Sit to stand and chair transfer assessment and prognosis after stroke

Executive interpretation

Chair rising is a clinically important activity in its own right. A person may regain an independent transfer while retaining substantial paresis, loading asymmetry or compensatory trunk movement; another may improve strength without being able to rise safely. Accordingly, the most defensible assessment package records three separate things: whether and how the person transfers; the time or repetitions achieved under a reproducible protocol; and, when needed, the mechanism of performance measured with appropriate instrumentation. These are complementary outcomes rather than interchangeable expressions of “leg power.”

For people who can complete repeated transfers safely, the five-times sit-to-stand test (5STS) is a useful brief performance measure. However, an influential small chronic-stroke reliability study partly evaluated repeated timing of the same videos; its near-perfect coefficients should not be represented as proof of negligible day-to-day patient variation. Thirty-second chair-stand testing has direct acute-stroke reliability evidence, but the principal study pooled a 43-cm, arms-crossed version with an arm-assisted 46-cm modification. Stroke-specific minimal important change estimates exist for 5STS, yet they depend on walking-recovery anchors, recovery stage, walking subgroup and the handling of patients unable to complete the task. There is no defensible single change threshold for every person after stroke. [1–4]

The prognostic literature is more substantial than a claim of “no evidence” would suggest, but less mature than a clinical calculator would require. Prospective work connects 5STS to subsequent falls and community-walking status, and smaller instrumented cohorts examine future falls. The strongest established distinction is between association or development-sample discrimination and an externally validated prediction rule. A one-year falls cohort found only modest discrimination for 5STS, with an area under the curve (AUC) of 0.61. Therefore, a slow chair-rise result can prompt broader assessment; a fast result cannot clear a patient of falls risk. [5–8]

The practical priority in acute and severe stroke is to avoid losing the very patients for whom transfer assessment matters most. “Unable,” unsafe, physically assisted, arm-assisted and untested for medical or communication reasons must remain visible. Giving inability an arbitrary long time may simplify statistical analysis but does not turn it into an observed performance. A separate assistance or capability outcome is usually more interpretable than a single continuous score spanning independent completers and non-completers.

Scope search approach and review landscape

Questions and boundaries

This is a critical narrative synthesis of stroke-specific assessment and prognosis, with an auditable bounded search, rather than a registered systematic review. It addresses single sit-to-stand, five-repetition testing, fixed-duration chair stands, supported or assisted transfers, stand-to-sit, and selected instrumented methods. Sit-to-walk is included where the transition changes what is measured; walking tests, floor transfers and whole mobility scales are discussed only where necessary to clarify boundaries. Intervention trials provide context about change and feasibility, but use of a test in a trial is not itself evidence of reliability, validity, responsiveness or prognostic accuracy.

The population is adults living with stroke. Evidence from Parkinson’s disease, healthy older adults, cerebral palsy, cardiopulmonary disease or mixed neurological samples is not imported as a stroke threshold. Time since stroke is reported in the terms and units used by individual studies because labels such as “subacute” are not always used consistently. Stroke type, residual severity, capacity to stand, ability to walk, cognition and communication exclusions are treated as modifiers of applicability, not background demographics that can be ignored.

Search coverage and reconciliation

Two native Scopus searches covered (i) measurement properties and prognosis and (ii) protocol, repeated-task, biomechanical and technology terms. All eight and six pages were archived respectively, yielding 187/187 and 137/137 unique source records. These are complete results for the specified bounded queries, not a claim that all stroke-transfer literature is indexed or retrieved. DOI/title checks, public publisher and repository searches, review-reference inspection and limited forward/backward citation chasing supplemented these sets. Citation-network calls were capped discovery routes and are not described as complete citation searches. Exact strings and reconciled counts are summarised in Appendix A.

Sources were prioritised for direct relevance and decision value: original stroke measurement studies; explicit longitudinal outcomes; original protocol-manipulation experiments; and primary validation against an appropriate reference method. Reviews were used to map the field and locate sources, not to upgrade their underlying studies. Abstract-only information is labelled and is not used to fill missing methods. There was no independent duplicate screening, pooled effect estimate or formal certainty-grade procedure for this narrative report. Selected papers undergo domain-based appraisal of sampling, protocol, reference standard, statistics, follow-up, validation and practical transportability.

What earlier reviews establish

Silva and colleagues' neurological-disease review found that five-repetition testing was the most widely investigated clinical sit-to-stand measure, while identifying substantial protocol heterogeneity and then-limited responsiveness evidence. Its mixed-disease conclusions cannot be read as if every finding applied to stroke. Original-source checking also matters: its summary of the Mong protocol describes crossed arms, whereas the primary paper specifies hands on the lap without using them. The primary method takes precedence. [1, 9]

Boukadida and colleagues' stroke-focused narrative review synthesised 46 papers after a MEDLINE search and reference chasing. It provides a useful account of interacting strength, postural-control, trunk-motion and loading determinants, but it was not a pooled prognostic analysis. Onursal Kılınç and colleagues' later systematic biomechanics review included 21 comparative studies, 450 stroke participants and 329 controls. Fifteen studies were chronic, three subacute, one acute, one mixed and one unspecified; functional status was unreported in nine. The included literature largely compared selected independent performers with controls. Heterogeneity in chair setup, phases and outcomes prevented meta-analysis. This supports a multidimensional account of the task, while warning against a universal “stroke movement signature.” [10, 11]

The 2014 Cochrane intervention review and 2025 European Stroke Organisation motor-rehabilitation guideline address training effects. They support the relevance of practising functional tasks, with important limitations in transfer-independence and falls evidence, but do not validate a particular timer, app-derived power equation or individual prognostic threshold. Similarly, consensus measurement recommendations distinguish lower-limb impairment, balance and mobility domains; the need for a common outcome does not erase the distinction between a task-performance score and the physiological quantity it may correlate with. [12–14]

What a chair rise result means after stroke

Capacity independence movement strategy and everyday performance

An independent transfer is an activity-level capability. Completion time quantifies the speed of a specified task. Repetition count adds sustained performance, pacing and fatigue exposure. A force platform measures external loading, while motion capture describes movement and can support model-based kinetic calculations. None of these directly measures all the others.

A person may become faster by shifting demand toward the less-affected limb, using a different foot position, producing more trunk momentum or using arm support. That may be a useful improvement in daily independence even if movement becomes less symmetric. Conversely, a more symmetric trial may be slower or require more effort. The clinical interpretation must therefore state the goal: restoring a movement component, improving safe independent transfer, or increasing everyday mobility. Declaring a transfer “normalised” from time alone overstates what was observed. [10, 11, 15, 16]

Capacity in a clinic also differs from performance in daily life. The number of daily transfers is influenced by care routines, chair availability, social activity and opportunities to move. More transfers might reflect greater activity, but also more fragmented sedentary time. A laboratory maximum-repetition test does not establish how often someone transfers at home, and a daily transition count does not establish the safety or quality of each transfer. Janssen's longitudinal work is particularly valuable because it measured these domains separately. [16, 17]

Both lower limbs matter but asymmetry needs its own denominator

The less-affected limb should not be treated as a normal control by definition. It can be weak or deconditioned and can compensate disproportionately. Whole-task speed may improve through bilateral change, whereas a ratio can improve because the stronger side deteriorates. An asymmetry measure must therefore be accompanied by the raw values from each side and an explicit equation.

Affected/unaffected peak force, affected force as a percentage of total bilateral force, difference divided by the mean, and absolute percentage asymmetry are different quantities. Ratios also depend on whether peaks occur simultaneously. Dividing two separate limb maxima is not necessarily the load split at seat-off. The task phase, signed versus absolute direction and reference denominator belong in the report. A target of perfect symmetry is not justified solely by a group difference from controls. [15, 18–20]

Failure and assistance are outcomes not inconvenient missing data

At the lower end of function, a stopwatch can be a poor primary instrument. The person who cannot leave the seat, the person who reaches standing with contact assistance, and the person who performs one independent repetition but cannot complete five have distinct capabilities. Record the number attempted and completed, the reason for stopping, the support provided and whether the failure concerned initiation, lift-off, extension, stabilisation or controlled descent.

A numeric ceiling such as 60 seconds creates a mixture of observed time and assigned status. This can be useful if explicitly prespecified, but changes crossing the ceiling have a different interpretation from changes between two completed trials. The transition from “unable” to 25 seconds is clinically important; it is not an observed 35-second reduction in execution time. Analyses and clinical displays should show capability transitions separately. A missing test due to medical instability or inability to understand the instructions must not be coded as motor inability. [3, 16, 21]

Table 1 Match the measure to the question

Protocol is part of the measurement

Chair geometry and physical environment

Chair height changes the initial joint configuration and the mechanical demand. A fixed 43-cm chair offers a common external challenge, whereas a chair adjusted to knee or popliteal height offers a more standardised relative configuration. These choices answer related but different questions. A tall person and a short person tested on the same chair do not perform an identical relative-range task. Reporting “standard chair” is insufficient; record height, depth, backrest, armrests, cushion compression and whether the chair was secured.

Ng and colleagues manipulated seat height in 43 independent chronic-stroke participants. Lowering the seat to 85% of knee height lengthened 5STS time compared with 100% and 115%; 100% and 115% did not differ significantly. The two tested arm positions, crossed arms and hands on thighs, did not significantly change time in that selected sample. Pushing on the thighs was not restricted in the hands-on-thighs condition. This does not establish that arm assistance is unimportant among people unable to rise without support. It establishes a protocol effect among established completers. [22]

For serial care, use the same chair when possible and record its actual height. If a change of chair is necessary, do not interpret the resulting difference using the old protocol's error or change threshold without qualification. Wheelchair transfers, soft lounge-chair rises and toileting transfers may be more ecologically relevant but should be named as separate task conditions. Their value lies in the actual daily challenge, not in pretending they replicate a standardised 5STS.

Foot position upper limbs and affected side loading

Both feet should be visible and their position reproducible. Specify whether feet are self-selected or prescribed, the distance between them, their fore-aft relation, and whether the affected foot is behind, level with or ahead of the other. If posterior placement is used, quantify the displacement rather than saying “feet back.” Observe heel lift, foot repositioning, orthosis use and loss of contact.

Kwong and colleagues tested 45 chronic-stroke survivors who could rise without external support. Moving both heels 10 cm posteriorly shortened 5STS time across three arm conditions. Hands-on-thighs also differed from an augmented-arm condition. Mean times across the six conditions ranged from 15.2 to 17.1 seconds. A change of setup can therefore create a change comparable with some published important-change estimates. The study does not provide a correction factor transferable to another patient; it demonstrates why foot and arm conditions must be fixed or deliberately analysed. [23]

Hand placement and hand use require separate entries. Resting hands on the thighs is not equivalent to pushing through them. Arms crossed, hands on lap without assistance, pushing on armrests and pulling on a walking frame all alter the task or permit different compensations. An affected arm in a sling can affect balance and movement. A support used only for balance still changes the test and should not disappear into a generic “independent” label. Changes in assistance should be celebrated when relevant but should not be combined with the timed change as though protocol were unchanged.

Timing repetition practice and stopping rules

Start rules include an external verbal command, the first visible body movement and the back leaving the backrest. End rules include the fifth full stand, the fifth return of the buttocks to the seat and the fifth contact with the backrest. They differ in reaction-time content and in whether the last descent is included. The test name alone does not resolve them. Record the rule in ordinary language and, for software, define the event algorithm and visual verification procedure.

The number of practice trials and scored trials also matters. A mean of three trials, best of three, first trial and median of three are not interchangeable. Best performance describes maximum observed capacity and may be optimistic; a mean reduces some random error but adds burden; a median resists a single unusual trial. Acute-stroke repeated testing can produce learning or fatigue within a session. Standardise rest intervals, instructions, encouragement and speed instruction, and record any deviation. [1, 2, 15, 22, 23]

Before testing, confirm medical stability and the level of guarding required. The protocol should state when pain, dizziness, symptoms, unsafe loss of balance, inability to continue or participant request stops the task. Do not force repeated maximal transfers to obtain a numeric score. Distinguish a safety stop from a completed slow trial and from inability. Remote testing additionally needs an appropriate chair, clear floor space, visible whole-body movement, a communication and emergency plan, and an appropriate adult helper when indicated by the validated procedure. The absence of adverse events in a small selected study is not proof of safety for unsupervised acute or severely impaired patients. [2, 21, 24, 25]

Table 2 Protocol details that change interpretation

Clinical assessment what the original studies support

Five times sit to stand an established test name with several protocols

The 5STS is best described as timed repeated-transfer performance under specified conditions. It combines getting up, stabilising, returning to sitting and repeating the task. Its practical advantages are short duration, inexpensive equipment and an outcome clinicians and patients can understand. Its limitations include inability to complete, protocol sensitivity, variable contributions from balance and strength, and the difficulty of distinguishing restitution from compensation using total time alone.

Mong, Teo and Ng studied only 12 people with chronic stroke, alongside 12 healthy older and 12 young controls. All could rise without hand support; stroke participants could walk 10 m without physical assistance. The original protocol used a 43-cm chair, hands on lap without use, two familiarisation trials and the mean of three scored trials. Timing ran from the back leaving the backrest to the final return. Rater-related ICCs were very high, but the paper explicitly attributes some excellent repeatability to assessors viewing the same videos on both occasions. This is a narrower claim than day-to-day biological stability. The roughly 12-second cutoff separated stroke participants from healthy older controls in this sample; it was not a falls or independence threshold. [1]

This small study is frequently cited as a complete validation package. It is more appropriately treated as early support for standardised timing and known-groups discrimination in selected chronic-stroke completers. Precision and transportability require further evidence. A high ICC can coexist with clinically relevant absolute differences when between-person variability is large. Nor does reliability establish which impairment caused slow performance.

Construct validity does not identify one exclusive impairment

Ng's separate study of 68 community-dwelling chronic-stroke survivors illustrates how the interpretation depends on the sampled population and comparator. The 5STS correlated strongly with Berg Balance Scale (BBS) scores and moderately with a strength index and six-minute walk distance. After adjustment, BBS remained the significant independent correlate; the full model explained 71% of 5STS variance. The study was cross-sectional and selected people who could walk 10 m without physical assistance. It supports a substantial balance contribution in this sample, not the claim that strength is irrelevant or that 5STS can replace a comprehensive balance assessment. The BBS itself includes related transfer tasks, creating some content overlap. [28]

The apparent disagreement between knee-flexor associations in Mong's small study and the stronger balance association in Ng's larger study is clinically informative. Sample size, restricted ranges of ability, muscle-testing technique and the task protocol all affect correlations. It is unsafe to resolve this by labelling 5STS either a pure strength test or a pure balance test. A short total time does not rule out a unilateral weakness, and a slow time does not identify whether strength, confidence, coordination, pain, sensory loss or cautious stabilisation is the main constraint.

Mentiplay, Clark, Bower, Williams and Pua provide a particularly useful later comparison of strength and balance. This cross-sectional secondary analysis included 61 participants from hospitals in Australia and Singapore, with mean time since stroke 39 months. Participants had to walk at least 10 m without assistance or gait aids and have a non-cerebellar stroke at least three months earlier. Two of 63 recruited participants were excluded because they could not perform 5STS. The same assessor tested both sites, reducing one source of procedural variation, although the sites differed significantly in 5STS performance. [26]

The original used a 45-cm chair and one scored 5STS trial, with no hand use, complete standing and seated contact between repetitions. Timing began and ended seated with the back against the chair. Trials were repeated for arm use, incomplete standing or failure to sit between repetitions. This differs from Mong's 43-cm, multi-trial protocol and from studies assigning inability a 60-second ceiling. Observed times ranged from five to 61 seconds. Seven muscle groups on both sides were tested using handheld dynamometry, while static eyes-open balance was measured on a Wii Balance Board. Hip-extension data were missing for 12 participants unable to lie prone. [26]

Bilateral knee-extension strength added 17.9 percentage points of explained variance beyond the study's covariate model; the strongest balance variables added approximately 16.6–16.8. The model combining bilateral knee-extension strength with anteroposterior sway velocity had total R² 0.565, compared with the covariate base of 0.310. Thus the reported 25.5% is an increment beyond covariates, not the full model's total explained variance. Removing strength or balance reduced R² by 0.087 and 0.076 respectively, supporting an independent contribution from each under this model. [26]

The methodology helps explain why this result need not agree with a BBS-based comparison using a composite strength score. The authors entered both limbs for a specific muscle group and used posturography rather than a scale containing transfer items. Nevertheless, static sway captures only one balance component. Covariate handling combined sex, time since stroke and recruitment country into a principal component that retained 42% of their variance, so it should not be described as unrestricted adjustment for every detail of those variables. Selecting the best-performing strength and balance measures in this same sample also favours a cautious descriptive interpretation. The study supports a multifactorial task and directly challenges a pure-strength or pure-balance label; it neither measures chair-rise mechanical power nor shows that treating either impairment causes the predicted improvement.

Bohannon's study of 61 acute-rehabilitation patients examined current transfer independence against bilateral knee-extension force and body weight. Independence was more frequent when hand use was permitted than prohibited. This supports the importance of bilateral strength relative to the load that must be moved, but the reported force cut points classify contemporaneous independence. They do not establish future independence, and they are not chair-rise time thresholds. The functional task and impairment measurement should remain distinct in rehabilitation documentation. [29]

Between day 5STS error the newer spasticity stratified study

An and colleagues provide direct seven-day test-retest information in 54 chronic-stroke inpatients able to walk at least 10 m, stratified by ankle plantar-flexor Modified Ashworth Scale (MAS) score. The seat was adjusted to popliteal height to approximate 90-degree knee flexion; participants kept arms crossed and feet flat. ICC(2,1) values were 0.96, 0.98 and 0.96 for MAS 0, 1–1+ and at least 2 respectively. Corresponding SEMs were 1.64, 1.54 and 1.99 seconds, and MDC95 estimates were 4.56, 4.26 and 5.51 seconds. These values support good relative reproducibility while showing appreciable individual-level uncertainty. [4]

Several cautions matter. Subgroup sizes were only 16, 23 and 15; participants were selected ambulators with relatively preserved cognition. Ongoing inpatient rehabilitation over seven days can produce genuine change, so a retest difference need not be pure measurement noise. The primary description does not provide the level of timing-event detail found in the original Mong protocol. Most importantly, these are distribution-based detectable-change values, not patient-anchored important-change estimates.

The report also requires numerical care. Its broad statement that measurement error was below 20% should not replace inspection of the actual 5STS results. Dividing the reported MDCs by the corresponding mean 5STS times gives approximately 27%, 22% and 28%, not uniformly below 20%. These are reviewer calculations from the published tables, not new validated thresholds. Some demographic counts are internally inconsistent, further favouring cautious use of the exact tabulated numbers rather than a blanket reassurance. Do not select the smallest MDC from this paper for every patient or transfer it to a fixed-height, hand-assisted or remote test. [4]

Thirty second chair stands in hospitalised stroke

Lyders Johansen and colleagues tested hospitalised patients at a median five days after stroke. Sixty-two participants were included in the final overall analysis; the chair-stand reliability table contained 61. Patients attended two sessions separated by one hour on the same day, with two raters and standardised rests and instructions. The standard 30-second test used a 43-cm chair, arms crossed, knees around 90 degrees, feet approximately shoulder-width apart and no initial contact with the backrest. Participants were instructed to stand fully and sit fully between repetitions. [2]

The important protocol qualification is that 29% used a modification allowing armrests and a 46-cm seat. The pooled results therefore describe a pragmatic mixed-protocol cohort, not exclusively the original arms-crossed test. Intrarater ICCs were 0.87 and 0.91; interrater ICCs were 0.88 and 0.94. The reported smallest real difference for individual change was approximately three repetitions. A small learning effect occurred, with the first actual trial averaging 7.1 stands and subsequent trials around 7.7–7.8. A numerical increase can therefore reflect familiarisation as well as recovery. [2]

The study's acute setting is a strength, but eligibility required independent sitting and standing. Most participants could also walk, commonly with aids. It does not establish reliability or safety in all acute stroke, particularly those requiring physical lift assistance, unable to stand or unable to understand the protocol. Same-day repeatability is also not the same as stability over a week of rapid neurological and medical change.

Three repetitions is an error-based threshold for this study's procedures. The discussion speculates about clinical importance using osteoarthritis literature, but that speculation is not an anchored stroke MIC. Likewise, the authors' SEM95 terminology for groups should not be substituted for the standard error of a study's group mean without considering sample size and design. For an individual result, preserve the original/modified condition and interpret detectable change within the matching protocol. If a person changes from arm-assisted to unassisted testing, report the improvement in support requirement separately instead of directly subtracting counts as if nothing else changed.

When a timed test is not appropriate assistance and ordinal mobility

An assistance scale can represent clinically valuable change before repeated timing is feasible. The Cumulated Ambulation Score (CAS) includes bed transfer, chair sit-to-stand and indoor walking, each rated on a three-level scale. Arens and colleagues evaluated 60 patients at median day three after stroke, with an experienced clinician and an inexperienced but trained assessor observing the same performance. Total-score weighted kappa was 0.816; activity-level values ranged 0.733–0.904. Observed agreement was 85%–92% for the individual activities, but only 70% for the total. [21]

This is useful evidence for shared clinical grading of basic mobility, including people who need assistance. However, the total CAS is not a chair-rise-only score, and its total-score SEM of 0.37 and MDC of 1.03 points must not be relabelled as transfer-item properties. Simultaneous observation isolates between-rater scoring variation more than day-to-day patient variability. Since the scale is ordinal and bounded, a one-point difference has a practical category meaning; it should not be portrayed as a universal interval-scale quantity. The paper's recommendation around one point deserves particular care because the calculated total MDC is slightly above one.

For clinical transfer monitoring, pair an established assistance description with a brief explanation: physical lift, balance assistance, supervision, cueing, arm support or independent performance. A person who becomes independent with a raised chair may still need assistance from a low sofa. That is useful ecological information, not an inconsistency to conceal. Ordinal assistance and standardised performance are often both needed as ability improves.

Table 3 Selected clinical reliability and error estimates

Responsiveness and meaningful change

A hierarchy of change claims

Four statements should be distinguished. First, a measured number changed. Second, it changed more than expected measurement error. Third, the change is important to the patient or clinical decision. Fourth, the change was caused by a particular treatment. A single study rarely establishes all four. An intervention group improving on a task does not establish an MIC; exceeding an MDC does not establish that the change matters; and an important change in early stroke may reflect spontaneous recovery and multiple co-interventions.

Responsiveness is best evaluated against prespecified hypotheses and relevant external change measures. An effect size or significant pre/post difference can be influenced by the treatment and the sample's room for improvement. Anchored interpretation is stronger when the anchor concerns the same construct, correlates adequately with change and separates minimally important improvement from a large recovery. Confidence intervals, baseline dependence, false-positive/negative consequences and handling of non-completers should be inspected alongside the headline value.

The 2021 5STS responsiveness and MCID study

Martín-SanAgustín and colleagues followed patients in an outpatient rehabilitation programme at admission, four weeks and eight weeks. At first assessment, 111 patients were included, with mean time since stroke approximately 52 days; 44 could not complete the 5STS. The protocol used a 43-cm armless chair with a 47.5-cm seat depth and backrest-based timing. Inability to complete five repetitions within one minute was assigned 60 seconds. Severe cognitive/language limitations were excluded. People who already walked everywhere independently were excluded at the baseline of each stage, so the later sample was not simply the original cohort measured again without selection. [3]

The study used a patient global rating of change in walking ability, with scores above +5 identifying important improvement. It also examined changes against gait speed and Functional Ambulation Category. The chosen ROC-derived 5STS values were 1.18 seconds for all patients over the first four weeks and 0.76 seconds over the next four. For household ambulators, they were 1.90 and 0.72 seconds; for limited-community ambulators, 2.92 and 3.09 seconds. AUCs against the patient anchor were approximately 0.70–0.76. Change correlations with walking comparators were generally modest to moderate. [3]

These are genuine anchor-based estimates, but their meaning is more constrained than the label “the stroke 5STS MCID” suggests. The patient anchor was walking recovery, not perceived transfer improvement. The threshold distinguished a marked positive response rather than the smallest conceivable benefit. The household category included non-ambulators. Many patients began at an imputed 60-second value, producing a pile-up at the ceiling and large assigned improvements when completion first became possible. The authors explicitly acknowledged these distributional problems. The choice between two candidate methods used their association with the same anchor in the development data, rather than independent confirmation.

The stage-specific results should therefore be retained rather than averaged into one number. They may help interpret change in a broadly matching early outpatient setting, but do not establish a universal patient-level transfer MIC in chronic stroke or remotely administered tests. A 0.72-second decrease and a transition from inability to an independent transfer have fundamentally different clinical content even if both count as improvement. Also, an important-change estimate smaller than another study's MDC is not a contradiction: importance and detectability answer different questions, and both the population and protocol differ.

Table 4 5STS anchored change estimates require the original context

How to use these estimates without false precision

A defensible serial report first states the two observed results, their dates, the exact protocol and whether the task was completed independently on both occasions. Next, it compares the difference with an error estimate from the closest matching study, if one is available. Finally, it describes what changed in everyday transfer ability and whether the patient considers that useful. If no matching error or anchor estimate exists, report that uncertainty rather than borrowing a number from Parkinson’s disease, geriatric norms or an unrelated version of the task.

For example, a modest improvement in 5STS with unchanged assistance may warrant repeat testing and corroboration rather than a binary declaration of meaningful recovery. A reduction in physical assistance may be clearly important even though an independent 5STS remains impossible. A faster time achieved using newly introduced arm support is a valid result for the new supported condition, but it does not establish improved unsupported lower-limb capacity. The best electronic record can accommodate all three situations.

Remote administration and fixed duration extensions

Remote scoring is not automatically interchangeable with in person testing

Deshmukh and colleagues compared laboratory and teleassessment in 17 chronic-stroke participants who completed the study, with a mean nine years since stroke. The virtual setting required an adult present and suitable videoconferencing equipment. The 30-second chair stand showed ICC 0.885 (95% CI 0.712–0.957), but limits of agreement were approximately −3.69 to +3.93 repetitions. For 5STS, agreement was poor: ICC 0.148 (−0.340 to 0.575), with very wide time differences. The five-rise result was collected within the 30-second test rather than necessarily as an independently standardised maximal five-rise trial. [24]

These results demonstrate why an acceptable correlation or a nonsignificant mean difference does not ensure interchangeability for an individual. Even the better 30-second result permits differences of several repetitions, important when some error thresholds are around three. Home-chair geometry, camera view, instructions and the way a five-repetition result is extracted can matter. The study's extremely weak ROC results should not be promoted as falls thresholds; the sample was preliminary and the ROC exercise did not validate prospective prediction.

Da Silva and colleagues reported much better in-person versus teleassessment agreement for 5STS in 25 chronic-stroke participants, including moderate or severe lower-limb impairment, but still requiring independent short-distance walking. Upper-limb compensation was permitted when needed, and the test was repeated three times. The cross-mode ICC was 0.95. The near-perfect intra/inter-rater coefficients came from reviewing recorded assessments, so they principally support reproducibility of scoring the same performance. They do not cancel the conflicting cross-mode study or validate every home protocol. [25]

The appropriate conclusion is conditional feasibility. A service should select one documented procedure, train staff and helpers, verify the home setup, record supports and keep the mode of testing consistent where possible. Changing from clinic to home is a method change worth recording. Larger validation studies need prespecified acceptable limits of agreement and representation of aphasia, cognitive impairment, severe motor limitation and digital-access constraints.

Virtual 30 second responsiveness

Noguchi and colleagues performed a secondary analysis of two TRAIL telerehabilitation trials. Baseline construct and known-groups validity used 67 participants, whereas responsiveness used 32 with relevant post-intervention data from the proof-of-concept study. Mean time since stroke was 9.3 months. Participants had to walk 10 m without physical assistance, tolerate 50 minutes of activity, have adequate cognitive-communication capacity and have a helper. This is more representative than a very small feasibility series, but it remains a selected ambulatory group with access to remote rehabilitation. [27]

The protocol required crossed arms, feet flat, full standing and a return to sitting with backrest contact for each complete repetition. Helpers stabilised the chair and supervised safety without providing movement assistance or encouragement. Home-chair height varied between participants, but each used the same chair at serial assessments. Timing was performed by both assessor and helper. These details are important: this was not an arbitrary webcam count, and a count of full stand-and-sit cycles may differ from a protocol crediting a last partial cycle.

The authors reported 14/17 construct-validity hypotheses confirmed (82%) and 12/15 responsiveness hypotheses confirmed (80%). Baseline correlations were stronger with TUG (−0.64) and balance confidence (0.54) than with self-reported strength (0.32). Change correlations were generally small, with the largest 0.35 for the Stroke Impact Scale strength domain; TUG and Functional Reach changes were approximately 0.33 in absolute magnitude. Known-groups differences were 4.0 repetitions by modified Rankin grouping and 3.5 by NIHSS grouping. These results support a multifactorial functional measure rather than direct measurement of muscle force. [27]

A source-level inconsistency limits confidence in the reported responsiveness percentage. In the original Table 3 on journal page 237, hypotheses 13 and 14 predict that the FM-tele correlation will exceed the FRT and cognition correlations. The printed comparisons are 0.18 versus 0.33 and 0.18 versus 0.20, yet both are marked confirmed. A literal recount of the printed comparisons would be 10/15, not 12/15, and would not meet the study's stated criterion of more than 75%. This is an audit of the published table, not a reanalysis of raw data; a typographical or labelling error could explain it. The report therefore preserves the authors' result while treating the categorical conclusion of acceptable responsiveness as unresolved pending clarification. This does not erase the observed correlations or known-groups findings.

This study does not compare remote with in-person 30-second testing, derive a stroke-specific anchored MIC, or establish predictive validity for falls. Its use of an earlier virtual MDC of 3.1 repetitions and osteoarthritis MICs to formulate hypotheses should not be turned into a newly validated stroke-important-change threshold. [27]

Longer timed repetition tests and cardiorespiratory inference

Thirty-, forty-five- and sixty-second tests impose different pacing and fatigue demands. Evidence cannot be transferred merely because all require repeated chair rises. The bounded search did not identify a mature stroke-specific validation package for a one-minute test covering reliability, agreement, important change and prospective outcomes. This is a scoped evidence gap, not proof that no stroke paper has ever used the test. One-minute results from COPD, heart failure or cerebral palsy do not resolve it.

An instructive direct stroke study is Machado and colleagues' 45-second test coupled to a heart-rate prediction equation. In 30 participants, predicted mean oxygen consumption was 36.2 mL/kg/min versus 28.0 measured during a graded exercise test; concordance was only 0.11 (95% CI −0.07 to 0.28). This is evidence against that particular cardiorespiratory estimate in this population. It does not invalidate counting repetitions as a functional outcome, but it prevents advertising the equation as a substitute for direct fitness testing. [30]

The same principle applies to three-repetition or ten-repetition versions and to minimum-seat-height challenges. They may solve a practical problem, but each changes the construct or range. Name the modification, retain the raw result and avoid attaching 5STS error values or risk cutoffs unless that exact adaptation has been validated in a matching stroke population.

Instrumented assessment useful detail with measurement specific claims

What additional instrumentation can resolve

Instrumented testing is most useful when a clinical question cannot be answered by completion status or total time. Examples include whether the affected limb contributes at seat-off, whether extension is delayed after forward trunk movement, whether a faster transfer masks greater unilateral compensation, or whether descent becomes uncontrolled. These questions call for specific variables, not a large undifferentiated sensor score.

The measurement chain must remain visible. A force plate measures external force and moment; a pressure mat measures pressure distribution subject to its sensor design; an IMU measures acceleration and angular velocity; video or motion capture estimates positions of anatomical or modelled points. Joint torque, whole-body centre of mass (CoM), work and power then require models, transformations or integration. In some cases these estimates are strong and clinically useful, but they are not raw measurements. The validation must match the final output, population and task rather than the sensor's advertised precision alone.

Force platform loading and genuine between day reliability

Eng and Chu studied 15 chronic-stroke participants who could rise from a chair and stand independently. Vertical force at seat-off was assessed with separate plates under each foot and a third under the chair. Five trials were averaged, with barefoot foot position and thigh support standardised. Between-day ICCs for chair-rise loading were 0.99 for the affected limb and 0.97 for the other limb, with SEMs of 8.0 and 16.2 N. This is direct support for a tightly controlled force-at-seat-off outcome, not for every force-plate variable or for an individual single trial. [18]

The study is small and selected, but its design explicitly includes variation in patient performance across separate days, making it different from retiming a recording. The measurement is loading at a defined event, not maximal muscle strength. A low affected-side force could reflect capacity, strategy, pain, sensory integration or support choices. Relating such a result to an impairment requires additional examination. Force normalisation also needs attention: the study expressed bilateral-task loading relative to half of body weight in some analyses, which must not be confused with percentage of total body weight.

Low cost dual board measurement the Bower study

Bower, McGinley, Miller and Clark tested 30 outpatients more than three months after stroke on two occasions one week apart. The participants could stand unsupported and all could walk with or without aids. The dual Wii Balance Board sit-to-stand condition yielded paired reliability data in only 23, because some could not complete the task or lacked time. This selective loss is important: the most impaired participants can disappear from the apparently excellent measurement-property result. [15]

The supplementary protocol is unusually useful. Heels were 17 cm apart, toe-out angle 14 degrees, and seat position was adjusted to 75-degree knee flexion and 5–10-degree ankle dorsiflexion. Participants stood at a self-selected pace without upper-limb support for three trials. A nearby chair on the less-affected side was available for safety; touching it or needing physical assistance made a trial unsuccessful. The median of three trials was analysed. Board calibration, 40-Hz sampling and a 12-Hz low-pass filter formed part of the measurement system. These conditions differ materially from a rapid fixed-height 5STS. [15]

For the affected limb, peak force had ICC 0.90, SEM 2.21 percentage points of body weight and MDC 6.14 percentage points. Peak rate of force development (RFD) had ICC 0.94, SEM 30.5 and MDC 84.53 percentage points of body weight per second. Affected/unaffected peak-force ratio had ICC 0.96 and MDC 0.12; the corresponding peak-RFD ratio had ICC 0.89 and MDC 0.19. These are endpoint-specific results. An ICC range for the entire battery should not be used as if it applied to a new app's total score. [15]

The study did not compare the stroke sit-to-stand outputs against another force platform. Earlier technology-validation studies provide context, but they do not turn this into direct criterion validation of every dynamic endpoint in stroke. Moreover, the selected sit-to-stand force variables were not significantly correlated with the clinical mobility/balance tests examined. That does not prove they are invalid: they may measure a different aspect. Equally, it does not prove they add clinically useful information, respond to important recovery or predict future falls. Those require separate tests.

Raw units are safer than an unqualified percentage-change label. The paper's tabulated peak-force-ratio MDC is 0.12, but its accompanying percentage cannot be straightforwardly reproduced by dividing 0.12 by the reported mean ratios. Accordingly, the raw ratio MDC is retained and the percentage is not operationalised as an app threshold. This does not undermine the whole study; it prevents one ambiguous derived number from becoming a universal clinical rule.

Force RFD work and power are different outputs

A force trace alone is not power. External mechanical power requires a force and a velocity belonging to the same mechanical system; joint power requires joint moment and angular velocity. A rate of increase of force is RFD, expressed in force per time, not watts. Integrating force to estimate acceleration and velocity requires body mass, gravity removal, valid initial conditions and accounting for external support. Before seat-off, the chair contributes external force; with arm assistance, unmeasured forces may pass through armrests, a walking frame or another person.

Equations based on body mass, estimated vertical displacement and time can produce a useful estimate of average external work rate. However, the result depends on the assumed CoM trajectory and on how much of the timed test actually represents upward work. Five full cycles include descent and pauses; substituting the entire completion time into an equation intended for rise time changes its meaning. A power estimate must identify whether it concerns the whole person, one limb, a joint or a movement phase, and whether it is absolute or normalised to mass.

Two examples prevent common mislabelling. Janssen's “power chair stand up” is the inverse of rise duration, in seconds to the power of minus one. It is a rising-speed index, not measured mechanical power. Fujita and colleagues' “new test for muscle power” evaluates a nine-second modified Wingate cycling test; its association with 5STS offers construct-related information but does not validate the stopwatch as a wattmeter. Neither paper supports a generic stroke chair-rise power calculator. [16, 31]

Smartphone and IMU outputs separate timing from derived biomechanics

Merchán-Baeza and colleagues' acute-stage smartphone study involved eight participants. Its abstract reports rater reliability for derived trunk-motion variables, providing feasibility evidence rather than a broad validation of all mobile assessments. The small sample and limited access to complete methods restrict detailed protocol or error claims. It should not carry a marketed assertion that smartphone sit-to-stand measurement is validated in acute stroke generally. [32]

Sánchez-Sánchez and colleagues provide a larger, fully accessible example: 36 chronic-stroke participants and 33 controls completed a multi-phase mobility procedure using an Android device. The study examined known-groups differences and correlations with clinical measures. Its “sit-to-stand power” was estimated using a modelled CoM trajectory, participant characteristics and movement time. The study did not directly validate that power output against force-plate/motion-capture mechanical power in the stroke participants. Indeed, the estimated sit-to-stand power did not significantly distinguish the stroke and control groups and was not significantly associated with 5STS time in the stroke group. [33]

This is not a reason to discard the device. Several other kinematic or temporal outputs distinguished groups or related to mobility, supporting specific construct hypotheses. It is a reason to label the quantity as estimated and to avoid presenting correlation-based construct validation as agreement with a physical reference. Software version, sensor location, sampling, filtering, segmentation, body-mass input and treatment of atypical movement must remain part of the protocol. A sensor attached to the lower back, sternum or thigh is not sampling the same motion, particularly when trunk compensation is substantial.

For individual monitoring, an algorithm needs more than healthy-control separation. Relevant evidence includes agreement for event timing, between-day reliability of derived values, robustness to chair/foot/arm conditions, missing-trial rates and change beyond technical and biological variation. A model may perform well on clean independent rises yet fail precisely when a patient rocks repeatedly, pauses, uses an arm, fails to rise or is physically assisted. Those events should be surfaced for review rather than automatically converted into a confident score.

Markerless video a promising method with model dependence

Majoni and colleagues' 2026 study compared simultaneous markerless and marker-based CoM estimates in 17 people with subacute or chronic stroke across several balance and mobility tasks, including sit-to-stand. It used calibrated multi-camera systems, a safety harness and specified processing pipelines. Two markerless models were considered: a default 17-segment model and an 11-segment “fit” model matching the reference's segmentation and anthropometry. [34]

Matching the models reduced several systematic differences, particularly in the superior–inferior direction. Across tasks, the default model had vertical RMS differences around 4.27–6.16 cm, whereas the matched model's range was approximately 0.97–1.38 cm. The reported high waveform correlations are therefore not sufficient on their own; anatomical model choice changes the absolute result. These ranges refer to the tested task set and are not a stand-alone error threshold for chair rise. [34]

The comparator is itself a model-based CoM estimate, not direct observation of every segment's true mass distribution. The study selected tasks with generally visible full-body capture and limited occlusion. It does not validate a single casual phone video, every pose-estimation algorithm, transfer assistance behind an occluding caregiver, or camera-derived limb forces. Nonetheless, it provides meaningful primary evidence that carefully configured markerless measurement can be feasible in stroke, with explicit method-dependent uncertainty. Product claims should name the demonstrated camera/model/task combination.

Observational movement quality and stand to sit

A stopwatch can hide the descent strategy. Stand-to-sit requires control of lowering, alignment with the chair and termination of movement; a person may achieve a fast cycle by dropping heavily into the seat. For a test intended to capture safe transfer ability, full sitting, stable contact and the assistance required during descent deserve observation even if they are not part of the numeric score.

Hou and colleagues compared 13 people with stroke and 13 controls using lower-limb kinematics, kinetics, ground reaction force and centre of pressure during both directions of transfer. They demonstrated direction- and joint-specific asymmetry. This is evidence for analysing descent separately, but not a normative “abnormal asymmetry” threshold or a prospective risk score. The small sample consisted of participants able to complete the tasks; estimates from many joints/phases should not be treated as independently replicated clinical findings. [19]

Mao and colleagues' subacute biomechanical study likewise reveals information concealed by total duration: phase timing and affected/unaffected knee moments differed in 25 participants compared with 17 controls. However, eligibility required repeated independent rises and high cognitive screening scores. The work describes an important selected subgroup, not the full spectrum of early stroke transfer failure. It also uses a particular multi-phase definition that cannot simply be pooled with two-phase timing from another laboratory. [20]

The 2026 Movement Compensatory Screening (MoCS) framework includes chair-transfer domains within a broader 65-item observational instrument. The accessible abstract supports content validity through expert review. Content validity is a useful first step, but does not establish inter-rater agreement, detectable change, responsiveness, causal interpretation or future prognosis. A clinician may document visible strategy without implying that an emerging observational score is already validated for serial or prognostic use. [35]

Assistance itself can change the movement being measured. Burnfield and colleagues compared four clinician/device-assisted transfer conditions in ten patients with recent stroke. The original abstract reports that device-assisted transfers took nearly twice as long as clinician-assisted transfers and constrained aspects of trunk and ankle motion; clinician encouragement increased lower-limb activation during device assistance. One device and one clinician were studied, so these findings cannot be generalised to every lifting aid. They nevertheless illustrate why assisted duration and EMG cannot be interpreted as if they represented an independent rise. This is a within-person mechanical comparison, not evidence that one assistance method causes better recovery or predicts future independence. Complete protocol details were not available, so no device-specific prescription follows. [36]

Daily life transition monitoring

Activity monitors can address the gap between what someone can do in a test and what they actually do. The AMoR study compared sensor classification with video during simulated home activities in 21 chronic-stroke participants. Its abstract reports good transition-count agreement, including an ICC of 0.859 and mean absolute percentage error of 7.13% for sit-to-stand. This is encouraging for event detection, but it validates a particular simulated protocol and algorithm, not free-living transfer quality, independence or fall risk. [17]

Daily-life monitoring also requires a definition of the event: full seat-to-stand, a failed attempt, sit-to-walk, a brief perch, or an assisted transfer. Sensor wear time, missed slow movements, repeated rocking, classification minimum-duration rules and days of observation affect counts. Reporting “15 transitions” without valid observation time and event rules is incomplete. An app should retain uncertainty flags and distinguish a detected event from a clinically successful, independent and safe transfer.

Table 5 Instrumented outputs strongest claim and missing step

Sit to walk is a related but different transition

Why movement intention changes the measurement

In stand-alone sit-to-stand, the intended endpoint is stable upright stance. In sit-to-walk, forward motion can continue into stepping. Pausing after standing can therefore be an adaptation or task instruction rather than an inherently pathological interruption. Foot choice, walking-aid use and whether the person turns immediately after rising also change the transition. A rise embedded in TUG is not automatically interchangeable with a stand-alone rise.

Chandler and colleagues analysed early-stroke sit-to-stand, sit-to-walk and walking. Only 48 of the original 105 trial participants contributed to the relevant task analysis, after attendance and performance selection; the analysed group was around 64 days after stroke. Sit-to-stand and sit-to-walk were performed without human assistance or a walking aid, while temporary self-steadying with the arms was retained. Three participants could not walk 3 m unaided and were assigned walking speed 0. The tasks showed distinct movement characteristics, and relationships with walking speed were not strong. This supports task specificity and cautions against inferring walking quality from chair-rise timing or symmetry alone. It also exposes selection that a simple label of “early stroke” would conceal. [37]

Event detection can fail systematically after stroke

Jones and colleagues tested four gait-initiation-onset methods in 20 ambulatory stroke survivors and 21 controls. The stroke participants could rise without their arms and walk without an aid. A mediolateral-force threshold failed to identify onset in 48% of their sit-to-walk trials; five participants had no trial successfully identified by that method. Alternative peak-force methods detected events more consistently, but within-person single-trial reliability remained poor to moderate, with stroke ICCs around 0.41–0.50. [38]

The failure mode matters more than the sophistication of the algorithm. A threshold developed around healthier movement may preferentially exclude atypical stroke strategies, making the remaining sample appear more orderly. The authors recommend averaged repeated measures for the more usable methods. That does not validate a single-trial home measure or a fall-prediction threshold. When a platform reports a transfer-to-gait transition, it should expose event definitions, detection failures and the number of usable trials, rather than hiding them inside a total score.

Recovery and prognosis separate the outcome being predicted

Three different prognostic questions

“Prognosis” can mean the future time needed to perform the same test, future transfer independence, or a different outcome such as walking or falls. These questions have different evidential requirements. Baseline 5STS predicting later 5STS is not automatically a prediction of independence. A concurrent relationship with Barthel Index or walking speed is not longitudinal prediction. An association with past falls is a retrospective classifier even when the paper uses the language of risk.

A usable prognostic model should specify the population, baseline setting, exact test procedure, outcome definition, time horizon, handling of inability and loss to follow-up, predictor selection, calibration and validation. A high development AUC or percentage correctly classified is only one part of that package. Before a model is deployed, it needs performance in new patients and evidence that using it helps decisions. Models derived during one rehabilitation programme also predict outcomes under that pattern of care and discharge practice, not an untreated biological destiny.

The first year independent rising rising speed and real life activity

Janssen and colleagues enrolled 50 patients within four days after stroke and followed transfer-related functioning through 48 weeks. Observed independent rising increased from 27/50 (54%) to 33/40 (83%) among those assessed at 48 weeks. Six accelerometer-recorded rises at a comfortable pace, with and without arms, were assessed using patients' usual chairs and preferred foot positions. The design favoured ecological relevance over a fixed-chair performance test. [16]

Most change occurred early, but some speed and broader functional improvements continued later. The whole-cohort rising-speed index included zeros for those unable to rise, so its improvement reflects both regained capability and faster performance among completers. Daily transfer counts were measured over eight hours, not a full day; activity analysis involved fewer participants. The study describes recovery trajectories rather than an individually validated prognosis. Its greatest contribution is demonstrating why regained independence, speed and everyday activity must be tracked separately. It does not support declaring recovery complete at 12 weeks. [16]

Mechanisms of successful and failed early transfers

Kerr, Clark and Pomeroy analysed repeated kinematics and muscle activation in 91 people early after stroke before and after rehabilitation. Participants were classified as always able (51), never able (19) or becoming able after baseline (21). Forward body positioning and the timing of quadriceps/hamstring activity distinguished important aspects of success and change. This adds a group often missing from biomechanics studies: those initially unable to perform an independent rise. [39]

The inference should remain mechanistic and exploratory. Group membership uses subsequent performance, so a comparison between these groups is not itself a prospectively validated classifier. Analyses identifying baseline variables associated with later movement time in the always-able subgroup apply to that subgroup; they do not establish a decision rule for patients initially unable to stand. Rehabilitation exposure, spontaneous recovery and multiple movement variables complicate causal interpretation. The paper helps clinicians formulate hypotheses about phase-specific constraints, but does not justify predicting inevitable failure from one EMG or CoM observation.

Prediction of later 5STS performance

The 2024 Jerez outpatient study followed 56 initially enrolled patients, 55 surviving to the relevant follow-up, who could walk four metres and complete 5STS. Mean initial time since stroke was 48.4 days. Mean 5STS time improved from 16.1 seconds at admission to 13.2 after one month and 11.6 at discharge. Baseline timing explained 47.6% of discharge-time variance in a simple model; one-month timing explained 72.6%. Multivariable models including age and time to rehabilitation admission explained 57.7% and 75.5% respectively. [40]

This is longitudinal evidence, but its endpoint is later performance on the same test in established completers. It does not predict whether a non-completer will regain independence, nor whether someone will fall. Reassessment being more predictive is plausible because it is temporally closer to discharge and incorporates early recovery. The greater predictive value should not be presented as proof that the reassessment itself improves outcome.

The models were developed in a small single-centre sample with no independent validation reported. Discharge timing and rehabilitation exposure varied. Important lower-limb impairment variables were not measured as covariates. A relationship between a poor baseline score and a larger change score also requires caution: baseline is mathematically part of change, and regression to the mean and greater room for improvement can create or strengthen that relationship. It is not sufficient evidence that slower patients benefit more from a particular treatment. [40]

Future community walking promising development evidence

Medina-Mirapeix and colleagues prospectively recruited 80 people within 60 days after a first stroke into the Jerez outpatient programme during October 2016–October 2018. Mean time since stroke was 34 days. Seventy-four were followed to discharge; two died and four dropped out. Median time to discharge was 64 days, with a wide IQR of 85 days. Only 43/80 could perform the baseline 5STS; inability to complete within one minute was assigned 60 seconds. The test used an armless chair, but the original does not fully specify chair dimensions, arm position or timing events. Neither walking aids nor physical assistance were permitted during the assessed performance tests. [5]

The outcome was a performance-defined classification using gait speed of at least 0.8 m/s and FAC level 5. It was not direct measurement of distance walked in the community, participation or safety in actual outdoor situations. The discharge assessor was blinded to baseline 5STS and gait-speed results, a useful safeguard. Forty-seven of the 74 participants were classified as non-functional community walkers at discharge.

Only baseline 5STS remained in the data-selected final logistic model, with an odds ratio of 1.28 per recorded second (95% CI 1.09–1.50). Apparent baseline AUC was 0.956 (0.91–0.99); the at-least-21-second cut point had sensitivity 87.2% and specificity 88.9%. One-month 5STS gave AUC 0.952 (0.90–0.99), with a different cut point of 15 seconds, sensitivity 85.1% and specificity 92.6%. These are substantial development-sample signals, but the paper explicitly acknowledges absence of external validation. Calibration is not established by an AUC, and treating assigned 60-second inability scores as continuous seconds imposes an assumption about the predictor that requires scrutiny. [5]

The headline combined sensitivity of 98.1% needs a stronger qualification. It was calculated using a parallel-testing formula that assumes knowledge of the baseline result does not change the probability of the reassessment result. It was not measured from an independently validated combined classification rule. Serial tests of the same activity in the same recovering patients are plausibly dependent. Multiplying the marginal error probabilities therefore cannot establish the actual false-negative rate of the combined strategy. The corresponding combined specificity of about 82% is formula-derived too.

There is also an implementation-critical inconsistency in the printed probability equation on journal page 371. As printed, the exponent is −4.84 + 0.25 × 5STS within an inverse-logistic denominator, which would make the labelled poor-outcome probability decrease as 5STS time increases. That conflicts with the positive odds ratio and narrative interpretation. The equation must not be copied into software or silently corrected without confirmation of the intended sign/outcome coding. This does not negate the ROC association; it limits deployment of the published probability formula.

The study excluded prior poor walking and significant cognitive/language impairment, and required retention through the one-month assessment. Discharge is a care-process endpoint whose timing and treatment context can vary across services. Furthermore, the 2021 responsiveness study and 2024 discharge-performance study share this hospital, the 2016–2018 recruitment period and ethical reference EST-42/16. Participant-level overlap is not resolved, but these publications cannot be counted as clearly independent replications. The 21-second cut point remains a provisional classifier for this specific future outcome, not a falls cutoff, a change threshold or a treatment-allocation rule. [3, 5, 40]

A related prospective report by Medina-Mirapeix and colleagues examined transition to a higher ambulation class in 109 patients, including 32 limited-community ambulators. In that subgroup, a 14.8-second 5STS cutoff had AUC 0.822 at three months and 0.857 at discharge. The test was less useful in the lower-function groups, where many could not complete it. Cox analyses adjusted for age and time since stroke, and an association with transition time was reported. [41]

The verified abstract and indexed original Results/analysis/discussion sections support this bounded conclusion; complete clinical Methods and predictor coding were not recovered. Consequently, the reported hazard ratio is not translated into a per-second individual prediction here. Small subgroups, imputed failure times and absence of external validation restrict generalisability. This paper further demonstrates that baseline capability changes which test is informative. It does not provide a general 14.8-second threshold for all stroke, and its relationship to the other Jerez cohorts needs to remain visible rather than assumed independent.

Reduced caregiver assistance Perrys retrospective longitudinal study

Perry and colleagues retrospectively examined 55 consecutive complete charts of patients with first unilateral stroke, with mean onset-to-admission interval 9.2 days and mean rehabilitation stay 18 days. Transfer performance was assessed from a mat adjusted so the thighs were parallel to the floor, using a self-selected method. Three caregiver-assistance categories grouped independence/modified independence, supervision, and physical assistance. Forty-seven patients improved at least one category and eight did not. [42]

An age-adjusted model containing cognition and bilateral ankle dorsiflexion range classified 48/53 outcomes correctly (90.6%). The corresponding adjusted odds ratios were 1.67 (95% CI 1.08–2.56) per cognition-score unit and 1.24 (1.04–1.47) per recorded bilateral-range unit. These numbers were checked against the original tables. They support further consideration of cognition and ankle mobility when planning assistance, but not a ready-to-use risk calculator. [42]

Several issues constrain inference. There were only eight non-improvers, many candidate predictors and data-driven selection. The majority outcome was already improvement in 85.5% of the full sample, so high overall classification accuracy is less impressive than it initially sounds. There was no external validation or calibration assessment sufficient for individual probabilities. A one-category improvement can mean moving from physical assistance to supervision, not becoming independent. Patients unable to comply with testing because of language/cognition were excluded, restricting application to those most likely to need help.

The original also recoded ankle dorsiflexion so neutral was assigned 11 degrees before bilateral summation. A plotted value from that model must not be presented as an untransformed anatomical range threshold. Cognitive-score denominators are described inconsistently in parts of the text. These details reinforce using the study to inform assessment priorities and caregiver planning, rather than reproducing its plotted probabilities in software without clarification. A prognostic association also does not prove that increasing ankle range alone will cause independent transfer recovery.

Falls what is retrospective prospective and clinically usable

Retrospective force and clinical test associations

Cheng and colleagues compared 33 stroke participants, 18 with a history of falls and 15 without, along with 25 controls. Comfortable chair-rise/descent force-plate outcomes showed slower force development and greater mediolateral sway among fallers. The study was explicitly retrospective. It supports an association between transfer kinetics and fall history; it does not demonstrate the accuracy of baseline kinetics for predicting future falls. [43]

Beninato and colleagues' 27-person community case series similarly used multiple-fall history. The 5STS was only weakly associated with that history, while other self-report or functional measures performed better. This is a useful counterweight to selective citation of positive results. Neither study establishes a reason to apply a Parkinson-disease, older-adult or vestibular cutoff to stroke. The observation that two unrelated studies produce numerically similar chair-rise thresholds does not make their outcomes equivalent. [44]

Goto 2019 prospective falls with modest discrimination

Goto and colleagues offer stronger temporal evidence. They recruited 164 ambulatory stroke survivors attending one Japanese adult day-rehabilitation centre; 144 remained in the analysis after 20 stopped attending or died. Mean time since stroke was 5.21 years. Falls were recorded for one year using diaries brought to the centre at each visit, with staff collecting further details when a fall occurred. This active ascertainment is an important strength compared with a one-off long-recall question. There were 62 fallers, 27 recurrent fallers and 126 total falls. [6]

Baseline 5STS was slower in future fallers than non-fallers: 15.4 versus 13.1 seconds on average. However, the ROC AUC was only 0.61 (95% CI 0.52–0.71). The sample-derived 12.4-second cut point had sensitivity 0.63, specificity 0.60, positive predictive value 0.54 and negative predictive value 0.68. These figures mean that considerable overlap remained. They are not consistent with using a quick chair test to rule falls risk in or out with confidence. [6]

The study's falls analysis compared baseline characteristics and constructed ROC curves for variables with significant group differences. It did not establish a multivariable, calibrated and externally validated individual-risk model. Ongoing rehabilitation, activity exposure, cognition, environment and aids can affect both performance and falls. Selection was limited to ambulatory attendees who understood the protocol; prolonged absence and recurrent stroke were exclusion criteria. Attrition may therefore be informative rather than random.

The original 5STS description specifies rapid repeated standing/sitting and permits an orthosis, but does not report all chair geometry, upper-limb positioning or exact timing events. These missing implementation details further limit transport of the 12.4-second cutoff. A small numerical difference between this falls cutoff and Mong's roughly 12-second stroke/control discriminator should not be interpreted as a shared physiological threshold. They classify different outcomes, in different samples, with incompletely matched procedures.

Falls occurred during a range of activities, including standing up and sitting down, but the study's primary falls outcome was not confined to transfers. A slow chair-rise test may index broader vulnerability. Improving chair-rise time alone is therefore not demonstrated to prevent falls. Equally, a fast time does not rule out hazards during walking, turning, toileting, divided attention or getting up from the floor. Transfer assessment should contribute to a broader falls evaluation.

Wearable mixed battery prediction Abdollahi 2024

Abdollahi and colleagues assessed 21 chronic-stroke survivors using eight IMUs during balance, TUG, walking and repeated chair rises, with and without a cognitive task. The Results explicitly state that 11 fallers and 10 non-fallers were classified using six-month follow-ups. It would therefore be inaccurate to dismiss this as purely retrospective fall-history classification. However, the report provides limited detail on fall definition and ascertainment compared with the diary-based Goto study. [7]

The investigators considered 92 candidate features and reported leave-one-subject-out cross-validation, including feature selection and tuning within each training set. That reported nesting is a methodological strength and should not be ignored. The best model reached approximately 91% accuracy using dual-task balance sway and TUG walking features. An alternative single-thorax-sensor battery included chair-rise variables alongside balance, TUG and age. These results do not isolate the incremental contribution of chair rise as a stand-alone prognostic test. [7]

With only 21 participants, each held-out person has a substantial effect on overall accuracy. Extensive exploration of features, algorithms, sensor combinations and task subsets also creates uncertainty even when individual cross-validation steps are nested. No independent cohort, calibration performance or clinical utility was demonstrated. The selected participants could walk 10 m without assistance, and important comorbidity restrictions limit generalisability. The appropriate claim is an exploratory prospective mixed-battery model, not a validated sensor-based stroke falls prediction product.

Directional variability the 2026 pressure mat pilot

Lee, Yuk and Lee conducted a prospective observational pilot in a Korean rehabilitation unit. Twenty-five people with first-ever stroke completed baseline assessments; one was lost to contact, leaving 24 with three-month post-discharge falls outcomes. Five reported at least one fall and 19 did not. Sixteen healthy controls contributed a separate comparison, but cannot enlarge the effective sample for predicting falls after stroke. Stroke participants had to follow two-step commands and maintain independent sitting for at least 30 minutes. Their median age was 74 years, and 22 of the 24 followed participants had ischaemic stroke. These are selected early-subacute patients, not a representative sample of all stroke severity or chronicity. [8]

The protocol was a distinct 12-second repeated-transfer assessment rather than a conventional five-rise time. A FAIV fabric pressure mat with 65,536 sensor nodes sampled at 100 Hz. Participants started at the front edge of a fixed-height armchair, with feet on shoulder-width marks equidistant from the centre reference point. The numerical chair height was not reported. After one to three practice trials, they repeatedly stood and sat at a comfortable, self-selected pace following a verbal start command. Compensatory strategies, including arm swinging, were encouraged; direct examiner physical support was avoided. The article does not establish a crossed-arms protocol or fully specify permissible armrest use. A trained examiner stood by and stopped unsafe testing. Participants finished a cycle already underway at 12 seconds, although only the exact 12-second signal window entered analysis. [8]

For each cycle, the investigators calculated the major-axis orientation of a 90% prediction ellipse fitted to the centre-of-pressure trajectory. The range of these angles across cycles was the STS Direction Range. This measures a particular feature of within-session pressure-trajectory variation. It does not directly measure neural feedforward control, leg power or an individual's future fall probability. Implementation would require the cycle-segmentation rules, treatment of an incomplete final cycle and handling of the equivalent 0°/180° ellipse-axis orientations to be explicit. Because people can complete different numbers of cycles in a fixed-duration window, the dependence of a maximum-minus-minimum range on cycle count also needs evaluation. These are reproducibility and validation requirements, not evidence that the published algorithm necessarily failed.

Fall occurrence was ascertained by a structured telephone questionnaire three months after discharge, using an unintentional event ending on the ground or a lower surface as the definition. Thus, the cohort is genuinely prospective, but ascertainment depends on follow-up recall rather than the repeatedly reviewed diaries used by Goto. Environmental circumstances and locations of falls were not systematically documented. Clinical assessments and instrumental assessments were also separated in time: group medians were 13–14 days after onset for the clinical evaluation and 21–22 days for the mat assessment. These group summaries are not paired estimates of the delay, but they show that the added instrument was generally measured later in a rapidly changing recovery phase. Apparent added prediction cannot be attributed solely to finer measurement resolution. [8]

Median direction range was 150° in future fallers and 115° in non-fallers, with a reported Cohen's d of 0.91. The Mann–Whitney comparison was p = 0.055, which did not meet the paper's prespecified 0.05 threshold, despite the discussion describing the difference as significant. The age-plus-BBS model had an AUC of 0.574 (95% CI 0.29–0.86); adding direction range increased the apparent AUC to 0.768 (0.52–1.00), a difference of 0.194. The main article reports neither uncertainty for that increment nor a validated decision threshold. Its model table contains predictor names and AUCs rather than deployable coefficients or calibrated probabilities. [8]

The authors appropriately acknowledge overfitting in their limitations: three predictors were fitted with only five fall events. The main Methods describe no resampling-based optimism correction, held-out test set or external validation. This severe information constraint and the broad confidence intervals matter more than assigning a qualitative label such as “acceptable” to the point estimate. The comparison was against a small-sample age/BBS model, not a demonstrated optimal clinical falls assessment. Non-significant BBS group differences do not establish that ordinal scoring, floor effects or ceiling effects caused its weak performance in this sample. Likewise, a larger direction range does not show that training people to reduce it will prevent falls.

The original's supplementary validation concerned quiet-standing COP displacement/path length, correlations with BBS and stroke/control discrimination. The main narrative says only the eyes-closed mean-displacement correlation reached significance and that eyes-closed measures separated groups; the supplementary table itself was not recovered. These analyses should not be treated as direct force-plate criterion validation, between-day reliability or validation of the dynamic direction-range outcome. Age and sex also differed between controls and stroke participants. The complete main article therefore strengthens the case for an interesting, hypothesis-generating prospective signal while making the limits much clearer: there is no validated pressure-mat cutoff, calibrated individual risk estimate or established therapeutic target ready for clinical implementation. [8]

What remains unestablished

The retrieved evidence does not establish an externally validated, broadly applicable chair-rise-only rule for future falls, future independent living, institutionalisation or mortality after stroke. There is useful longitudinal evidence for specific transfer or mobility outcomes, but it is setting- and outcome-dependent. Generic cohort associations involving chair stands in older adults, or stroke as one of many covariates, should not be presented as direct prognosis for people living with stroke.

The research priority is not simply another statistically significant association. It is a well-described cohort spanning non-completers and supported completers, prespecified outcome/time horizon, robust falls or independence ascertainment, appropriate handling of rehabilitation exposure and competing events, and validation of incremental clinical benefit beyond inexpensive baseline measures. Until then, chair-rise results can help identify concerns and monitor recovery while supporting, rather than replacing, individual clinical judgement.

Table 6 Longitudinal outcomes must stay separate

Practical selection across the stroke pathway

Acute stroke and people unable to rise independently

Begin with medical stability, task safety and the assistance required. Document whether the person initiates the movement, achieves seat-off, reaches upright stance, remains stable and lowers safely. Record chair setup, supports and the therapist's role. An assistance category or carefully described single transfer is often the primary outcome; repeated unsupported testing may be inappropriate or impossible. An unattempted test must retain its reason.

When repeated testing becomes feasible, choose a protocol that the person can perform safely and that answers the clinical question. The original 30-second test and its arm-assisted modification can both be useful, but they must be named separately. Retain the history of support reduction even if the timed outcome becomes temporarily slower under the harder condition. The research does not justify assigning a poor prognosis simply because a patient cannot perform an unsupported five-rise test early after stroke. [2, 16, 21, 39, 42]

Subacute rehabilitation and discharge planning

Pair a standardised performance result with an independence/assistance outcome and a relevant patient goal. Reassess under the same chair, foot, arm, aid and timing conditions, and interpret change in the context of time since stroke and concurrent recovery. Where published change estimates are used, identify the matching population and anchor. A walking-anchored 5STS change threshold should be described as such. [3, 5, 40]

For discharge planning, the outcome of interest may be rising from the actual home toilet, bed or lounge chair with an available caregiver. A clinic 5STS can add a reproducible indicator but cannot replace an environmental and assistance assessment. Predicted community-walking status, current transfer independence and need for caregiver training should appear as separate clinical judgements. A statistical model developed in one outpatient programme should not control access to rehabilitation or discharge destination.

Chronic stroke and higher functioning survivors

In independent completers, 5STS can efficiently track repeated-transfer performance, while 30-second testing may capture sustained repetition capacity. If the issue is unilateral contribution or movement strategy, add bilateral loading or kinematic assessment with a validated setup rather than inferring asymmetry from total time. Interpret the less-affected limb as a contributor with its own potential deficits.

For falls, use chair-rise findings to prompt questions about circumstances, prior falls, balance, walking, cognition, symptoms, home hazards and assistive devices. The prospective evidence supports concern when performance is poorer, but the observed discrimination is not sufficient for a binary “safe/unsafe” clearance. Physical activity exposure also matters: a relatively capable person may fall more because they undertake more activities. [6–8]

Remote follow up

Select a remote procedure supported by evidence for that version, and train the patient/helper before relying on the score. Verify that the whole movement is visible, the chair is stable, the same chair is used over time, assistance is documented and the instruction is understood. Keep a record of technical failure, incomplete view and ambiguity about repetition completion. A video score that cannot be confidently checked should be marked for review rather than silently accepted.

The evidence for remote 5STS is conflicting across protocols, and remote 30-second construct evidence is stronger than a claim of complete interchangeability. Rerating a recording is useful quality control, but it cannot estimate how the patient would perform on a different day. Do not combine the two forms of reliability in a single unqualified badge. [24, 25, 27]

Claim guidance for rehabtools and related clinical products

Claims that the evidence can support

A defensible product description can say that the system records performance on a specified chair-rise protocol and supports longitudinal review. If the software counts repetitions or measures duration, state the actual outcome. If a particular sensor/procedure has been validated, specify the population, task and endpoint. For example, evidence about a calibrated dual-board affected/unaffected loading ratio should be tied to that ratio and method, not expanded into “validated balance assessment” without qualification. [1, 2, 15, 34]

The interface should preserve the primary observation and provenance: raw time or count; date/time; stroke phase; affected side; chair dimensions; arms and feet; aids/orthoses; assistance; practice/scored trials; event definitions; and successful versus failed or terminated attempts. For instrumented tests, store device and software versions, sample/processing settings, analysed phase, units, denominator and quality flags. Derived values should be traceable to their inputs.

A longitudinal display can distinguish three outcomes: performance under the same conditions, a change in task difficulty/support, and change in everyday transfer goals. This is more faithful to rehabilitation than a single arrow that treats every numeric reduction as recovery. The display should also show when the score comes from an imputation or a changed protocol.

Claims that should be restricted or avoided

Avoid claiming that a stopwatch test directly measures affected-leg strength or power. If an equation estimates power, label it as an estimate and explain the assumptions and validation population. Avoid saying that symmetry proves neurological recovery, that a statistically significant correlation establishes agreement, or that a high rater ICC establishes low between-day patient error.

Do not present 12, 12.4 or 21 seconds as interchangeable stroke “risk cutoffs.” They concern different comparisons or future outcomes. Do not label a three-repetition 30-second change as a validated stroke MIC solely because it exceeds a reliability study's error threshold. Do not turn the smallest published 5STS MCID into a universal improvement trigger. A model should not display a personalised probability unless the complete prediction equation, relevant calibration and appropriate validation are available. [1–3, 5, 6]

No generic fall-risk score should be justified solely by mixing retrospective associations with prospective development studies. Nor should the high accuracy of a mixed sensor battery be attributed to chair rise alone. Claims about remote assessment must distinguish construct validity, cross-mode agreement and responsiveness. Markerless CoM validation in a calibrated multi-camera laboratory should not be cited as validation of arbitrary single-camera force or power estimates. [7, 24, 25, 27, 34]

Table 7 Examples of defensible and overextended claims

How to communicate an individual result

A useful report says what was done and what changed, then identifies uncertainty. For example: “Five full stand-and-sit cycles took 18.2 seconds using the same 43-cm chair and no hand support; the previous time was 21.0 seconds. Assistance and setup were unchanged. The difference should be interpreted alongside measurement error and the patient's transfer goals.” It does not need an unsupported label such as “power increased by 15%” or “falls risk is now low.” These example numbers illustrate reporting structure only; they are not thresholds or patient data.

When a support condition changes, say so directly: “Now completes a single rise without physical assistance from the raised chair; repeated unsupported testing remains incomplete.” That is often more clinically useful than converting the earlier failure into 60 seconds and displaying an apparently dramatic percentage improvement. Software can make this nuanced interpretation easier by preserving categorical transitions and allowing explanatory text.

Research priorities and overall conclusion

The immediate measurement priorities are larger between-day studies spanning severity; reproducible chair/arm/foot/event protocols; separate reporting of supported and unsupported versions; agreement and absolute error for sensor-derived outputs; and patient anchors that concern transfer ability itself. Early stroke studies should retain non-completers and explain how transitions into successful performance are analysed. Studies of meaningful change should distinguish a patient's minimal benefit from marked improvement, and prespecify handling of ceiling imputations and changing recovery strata.

Technology research should benchmark the final clinical output rather than only the raw sensor. Validation datasets should include atypical strategies, unsuccessful attempts, assistance, orthoses and occlusion, with external evaluation across devices and settings. For power, the reference should measure compatible mechanical work rate; correlation with TUG or a self-report strength scale is insufficient. For daily-life transitions, event validity and wear-time completeness precede claims about activity or recovery.

Prognostic studies should move beyond development-sample discrimination. Useful next steps include external validation, calibration, added value over existing clinical measures, explicit missing-data handling, and demonstration that predictions improve decisions without excluding people with severe disability. Future falls should be actively ascertained and separated from fall history; discharge walking, transfer independence and everyday participation should remain distinct outcomes.

The overall conclusion is positive but specific: chair-rise assessment is a valuable component of stroke rehabilitation when the test is matched to capability and the interpretation is matched to the evidence. Standardised time/count measures are practical; assistance outcomes make severe stroke visible; instrumentation can clarify mechanism. The strongest clinical use is to document reproducible performance and meaningful functional transitions. Universal power, recovery and risk claims go beyond the present evidence.

Appendix A Search boundaries and source interpretation

The exact PubMed measurement query was:

(stroke[Title/Abstract] OR hemipleg*[Title/Abstract] OR hemipare*[Title/Abstract]) AND ("sit-to-stand"[Title/Abstract] OR "sit to stand"[Title/Abstract] OR "chair rise"[Title/Abstract] OR "chair stand"[Title/Abstract]) AND (reliab*[Title/Abstract] OR valid*[Title/Abstract] OR repeatab*[Title/Abstract] OR responsiv*[Title/Abstract] OR "measurement error"[Title/Abstract] OR "minimal detectable"[Title/Abstract] OR "clinically important"[Title/Abstract])

The exact PubMed prognosis query was:

(stroke[Title/Abstract] OR hemipleg*[Title/Abstract] OR hemipare*[Title/Abstract]) AND ("sit-to-stand"[Title/Abstract] OR "sit to stand"[Title/Abstract] OR "chair rise"[Title/Abstract] OR "chair stand"[Title/Abstract]) AND (predict*[Title/Abstract] OR prognos*[Title/Abstract] OR prospective[Title/Abstract] OR longitudinal[Title/Abstract] OR falls[Title/Abstract] OR independence[Title/Abstract])

The exact PubMed review query was:

(stroke[Title/Abstract] OR hemipleg*[Title/Abstract] OR hemipare*[Title/Abstract]) AND ("sit-to-stand"[Title/Abstract] OR "sit to stand"[Title/Abstract] OR "chair rise"[Title/Abstract] OR "chair stand"[Title/Abstract]) AND (review[Title/Abstract] OR meta-analysis[Publication Type])

The exact Scopus measurement/prognosis query was:

TITLE-ABS-KEY((stroke OR hemipleg* OR hemipare*) AND ("sit-to-stand" OR "chair-rise" OR "chair-stand") AND (reliab* OR valid* OR "measurement error" OR "detectable change" OR prognos* OR predict* OR "clinically important"))

The exact Scopus protocol/technology query was:

TITLE-ABS-KEY((stroke OR hemipleg* OR hemipare*) AND ("sit-to-stand" OR "stand-to-sit" OR "chair rise") AND ("foot position" OR "feet position" OR "chair height" OR "seat height" OR "arm position" OR "30-second" OR "one-minute" OR "1-minute" OR "power" OR "video" OR "kinematic"))

Original connector PubMed pages overlapped and did not reconcile. Official NCBI ESearch identifier sets and EFetch of all identifiers resolved that discrepancy for the identical queries. Scopus pages reconciled directly. No date/language restriction was added to these bounded queries; non-stroke populations and unrelated uses of the word stroke were not treated as direct clinical evidence. Online-first and issue dates may differ, so DOI and source identity take priority in reference matching.

The reference list states whether a complete original, author manuscript, original sections or an abstract was examined. Full texts were obtained from public originals, repositories and permitted institutional copies. Abstract/excerpt-only sources are used for bounded findings and are not assumed to contain unreported protocols or validation. Selected primary-study summaries below retain the major access and applicability limits. Study counts should not be summed into a pooled evidence denominator, especially where overlapping clinical programmes or selected secondary analyses are involved.

Table 8 Bounded searches and reconciled coverage

Counts are per query before cross-query deduplication, not numbers of included studies. Search date: 2 October 2026. Connector PubMed overlap was resolved using the identical queries through official NCBI.

Appendix B Primary study map

The following grouped tables summarise the principal original studies discussed in the report. They are a reading guide to design, population, procedure and limitations, not a substitute for the detailed appraisal in the preceding chapters. An author calling a variable strength, power, prediction or responsiveness does not automatically establish that construct; the report uses the study design and reference standard to judge the claim.

Table 9 Primary study map clinical measurement

Selected appraised sources. Publications from the Jerez programme may overlap and should not be counted as independent replication.

Table 10 Primary study map remote and wearable clinical assessment

Selected appraised sources. Publications from the Jerez programme may overlap and should not be counted as independent replication.

Table 11 Primary study map instrumented mechanisms

Selected appraised sources. Publications from the Jerez programme may overlap and should not be counted as independent replication.

Table 12 Primary study map recovery and prognosis

Selected appraised sources. Publications from the Jerez programme may overlap and should not be counted as independent replication.

Table 13 Primary study map task and construct boundaries

Selected appraised sources. Publications from the Jerez programme may overlap and should not be counted as independent replication.

References

References are numbered in first citation order. Study specific source descriptions identify the material examined and do not constitute a study quality rating. Links identify the original publication or the explicitly named primary source version.

1. Mong Yiqin, Teo Tilda W, Ng Shamay S. 5-repetition sit-to-stand test in subjects with chronic stroke: reliability and validity. Archives of physical medicine and rehabilitation. 2010;91(3):407-13. DOI 10.1016/j.apmr.2009.10.030 Source examined: Full original text on web.

Source note: SRC-410076a94831 Mong Yiqin 2010

2. Lyders Johansen Katrine, Derby Stistrup Rikke, Skibdal Schjøtt Camilla, Madsen Jacqueline, Vinther Anders. Absolute and Relative Reliability of the Timed 'Up & Go' Test and '30second Chair-Stand' Test in Hospitalised Patients with Stroke. PloS one. 2016;11(10):e0165663. DOI 10.1371/journal.pone.0165663 Source examined: Full text.

Source note: SRC-e354e4dbcabc Lyders Johansen Katrine 2016

3. Agustín Rodrigo Martín-San, Crisostomo Mª José, Sánchez-Martínez Mª Piedad, Medina-Mirapeix Francesc. Responsiveness and Minimal Clinically Important Difference of the Five Times Sit-to-Stand Test in Patients with Stroke. International journal of environmental research and public health. 2021;18(5). DOI 10.3390/ijerph18052314 Source examined: Full text.

Source note: SRC-0625b74e491a Agustin Rodrigo Martin-San 2021

4. An, SeungHeon, Lee, DongGeon, Park, DongMin, Lee, Kyeongbong. Differences in Functional Performance and Minimal Detectable Change According to Levels of Ankle Plantar Flexor Spasticity in Patients with Chronic Stroke. Journal of clinical medicine. 2025;14(20). DOI 10.3390/jcm14207358 Source examined: Full text.

Source note: SRC-cf8951de2ab7 An 2025

5. Medina-Mirapeix, Francesc, Crisóstomo, María José, Gacto-Sánchez, Mariano, Escolar-Reina, M Pilar, Sánchez-Martínez, M Piedad, Martín-SanAgustín, Rodrigo, et al. The 5-STS is a prognostic factor of sub-acute stroke patients who will not become community walkers at discharge from rehabilitation. NeuroRehabilitation. 2023;53(3):367–375. DOI 10.3233/nre-230161 Source examined: Full text.

Source note: SRC-3a7aada66fc8 Medina-Mirapeix 2023

6. Goto Yuto, Otaka Yohei, Suzuki Ken, Inoue Seigo, Kondo Kunitsugu, Shimizu Eiji. Incidence and circumstances of falls among community-dwelling ambulatory stroke survivors: A prospective study. Geriatrics & gerontology international. 2019;19(3):240–244. DOI 10.1111/ggi.13594 Source examined: Full text.

Source note: SRC-715437443b72 Goto Yuto 2019

7. Abdollahi, Masoud, Rashedi, Ehsan, Jahangiri, Sonia, Kuber, Pranav Madhav, Azadeh-Fard, Nasibeh, Dombovy, Mary. Fall Risk Assessment in Stroke Survivors: A Machine Learning Model Using Detailed Motion Data from Common Clinical Tests and Motor-Cognitive Dual-Tasking. Sensors (Basel, Switzerland). 2024;24(3). DOI 10.3390/s24030812 Source examined: Full text.

Source note: SRC-92ae8870ce08 Abdollahi 2024

8. Lee, Hyun Haeng, Yuk, Doyoung, Lee, Jongmin. Quantifying directional variability during sit-to-stand for post-stroke fall risk stratification: a pilot study. Topics in stroke rehabilitation. 2026:1-10. DOI 10.1080/10749357.2026.2651791 Source examined: Full main text; supplement unavailable.

Source note: SRC-4803f3a9d780 Lee 2026

9. Silva Paula F S, Quintino Ludmylla F, Franco Juliane, Faria Christina D C M. Measurement properties and feasibility of clinical tests to assess sit-to-stand/stand-to-sit tasks in subjects with neurological disease: a systematic review. Brazilian journal of physical therapy. 2014;18(2):99-110. DOI 10.1590/s1413-35552012005000155 Source examined: Full text.

Source note: SRC-0c5da02477f4 Silva Paula F S 2014

10. Boukadida Amira, Piotte France, Dehail Patrick, Nadeau Sylvie. Determinants of sit-to-stand tasks in individuals with hemiparesis post stroke: A review. Annals of physical and rehabilitation medicine. 2015;58(3):167-72. DOI 10.1016/j.rehab.2015.04.007 Source examined: Full text.

Source note: SRC-3b90142aa2c6 Boukadida Amira 2015

11. Onursal Kılınç, Özge, De Ridder, Roel, Kılınç, Muhammed, Van Bladel, Anke. Trunk and lower extremity biomechanics during sit-to-stand after stroke: A systematic review. Annals of physical and rehabilitation medicine. 2023;66(3):101676. DOI 10.1016/j.rehab.2022.101676 Source examined: Full text.

Source note: SRC-4cf2c414fae0 Onursal Klnc 2023

12. Pollock, Alex, Gray, Charla, Culham, Elsie, Durward, Brian R, Langhorne, Peter. Interventions for improving sit-to-stand ability following stroke. The Cochrane database of systematic reviews. 2014;2014(5):CD007232. DOI 10.1002/14651858.cd007232.pub4 Source examined: Review abstract and sections.

Source note: SRC-948e103e6f00 Pollock 2014

13. Alt Murphy, Margit, Munoz-Novoa, Maria, Heremans, Charlotte, Branscheidt, Meret, Cabanas-Valdés, Rosa, Engelter, Stefan T, et al. European Stroke Organisation (ESO) guideline on motor rehabilitation. European stroke journal. 2025;10(4):1160-1188. DOI 10.1177/23969873251338142 Source examined: Full original text on web.

Source note: SRC-92188784ac7b Alt Murphy 2025

14. Van Criekinge, Tamaya, Heremans, Charlotte, Burridge, Jane, Deutsch, Judith E, Hammerbeck, Ulrike, Hollands, Kristen, et al. Standardized measurement of balance and mobility post-stroke: Consensus-based core recommendations from the third Stroke Recovery and Rehabilitation Roundtable. International journal of stroke : official journal of the International Stroke Society. 2023. DOI 10.1177/17474930231205207 Source examined: Full original text on web.

Source note: SRC-723517f23983 Van Criekinge T 2024

15. Bower Kelly J, McGinley Jennifer L, Miller Kimberly J, Clark Ross A. Instrumented static and dynamic balance assessment after stroke using Wii Balance Boards: reliability and association with clinical tests. PloS one. 2014;9(12):e115282. DOI 10.1371/journal.pone.0115282 Source examined: Full text.

Source note: SRC-e37bc164bdf9 Bower Kelly J 2014

16. Janssen, Wim, Bussmann, Johannes, Selles, Ruud, Koudstaal, Peter, Ribbers, Gerard, Stam, Henk. Recovery of the sit-to-stand movement after stroke: a longitudinal cohort study. Neurorehabilitation and neural repair. 2010;24(8):763-9. DOI 10.1177/1545968310363584 Source examined: Full original text on web.

Source note: SRC-80355895de66 Janssen 2010

17. Garcia Oliveira, Simone, Nogueira, Samuel Lourenço, Uliam, Nicoly Ribeiro, Girardi, Paulo Matheus, Russo, Thiago Luiz. Measurement properties of activity monitoring for a rehabilitation (AMoR) platform in post-stroke individuals in a simulated home environment. Topics in stroke rehabilitation. 2025;32(2):119-129. DOI 10.1080/10749357.2024.2377520 Source examined: Abstract only.

Source note: SRC-78c9a0b551f1 Garcia Oliveira 2025

18. Eng Janice J, Chu Kelly S. Reliability and comparison of weight-bearing ability during standing tasks for individuals with chronic stroke. Archives of physical medicine and rehabilitation. 2002;83(8):1138-44. DOI 10.1053/apmr.2002.33644 Source examined: Full original text on web.

Source note: SRC-a85b682abe0b Eng Janice J 2002

19. Hou, Meijin, He, Jian, Liu, Dongwei, Guo, Chenyi, Ma, Ye, Luo, Xiaobo. Bilateral lower limb symmetry during sit-to-stand and stand-to-sit tasks in stroke patients with hemiplegia. Frontiers in neurology. 2025. DOI 10.3389/fneur.2025.1494133 Source examined: Full text.

Source note: SRC-fc872664500b Hou 2025

20. Mao, Yu Rong, Wu, Xiu Qin, Zhao, Jiang Li, Lo, Wai Leung Ambrose, Chen, Ling, Ding, Ming Hui, et al. The Crucial Changes of Sit-to-Stand Phases in Subacute Stroke Survivors Identified by Movement Decomposition Analysis. Frontiers in neurology. 2018. DOI 10.3389/fneur.2018.00185 Source examined: Full text.

Source note: SRC-a957f7837a9b Mao 2018

21. Arens Christian Hedelund, Johnsen Nicole Milwertz, Milanesi Manuela, Weli Ali, Linnebjerg Connie, Christensen Hanne, et al. Inter-tester reliability and agreement of the Cumulated Ambulation Score in acute stroke: The InTRO-CAS-stroke study. Journal of stroke and cerebrovascular diseases : the official journal of National Stroke Association. 2024;33(5):107630. DOI 10.1016/j.jstrokecerebrovasdis.2024.107630 Source examined: Full text.

Source note: SRC-6ced24d49de6 Arens Christian Hedelund 2024

22. Ng, Shamay S M, Cheung, Susanna Y, Lai, Lauren S W, Liu, Ann S L, Ieong, Selena H I, Fong, Shirley S M. Association of seat height and arm position on the five times sit-to-stand test times of stroke survivors. BioMed research international. 2013. DOI 10.1155/2013/642362 Source examined: Full text.

Source note: SRC-558046641c25 Ng 2013

23. Kwong, Patrick W H, Ng, Shamay S M, Chung, Raymond C K, Ng, Gabriel Y F. Foot placement and arm position affect the five times sit-to-stand test time of individuals with chronic stroke. BioMed research international. 2014. DOI 10.1155/2014/636530 Source examined: Full text.

Source note: SRC-2a8bf6714cbb Kwong 2014

24. Deshmukh Shravni, Freels Sally, Madhavan Sangeetha. Teleassessments of Lower Limb Function in Adult Stroke Survivors: A Preliminary Study Evaluating Safety, Feasibility, and Validity for Telerehabilitation. Telemedicine reports. 2024;5(1):322-329. DOI 10.1089/tmr.2024.0052 Source examined: Full text.

Source note: SRC-fa74a1cd4a06 Deshmukh Shravni 2024

25. da Silva Bruna Nascimento Zanfir, Pinto Camila, Figueiredo Caroline Santos, da Rosa Thainara Cruz, Hsieh Katherine Lee, Pagnussat Aline Souza. Tele-assessment of Mobility and Balance is Reliable and Safe for Individuals with Chronic Stroke - A Guideline for a Systematic Physical Evaluation. International journal of telerehabilitation. 2025;17(2):6710. DOI 10.63144/ijt.2025.6710 Source examined: Full text.

Source note: SRC-dfec7c3929bb da Silva Bruna Nascimento Zanfir 2025

26. Mentiplay, Benjamin F, Clark, Ross A, Bower, Kelly J, Williams, Gavin, Pua, Yong-Hao. Five times sit-to-stand following stroke: Relationship with strength and balance. Gait & posture. 2020;78:35–39. DOI 10.1016/j.gaitpost.2020.03.005 Source examined: Full original text.

Source note: SRC-54f300daf6c5 Mentiplay 2020

27. Noguchi, Kenneth S, Liang, Allison, Wiley, Elise, Park, Sarah, Sakakibara, Brodie M, Tang, Ada. Measurement Properties of a Virtually Administered 30-Second Chair Stand Test in People With Stroke. Journal of neurologic physical therapy : JNPT. 2025;49(4):232–239. DOI 10.1097/npt.0000000000000520 Source examined: Full text.

Source note: SRC-a46f3e578634 Noguchi 2025

28. Ng Shamay. Balance ability, not muscle strength and exercise endurance, determines the performance of hemiparetic subjects on the timed-sit-to-stand test. American journal of physical medicine & rehabilitation. 2010;89(6):497-504. DOI 10.1097/phm.0b013e3181d3e90a Source examined: Full original text on web.

Source note: SRC-b028d0bca73b Ng Shamay 2010

29. Bohannon Richard W. Knee extension strength and body weight determine sit-to-stand independence after stroke. Physiotherapy theory and practice. 2007;23(5):291-7. DOI 10.1080/09593980701209428 Source examined: Abstract only.

Source note: SRC-4c411f0c3245 Bohannon Richard W 2007

30. Machado, Natasha, Williams, Gavin, Olver, John, Johnson, Liam. A timed sit-to-stand test and prediction equation had poor agreement with a graded exercise test in people with stroke. Disability and rehabilitation. 2026;48(5):1463-1475. DOI 10.1080/09638288.2025.2531301 Source examined: Abstract plus original excerpts.

Source note: SRC-16fd397e5488 Machado 2026

31. Toshifumi Fujita, Manabu Iwata, Michitaka Fukuda. Reliability and Validity of a New Test for Muscle Power Evaluation of Stroke Patients. Journal of Physical Therapy Science. 2011. DOI 10.1589/jpts.23.259 Source examined: Full text.

Source note: SRC-9b61ff582e22 Fujita T 2011

32. Merchán-Baeza Jose Antonio, González-Sánchez Manuel, Cuesta-Vargas Antonio Ignacio. Using Smartphones to Collect Quantitative Data on Lower Limb Functionality in People Who Have Suffered a Stroke. Journal of stroke and cerebrovascular diseases : the official journal of National Stroke Association. 2018;27(12):3555-3562. DOI 10.1016/j.jstrokecerebrovasdis.2018.08.012 Source examined: Abstract only.

Source note: SRC-3f02a1433d36 Merchan-Baeza Jose Antonio 2018

33. Sánchez-Sánchez M Luz, Ruescas-Nicolau Maria-Arantzazu, Arnal-Gómez Anna, Iosa Marco, Pérez-Alenda Sofía, Cortés-Amador Sara. Validity of an android device for assessing mobility in people with chronic stroke and hemiparesis: a cross-sectional study. Journal of neuroengineering and rehabilitation. 2024;21(1):54. DOI 10.1186/s12984-024-01346-5 Source examined: Full text.

Source note: SRC-a9761f3d0cf4 Sanchez-Sanchez M Luz 2024

34. Majoni, Nigel, Inness, Elizabeth L, Jagroop, David, Danells, Cynthia J, Mansfield, Avril. Differences in centre of mass measurements between markerless and marker-based motion capture systems during balance and mobility assessments in individuals with chronic and sub-acute stroke. Gait & posture. 2026. DOI 10.1016/j.gaitpost.2026.110256 Source examined: Full text.

Source note: SRC-36bab49248f1 Majoni 2026

35. de Andrade, Pedro Henrique Sousa, Bazan, Rodrigo, Sande de Souza, Luciane Aparecida Pascucci, Luvizutto, Gustavo José. Development and Content Validity of the Movement Compensatory Screening (MoCS): A Structured Framework for Identifying Compensatory Strategies After Stroke. Neurorehabilitation and neural repair. 2026:15459683261464075. DOI 10.1177/15459683261464075 Source examined: Abstract only.

Source note: SRC-71488974291e de Andrade 2026

36. Burnfield, Judith M, McCrory, Bernadette, Shu, Yu, Buster, Thad W, Taylor, Adam P, Goldman, Amy J. Comparative kinematic and electromyographic assessment of clinician- and device-assisted sit-to-stand transfers in patients with stroke. Physical therapy. 2013. DOI 10.2522/ptj.20120500 Source examined: Abstract only.

Source note: SRC-e199a38322e2 Burnfield 2013

37. Chandler Elizabeth Ann, Stone Thomas, Pomeroy Valerie Moyra, Clark Allan Brian, Kerr Andrew, Rowe Phillip, et al. Investigating the Relationships Between Three Important Functional Tasks Early After Stroke: Movement Characteristics of Sit-To-Stand, Sit-To-Walk, and Walking. Frontiers in neurology. 2021;12:660383. DOI 10.3389/fneur.2021.660383 Source examined: Full text.

Source note: SRC-4f24119e15d7 Chandler Elizabeth Ann 2021

38. Jones Gareth D, James Darren C, Thacker Michael, Perry Rhian, Green David A. Gait-initiation onset estimation during sit-to-walk: Recommended methods suitable for healthy individuals and ambulatory community-dwelling stroke survivors. PloS one. 2019;14(5):e0217563. DOI 10.1371/journal.pone.0217563 Source examined: Full text.

Source note: SRC-40f671ac458a Jones Gareth D 2019

39. Kerr, Andy, Clark, Allan, Pomeroy, Valerie M. Neuromechanical Differences Between Successful and Failed Sit-to-Stand Movements and Response to Rehabilitation Early After Stroke. Neurorehabilitation and neural repair. 2019. DOI 10.1177/1545968319846119 Source examined: Full original text on web.

Source note: SRC-ef3910eee5b4 Kerr 2019

40. Sánchez-Martínez Maria Piedad, Crisostomo María José, Martín-San Agustín Rodrigo, Montilla-Herrador Joaquina, Escolar-Reina María Pilar, Valera-Novella Elisa, et al. Determination of Five Sit-to-Stand Test Performance at Discharge of Stroke Patients. Diagnostics (Basel, Switzerland). 2024;14(5). DOI 10.3390/diagnostics14050521 Source examined: Full text.

Source note: SRC-fda9786c6b12 Sanchez-Martinez Maria Piedad 2024

41. Medina-Mirapeix, Francesc, Crisostomo, M José, Martín San Agustín, Rodrigo, Sánchez-Martínez, M Piedad. Prognostic value of balance performance for improvements of community ambulation among stroke patients: a cohort study. European journal of physical and rehabilitation medicine. 2022;58(2):171–178. DOI 10.23736/s1973-9087.21.06996-3 Source examined: Abstract plus original excerpts.

Source note: SRC-56cea77586dc Medina-Mirapeix 2022

42. Perry Susan B, Marchetti Gregory F, Wagner Suzanne, Wilton Wendy. Predicting caregiver assistance required for sit-to-stand following rehabilitation for acute stroke. Journal of neurologic physical therapy : JNPT. 2006;30(1):2–11. DOI 10.1097/01.npt.0000282144.72703.cb Source examined: Full text.

Source note: SRC-8d43d1e1c6d3 Perry Susan B 2006

43. Cheng P T, Liaw M Y, Wong M K, Tang F T, Lee M Y, Lin P S. The sit-to-stand movement in stroke patients and its correlation with falling. Archives of physical medicine and rehabilitation. 1998;79(9):1043-6. DOI 10.1016/s0003-9993(98)90168-x Source examined: Abstract only.

Source note: SRC-e65afd11dd1d Cheng P T 1998

44. Beninato, Marianne, Portney, Leslie G, Sullivan, Patricia E. Using the International Classification of Functioning, Disability and Health as a framework to examine the association between falls and clinical assessment tools in people with stroke. Physical therapy. 2009;89(8):816-25. DOI 10.2522/ptj.20080160 Source examined: Abstract only.

Source note: SRC-dec22a7648ac Beninato 2009