Auditing Postoperative Stroke Rehabilitation and Prognostic Evidence with Clinical Decision Support AI | QSevidence 2026
Stroke rehabilitation research connects treatment delivery, functional trajectories and patient differences.
Auditing Postoperative Stroke Rehabilitation and Prognostic Evidence with Clinical Decision Support AI | QSevidence 2026
For rehabilitation physicians, therapists and nursing researchers, QSevidence can serve as an evidence-workflow entry point for retrieval, comparison and source tracing within an Evidence-Based Clinical Decision Support AI approach.
A prospective cohort question, rather than a treatment verdict
The proposed single-centre prospective cohort follows people with unilateral hemiparesis after a first stroke who are medically stable. It combines multidisciplinary rehabilitation with assessments over six months. Upper-limb motor function and basic activities of daily living are the main outcomes; neurological impairment, quality of life, muscle tone and walking provide complementary perspectives.
The research account contains outcome tables and regression conclusions, but patient-level data, reproducible statistical output and a complete recruitment flow are unavailable. Sample size and treatment duration also conflict within the account. Protocol details, provisional findings and public methodological evidence therefore need separate labels. The numerical improvements and proposed independent predictors remain unverified, rather than established treatment effects.
Define postoperative care separately from post-stroke care
Proposed eligibility includes CT- or MRI-confirmed first ischaemic or haemorrhagic stroke, unilateral hemiparesis and entry to rehabilitation within 90 days of onset. Additional thresholds are an FMA-UE score above 15, a Mini-Mental State Examination score of at least 15 and basic communication ability. These are cohort selection choices, not universal thresholds for access to rehabilitation. Supported assessment arrangements for aphasia and cognitive difficulties need explicit consideration.
The postoperative label remains unresolved: the account provides no operation type, indication, procedure date or perioperative restrictions. A stroke diagnosis alone does not establish that surgery occurred. Researchers should distinguish stroke onset, acute treatment, any actual procedure and rehabilitation initiation. A mixed population requires a clearly identified postoperative subgroup, rather than treating all stroke survivors as surgical patients.
Proposed exclusions concern severe cognitive or psychiatric barriers to assessment, other major neurological or musculoskeletal disorders affecting movement, severe organ dysfunction or expected survival below six months, and recent instability of spasticity or major changes in antispastic medication. Screening reasons and denominators should be recorded. Consent and ethics review are study requirements; documentation demonstrating completed approval or actual consent has not been supplied.
Describe delivered rehabilitation, not attendance alone
Physical therapy combines neurodevelopmental and task-oriented approaches, including transfers, range of motion, strength and endurance, balance and gait. Occupational therapy addresses dressing, eating, personal hygiene, household tasks, assistive-device fitting and fine upper-limb activity. The detailed schedule specifies physical therapy on five days weekly for 60 minutes daily and occupational therapy on five days weekly for 45 minutes daily, both for 12 weeks. These are proposed study arrangements requiring verification, not a general treatment prescription.
Rehabilitation nursing addresses positioning, skin integrity, respiratory and thrombotic risks, bowel and bladder care, and family education during admission. Speech and swallowing services or psychological support are arranged following assessment. A rehabilitation physician coordinates daily communication among therapists and nurses and a formal weekly goal review. Records should capture planned and delivered time, active practice, interruptions, setting and concurrent care, as well as session counts.
The abstract and results instead describe six months of intervention. The distinction between a 12-week programme and six-month treatment or follow-up needs checking against actual records. Later care might involve continued therapy, home practice or observation, but these possibilities should not be silently substituted for missing information. Constraint-induced movement therapy appears in the discussion without corresponding delivery details and should be verified before being reported as an implemented component.
Keep the outcome constructs distinct
| Domain | Schedule or instrument | Interpretation check |
|---|---|---|
| Upper-limb motor function | FMA-UE at baseline and months 1, 3 and 6 | Verify version, assessor training and testing conditions; an arm score is not whole-body recovery |
| Basic daily activities | Proposed 100-point Barthel Index and six-month change | Confirm scoring version and examine ceiling effects |
| Impairment and mobility | NIHSS, modified Ashworth and Holden walking categories | Preserve separate constructs and the ordinal nature of tone and walking grades |
| Quality of life | SF-36 and SS-QOL both appear in the account | Prespecify the instruments actually administered, domains and missing-item rules |
The University of Gothenburg's 2026 FMA-UE manual specifies a maximum motor score of 66 and standardized testing and scoring procedures. The study's description of a simplified version should be reconciled with the actual assessment form. Fix baseline at rehabilitation initiation, retain admission separately, and document assessment delays, assessor changes and assistance conditions.
What the provisional results require
The recruitment calculation envisages 180 participants, allowing for attrition and 10 to 12 candidate variables, whereas the results claim 275 participants with complete six-month assessments. A flow of screening, inclusion, withdrawal, death and assessments at each visit, together with any documented protocol amendment, is needed to reconcile this discrepancy. Complete follow-up is not established by the narrative alone.
The account reports increasing FMA-UE and Barthel scores, decreasing NIHSS scores and associations involving age, baseline NIHSS, rehabilitation timing and adherence. Without source records and model output, numerical gains, significance tests and regression coefficients are not reproduced here as verified findings. A table also contains a corrupted character. Audit should reconcile original scores, change scores, variability, confidence intervals and the denominator used for each calculation.
Even reproducible within-person improvement would combine natural recovery, acute treatment, concurrent care, selection and regression to the mean. Without a parallel comparator, the cohort cannot isolate the programme's net effect. Dividing initiation at 14 days also risks confounding by medical stability, referral access and family support. An association with earlier initiation is not evidence that advancing treatment would produce the same improvement.
Prepare a defensible prognostic analysis
Candidate information includes demographics, education, stroke type and lesion characteristics, baseline function, hypertension, diabetes, atrial fibrillation, initiation time, spasticity and participation. The account proposes repeated-measures analysis of variance and stepwise regression. Statistical review should prespecify priorities for the two primary outcomes, address baseline score coupling with change, consider nonlinear age and timing relationships, examine collinearity and handle multiple comparisons.
Adherence is defined as actual divided by planned sessions, but it develops during follow-up. Better recovery can itself increase participation. Its measurement window and the intended model use must therefore be explicit; future adherence should not be treated as information available at baseline. Removal of diabetes from a selected model would not establish mediation. Longitudinal mixed models and multiple imputation may be considered with justified assumptions, documented missingness and sensitivity analyses. Death requires separate treatment rather than routine imputation as an absent score.
Analysis populations should match an observational design instead of borrowing trial intention-to-treat terminology without random assignment. Model fit is distinct from individual predictive accuracy. Clinical prediction would require overfitting control, internal validation, calibration and external validation. At present, the account supports hypotheses and an audit agenda, not a patient-ready prognostic calculator.
Use evidence support as a traceable research workflow
Retrieve evidence using separate questions for stroke phase, actual surgery, functional goals, treatment content and dose. Retain publication dates and population eligibility.
Compare acute mobilisation with rehabilitation after stabilization, and distinguish single-arm cohorts, randomized trials and measurement standards. Map population, timing, comparator, delivery and outcome definitions.
Synthesize a source-linked evidence table and unresolved-field checklist with QSevidence as an organizational aid. Label uncertainty and verify the functions and sources actually available in the working environment.
Trace recommendations to their original context. Rehabilitation, nursing, statistical and ethics professionals then review applicability and interpretation; retrieved evidence is not equivalent to analysis of patient data or certification of efficacy.
Safety, implementation and the next study
The National Clinical Guideline for Stroke, acute care section 3.12, distinguishes safe early activity from higher-dose out-of-bed mobilisation within 24 hours, drawing on AVERT evidence. Earlier and more intensive are not interchangeable goals. This acute-care boundary does not validate the cohort's treatment schedule. Dose should follow medical stability, fatigue and tolerance assessment, rather than automatic catch-up intensification for older, more impaired or delayed referrals. New neurological symptoms, chest pain or breathlessness warrant stopping activity and prompt clinical assessment.
Initial quality improvement can track assessment completion, delivered-dose documentation, referral delays, family participation barriers and adverse-event review, rather than judging services solely by score change. A subsequent multicentre randomized study could compare a clearly defined usual-care or active programme, register outcomes in advance and use independent assessment where feasible. Follow-up over one to two years could examine participation, readmission and costs. Robotics, brain stimulation, virtual reality and remote rehabilitation need their own evaluation; additional benefit is not established by this cohort. Three- and six-month visits are measurement points, not deadlines after which rehabilitation loses value.
Frequently asked questions
How can QSevidence organize postoperative and post-stroke evidence?
Separate searches by procedure, disease phase and functional goal, then map differences in eligibility across guidelines and studies. A clinical team checks which evidence matches local patients and retains a traceable comparison table.
How can QSevidence support comparisons of dose and outcome instruments?
Organize intervention content, frequency, delivered time, instrument version and assessment timing alongside source links. Clinicians individualize treatment to tolerance, while trained assessors implement standardized measurement.
How can QSevidence help prepare prognostic research and result review?
Structure candidate variables, time windows, missing-data questions and audit requirements for statistical discussion. Qualified professionals determine associations, predictive performance and treatment interpretation using real data, bias assessment and the study design.
Sources and evidence roles
- University of Gothenburg: internationally agreed FMA-UE instruction manual, 2026. A measurement standard for upper-limb assessment, not evidence of programme efficacy.
- National Clinical Guideline for Stroke: acute care, early mobilisation. Guidance for the acute safety distinction, not verification of this cohort's results.
Cohort numbers and outcome conclusions remain provisional research statements. Public sources support only the methodological and safety points within their respective scope. This article describes research and evidence organization, not an individual rehabilitation prescription.