life (r = 0.34–0.81) but not with length of stay [102, 124, 125]. In the same sample, GAS at 2 months predicted final GAS scores at the completion of a rehabilitation program ranging from 7 to 42 weeks [125]. Ratings between participants and significant others agreed on 70% of occasions [59]. GAS was more responsive than the FIM and BI (ES 9.0 SRM: 2.4 t value 10.0 z value 1.4) in MS [104] and was responsive to patient centred outcomes and program change in a mixed neurological sample [115].

Results pertaining to sample with upper limb spasticity. GAS was found to have moderate correlations with self-reported benefit (rho = 0.46, p < .001), low correlations with quality of life (rho = 0.07, p = 0.52), disability (rho = 0.19, p = 0.08), carer burden (rho = 0.14, p = 0.26), measures of pain (rho = 0.03, p = 0.77), mood (rho = 0.06, p = 0.61) and spasticity (rho = 0.35, p = 0.001 [156].

Medical Outcome Study 36-Item Short-Form Health Survey

The Medical Outcome Study 36-Item Short-Form Health Survey (SF-36) is a global scale assessing eight health concepts [165, 177]. A total of 24 studies investigated the psychometric properties of the SF-36, 10 included participants with MS [76, 77, 95, 127, 130, 138, 143, 145, 161, 162], 10 post-stroke [29, 60, 61, 67, 85, 96, 122, 133, 148, 168], 3 post TBI [74, 84, 123] and 1 discussed tool development with nil specific sample [165]. No studies specifically identified participants with upper limb spasticity.

Content validity. The development of the SF-36 [165] did not appear to consult participants on the comprehensiveness or comprehensibility of included items [165]. Relevance of items for the intended purpose was sufficient. The SF-36 contains 6 of 9 recommended dimensions for patient-based, health related quality of life, less comprehensive than the SIS [148].

Results for whole sample. The SF-36 was found to have a two-factor structure; with the eight dimensions falling within the two constructs of physical and mental health [177]. Mixed results were found for the use of the domain scores, with scaling assumptions met in the TBI population [84] but only 6 of 8 scales meeting the scaling assumptions in stroke [96]. Evidence for internal consistency of the 8 dimensions, Cronbach alpha >0.70 in majority of studies [29, 61, 74, 76, 84, 161], however dimensions of vitality and general health did not meet this criteria (α = 0.68, α = 0.66–0.68) [85, 96]. Test-retest reliability varied; higher for patient reported scores (ICC = 0.30–0.81) than proxy reported scores (ICC = 0.25 to 0.76) [61, 130, 162]. Individual domains of the SF-36 correlated with like subscales of global measures (all r = ≥ 0.50) post-stroke (EQ-5D) [60] post TBI (Symptom Checklist, Health Problem List, Beck Depression Inventory) [74] and with participants with MS (LHS, FIM, general health questionnaire) [76]. Correlations, however, were not as strong as hypothesized between individual domains and like dimensions for the BI, CNS and FIM post stroke [85, 122] nor with the MSFC in a MS population (r = 0.16–0.51) [145]. The SF-36 physical and mental summary scores had weak to moderate correlations with participants rating of severity of symptoms (r = 0.38, r = 0.18) and quality of life (r = 0.47, r = 0.29) [127, 168]. The ability to discriminate between subgroups of participants with varying levels of function across post-stroke, TBI and MS populations was demonstrated [95, 138, 145, 161, 162]. The SF-36 was more responsive in the first three months post-stroke [85] but less responsive in comparison to other tools measuring associated constructs in MS (ES = 0.01–0.30) [76]. SF-36 did not correlate with FIM change scores, suggesting the change captured within a HRQoL measure was not reflected in a global measure of activity [122]. There was evidence of significant floor and ceiling effects within MS [76, 77] and TBI [84], and varied reports post-stroke [60, 85, 96, 122, 133]. The minimal important clinical change varied across dimensions, reported to be 4–9 points within physical functioning, 6–8 within role physical, 6–7 social functioning and 6 points within the physical summary score [145].

Motor Activity Log

The Motor Activity Log (MAL) is a structured interview designed to capture use of the affected upper limb on two scales, Amount of Use (AOU) and Quality of Movement (QOM) [158]. Five studies evaluated the psychometric properties of MAL; all involved participants post-stroke [47, 63, 88, 157, 158], and one specifically identified participants with upper limb spasticity [88].

Content validity. The MAL was developed based on the non-use model to capture real-world arm function [157]. Item analysis suggests 2 items (put on makeup and write on paper) had greater than 20% missing data, with participants rating as not applicable, and had lower item-total correlations and reliability coefficients [158].

Results for the whole sample. The self-reported QOM scale correlated with performance based measures (ARAT r = 0.61, WMFT r = 0.65) with the AOU scale correlating less strongly with the WMFT r = 0.40 [63, 158]. The minimal detectable change was defined as 16.8% for the AOU and 15.3% for the QOM scales, but the minimal important change was not defined [47].

Results pertaining to sample with upper limb spasticity. The MAL correlated strongly with measures of activity (Chedoke Arm and Hand Activity Inventory (CAHAI) r = 0.82 p<0.01), weakly with measures of participation (Reintegration to Normal Living Index (RNL) r = 0.23 p<0.05) and of varying strengths (weak to moderate) with impairments, stronger than expected (spasticity r = -0.71, strength r = 0.61 to 0.84, pain r = -0.06, sensation r = -0.43, all p<0.01) [88].

Motor Activity Log-28

The Motor Activity Log-28 (MAL-28) is a revision of the MAL-30 with removal of redundant items ‘write on paper’ and ‘put makeup/shaving cream on face’ [158]. A single study evaluated the psychometric properties of this measurement tool involving participants greater than 6 months post-stroke, and without any participants with upper limb spasticity [158].

Content validity. Content analysis indicated appropriate range of items to cover basic (63%) and instrumental (41%) daily activities in addition to items that require finger movement, bimanual and unimanual tasks [158].

Results for the whole sample. Item analysis indicated that 98% of participants encountered included items in daily life [158]. There was evidence for internal consistency (α = 0.94–0.95) and increased test-retest reliability with self-ratings rather than proxy [158]. The MAL-28 held convergent validity with real life measure of hand performance and less with overall physical activity, patient ratings stronger than proxy [158].

Motricity Index

The Motricity Index (MI) is a brief scale of motor recovery [55]. Six studies evaluated the psychometric properties of MI [40, 48, 55, 98, 154, 163]; all involved participants post-stroke, and none specifically identified participants with upper limb spasticity.

Content validity. Demeurisse, Demol and Robaye [55] detailed the development of the MI with mixed results regarding its relevance and no evidence supporting either comprehensiveness nor comprehensibility.

Results for whole sample. There was evidence of the internal consistency of this tool (α = 0.97) [40] and high inter-rater reliability between an experienced and junior doctor (rho = 0.88) rating 20 participants six weeks post-stroke [48]. The Upper Limb MI (UL MI) correlated strongly with like measures of upper limb activity (RMA arm r = 0.73–0.76) [48] and with global measures of activity (BI r = 0.77) [163] whilst correlating moderately with measures of dexterity (NHPT r = 0.36–0.56) [98]. The UL MI correlated strongly with impairments also, including grip strength (r = 0.74–0.94) [40]. The MI, when combined with the visual neglect recovery index and age at 2–3 days post-stroke was a significant predictor of independence at 3 months (β = 0.042, p < .001) and 6 months (β = 0.038, p < .001) [154]. Evidence of a ceiling effect was noted, with 18% of the sample scoring the maximum score within the UL component of the MI on discharge from a rehabilitation ward post-stroke [98]. There was no evidence of a floor effect.

Nine-Hole Peg Test

The Nine-Hole Peg Test (NHPT) is a timed measure of unilateral upper limb dexterity through the placing and removal of nine pegs in/out of a board [178]. Ten studies evaluated the psychometric properties; 5 post-stroke [38, 94, 98, 129] and 5 included participants with MS [39, 51, 79, 139, 150]. One study specifically identified participants with upper limb spasticity [38].

Content validity. The NHPT was first discussed as being used in a study in 1985 [179]; no information was reported to inform the development nor content validity of the NHPT.

Results for whole sample. The NHPT when used with participants post-stroke correlated with both observed (r = 0.36–0.95) [38, 79, 94, 98, 139] and self-reported measures of activity and hand use (r = 0.53–0.66) [98], was more sensitive than the FAT [94], had poor predictive validity in comparison to like measures, and did not predict HRQoL [129]. The NHPT correlated highly with measures of tremor and dexterity in MS, common activity limitation features (r = -0.62 - -0.87 p<0.005) [180]. There was evidence for the reliability of the NHPT (inter-rater Rho = 0.75–0.99 and test-retest Rho = 0.68–0.90 and 0.83–0.99) when administered to participants 18 months post-stroke [94]. The NHPT was moderate to highly responsive within the first 6 months post-stroke (ES = 0.52–0.66) [38, 98], was more responsive than the upper limb MI [98] and measures of strength, equally responsive to the ARAT, Jebsen-Taylor test of hand function and less responsive than the SIS-hand [38]. True change was indicated by a change of 20% when administered to participants with MS [150]. There were no floor or ceiling effects found in the MS population.

Results pertaining to sample with upper limb spasticity identified. Strong correlations with measures of hand use, grip and dexterity were reported in stroke populations (r_s = 0.61–0.95) and with measures of strength (r_s = 0.61–0.82) [38] despite the NHPT being a simulated task performance measure. The NHPT was found to be equally responsive as like measures of upper limb activity performance (ARAT and Jebsen-Taylor test of hand function) (ES 0.52–0.66), more responsive than measures of impairment (pinch and grip strength) but less responsive than the SIS-Hand (ES = 0.55–1.018) in the first 6 months post-stroke [38].

Oxford Handicap Scale

The Oxford Handicap Scale (OHS) is a simple tool modified from the Rankin Scale to grade the ability of a person and the level of daily assistance required to live independently [181]. Two studies evaluated the psychometric properties of the OHS, both including participants less than 6 months post-stroke [144, 152]. Neither study specifically identified participants to have upper limb spasticity.

Content validity. No published information regarding the development nor content validity of the OHS was located.

Results for whole sample. The OHS was not a predictor of caregiver burden [144] but was found to predict both the number of services and amount of time required from services on discharge [152].

Rivermead Motor Assessment

The Rivermead Motor Assessment (RMA) [117] is comprised of three sections; for this review studies were separated into two categories 1) ‘RMA’ all three sections (upper limb, trunk and leg) administered and reported and 2) ‘RMA UL’ upper limb section of the RMA only administered and reported. A total of 7 studies were included [25, 26, 48, 117, 129, 147], all studies included participants post-stroke, 4 of the 7 studies included participants less than 6 months post-stroke [26, 48, 101, 147]. When separated into the two categories, evidence for the ‘complete RMA’ was drawn from 5 studies [25, 26, 101, 117, 147] and evidence for the ‘RMA UL’ section was drawn from 6 studies [25, 26, 48, 117, 129, 147].

Content validity. Test authors Lincoln and Leadbitter [117] detail the measurement tool development. This was completed via selecting a preliminary series of items ranging widely in difficulty ordered into the three sections; gross, leg and trunk and arm. All individual sections were found to have mixed results regarding relevance, reduced due to methods used to create items and nil information regarding comprehensiveness nor comprehensibility.

Results for whole sample. The hierarchical scale of the RMA in an acute and non-acute stroke sample found varying results. Evidence to support the scalability of the RMA was found for the gross function and arm section in acute stroke only [26]. Scalability was supported in the gross function section only, when used with participants 6 and 12 months post-stroke [25, 147]. The RMA correlated with ADL performance (r = 0.51) and balance (r = -0.45) [147], a related construct. Agreement between clinician and participants predicted scores with achieved scores was found (clinician ICC 0.965 Bland Altman 96.6; participants ICC 0.908 Bland Altman 79.3) [101]. The hierarchical scale of the RMA UL section was supported only when administered to participants in the acute phase post-stroke (Guttman scaling criteria met) [26], the scalability criteria was not met when used with participants 6 and 12 months post-stroke [25]. The UL section of the RMA was found to correlate strongly with measures of upper limb activity at 6, 12 and 18 weeks post stroke (r = Rho 0.73–0.76) [48] and greater than six months post stroke (r = - 0.80) [129]. The RMA UL correlated moderately with perceived physical activity (r = -0.47) and did not predict overall HRQoL [129].

Stroke-Adapted Version of the Sickness Impact Profile

The Stroke-Adapted Version of the Sickness Impact Profile (SA-SIP30) was derived from the original Sickness Impact Profile and contains the following 8 subscales: body care and movement, mobility, ambulation, social interaction, emotional behavior, alertness behavior, communication and household management [160]. Four studies evaluated the psychometric properties of the SA-SIP30 [58, 69, 148, 160], all involved participants post-stroke, and only one study specifically identified participants with upper limb spasticity [58].

Content validity. Test authors detailed the methodology applied to create the SA-SIP, based on statistical relevancy and homogeneity [160]. The scale was found to be relevant, however to lack comprehensiveness (as only 5 of 9 recommended dimensions for patient-based, health related quality of life measures were included) [148]. No information regarding comprehensibility was provided.

Results for whole sample. The SA-SIP accounted for 53% of variance in predicting participation (R² = 0.63, P<0.001) and was more sensitive to detecting stroke related changes impacting on independence at 6 months post-stroke [69].

Results pertaining to sample with upper limb spasticity. The SA-SIP30 was significantly associated with greater disability in hygiene, dressing, limb posture and pain (P < .05) [58].

Stroke Impact Scale

The Stroke Impact Scale (SIS) is a stroke-specific measure of global health outcome [64] and comprises of eight domains: strength, hand function, activities of daily living, instrumental activities of daily living, mobility, communication, emotion, memory and thinking, and participation. The SIS was found to be reported as either individual or collective domains which are administered and reported separately. To maintain consistency across all measures within this review, the SIS was required to be administered in full and in the form of version 3 to meet inclusion criteria. Ten studies evaluated the psychometric properties of version 3 of the SIS [64–66, 68, 71, 99, 110, 112, 148, 170], all included participants post-stroke and none specifically identified participants with upper limb spasticity.

Content validity. The SIS was originally developed following a comprehensive iterative process with the use of participants, caregivers and standardized instrument development guidelines implemented but specific details are not available (unpublished information) [68]. Rasch analysis led to revision of the measure [64] demonstrating comprehensiveness (containing 7 of 9 recommended dimensions for patient-based, health related quality of life) and to be more comprehensive than EQ-5D, SA-SIP and SF-36 [148].

Results for whole sample. Rasch analysis refined the SIS into version 3 producing unidimensional domains ranging in item difficulty and with the ability to discriminate [64]. A single index was proposed, aggregated from the 8 domains (α = 0.93) accounting for 68.76% of the variance [99]. These 8 domains were each found to be internally consistent (α ≥ 0.86–0.96) [66, 99], suggesting possible item redundancy and further investigations of shorter forms. Agreement between patient and proxy ratings were fair to excellent, being stronger in the observable physical domains (ICC 0.50 to 0.83) [65]. The tool was reliable between testing sessions when administered via mail (ICC 0.77–0.99) and telephone modes (ICC 0.90–0.99) [66]. The individual and related domains of the SIS were found to correlate with global measures of independence, activity and participation, both patient and proxy reported, (r = 0.69–0.78) [65, 110, 170]. The SIS was able to discriminate between participants deemed recovered by the BI [112] and held superior ability to discriminate between varying levels of disability compared to the FIM and SF-36V (modified version of the SF-36) when tools were administered via phone [110]. Floor and ceiling effects were varied ranging from nil floor effect and 0–32% ceiling effect [71, 110].

Upper-Limb Motor Assessment Scale

The Upper Limb -Motor Assessment Scale (UL-MAS) is a subscale of items 6, 7 and 8 of the Motor Assessment Scale, and it provides a task orientated performance-based measure of upper limb activity [46]. Ten studies evaluating the psychometric properties of the UL-MAS were included [46, 100, 103, 109, 116, 118, 119, 126, 137, 146], all involved participants less than 6 months post-stroke, and no studies specifically identified participants with upper limb spasticity.

Content validity. Evidence located for the development of the MAS and subsequent UL-MAS did not indicate participants were consulted on the comprehensiveness or comprehensibility of included items [46]. Relevance of items for the intended purpose was sufficient.

Results for whole sample. There was evidence to support the production of a single composite score from the UL-MAS items, which may be interpreted as a total score for UL function [116]. Inconsistencies were identified within the hierarchical scoring [126, 137, 146] with clinical recommendations to attempt and score every item [126]. Furthermore, task 2 within the Hand Movements item may not be indicative of upper limb motor recovery in adults aged 65 years and older [126]. The UL-MAS is a unidimensional scale measuring a single construct, upper limb motor performance, (α = 0.83 to 0.95, and with removal of wrist deviation 0.93) [100, 116, 126]. It was reliable between (Kendall Tau = 0.74–1.00) and amongst assessors (kappa 0.93–1.0, 88–85% agreement) [46, 118]. The UL-MAS was able to discriminate between differing levels of motor recovery both in the acute and subacute phase, with Rasch based scoring more precise [103]. Varying levels of floor and ceiling effects have been reported for the UL-MAS (floor effect 0–38%, ceiling effect 0–67%) [126, 137, 146].