Back to Evidence

Evidence-Based Medical AI for Postoperative Stroke Rehabilitation Cohort Review | QSevidence 2026

Evidence-Based Medicine23 min read

Research on rehabilitation after surgery in people with stroke-related hemiplegia needs to connect treatment delivery, functional measurement and differences between procedures.

Evidence-Based Medical AI for Postoperative Stroke Rehabilitation Cohort Review | QSevidence 2026

QSevidence can help organize evidence searches, intervention comparisons and source tracing for rehabilitation physicians, therapists, nurses and statisticians.

This Evidence-based medical AI workflow examines early activity, task-oriented practice, neuromuscular electrical stimulation, psychological support and family guidance together. The research narrative describes an 86-person cohort and functional scores, but study identity, original data and complete statistical outputs remain unverified. Its numerical findings below are research information awaiting verification, not established treatment effects.

Define the surgical population before comparing recovery

The proposed population has imaging-confirmed first stroke, hemiplegia and related surgery. Procedures include decompressive craniectomy, hematoma evacuation, extracranial–intracranial bypass and orthopedic surgery for spasticity or deformity. These settings differ in disease stage, loading restrictions and neurological risk. A reproducible study should record procedure type, stroke-to-surgery time and surgery-to-rehabilitation time rather than treating all postoperative patients as one homogeneous exposure group.

Proposed entry conditions include clinical stability, ability to participate in training and assessment, and informed consent. Severe organ dysfunction, uncontrolled infection and conditions affecting activity safety require exclusion or separate handling. The narrative's blood pressure, heart rate and oxygenation thresholds, together with two alternative cognitive screening cutoffs, are not a single reproducible eligibility rule. Specialists should reconcile these with surgical requirements and local practice. Communication support and adapted assessment should be recorded for aphasia or cognitive difficulties, rather than equating difficulty completing a scale with lack of rehabilitation potential.

Separate the components of the rehabilitation exposure

This is a prospective nonrandomized cohort. Conventional care includes positioning, passive joint movement and education. The comprehensive group additionally receives a structured programme proposed for 8 weeks on 5 days each week. Methods attribute allocation to patient preference and clinical circumstances; results instead mention admission order and resource availability. Both mechanisms can generate selection bias, and the actual allocation process needs clarification.

Proposed components and audit requirements
ComponentResearch scheduleWhat needs checking
Initiation24–72 hours after surgery when clinically appropriateStroke onset, stability assessment, surgical restrictions and reasons for delay; this is not a universal treatment window
Task practiceOnce daily for 60–90 minutes; bed mobility in weeks 1–2, sitting/standing and indoor walking in weeks 3–5, complex walking and upper-limb daily tasks in weeks 6–8Delivered minutes, assistance, fatigue and pain, not scheduled dose alone
Electrical stimulationMuscles selected after assessment; proposed 20-minute sessions once daily, 5 times weeklyIndication, device settings, contraction response, tolerance and professional adjustment
Psychological and family supportPsychological assessment at baseline and weeks 4 and 8; proposed home practice of at least 30 minutes daily and contacts at weeks 12 and 24Caregiver capacity, adherence, home environment and referral needs; these are research arrangements

The stimulation description also specifies 20–50 Hz, pulse widths of 200–400 microseconds and approximately 110% of motor threshold. Device context, contraindications and individualization have not been verified; these settings should not be transferred into patient operating instructions. A study should retain settings logs and qualified rehabilitation professionals should determine actual application. Recording contact time separately for task practice and adjunctive treatments helps distinguish a technology combination from an increase in therapy time.

Use one follow-up axis and consistent outcome measures

Primary outcomes are the Fugl-Meyer Assessment and Modified Barthel Index. Upper-limb, lower-limb and total motor scores should be recorded separately, while daily-activity independence is assessed using a fixed Barthel version. Secondary measures include the Modified Ashworth Scale, Stroke-Specific Quality of Life scale, Berg Balance Scale and Functional Ambulation Category. Scale versions, assessor training and assessor masking need documentation, alongside assistive devices and assessment conditions.

Baseline, 8-week and 24-week assessments are proposed, but the final visit is described both as 24 weeks after intervention and 16 weeks after discharge. These are not necessarily equivalent. A common origin and permitted visit window must be defined before analysis. Baseline NIHSS, preoperative function, lesion location, procedure, complications and access to rehabilitation belong in the same data dictionary.

Read the numerical narrative with its unresolved limitations

The following figures have not been independently verified. Recruitment years remain placeholders, ethics institution and approval details are unresolved, and the baseline table contains missing standard deviations and invalid entries. These figures therefore do not constitute a verified published study or a basis for individual prognosis.

Reported cohort information awaiting original-record verification
ItemComprehensive versus conventional groupInterpretation boundary
Enrollment and follow-up44 versus 42, total 86; final follow-up 42 versus 41, total 83; reported completion 96.5%Two losses in the comprehensive group and one recurrent stroke in the conventional group require event and missingness review
24-week FMA78.4±14.2 versus 65.1±13.5; P<0.001Reported, not independently recalculated; final-visit definition remains unresolved
8-week MBI68.7±15.4 versus 59.1±13.7; P=0.051Not statistically significant at the proposed 0.05 threshold
24-week MBI85.2±10.3 versus 72.8±12.6; P=0.001Analysis numbers, baseline adjustment and interval estimates need clarification
Balance and walkingBerg 48.3±6.7 versus 41.2±7.9; FAC≥4 in 78.6% versus 56.1%Denominators and missing-data handling are incomplete; these are not confirmed implementation effects
Nonserious adverse events3 (6.8%) versus 2 (4.8%); P=0.8No related serious events are claimed in the narrative; small numbers and uncertain ascertainment do not establish equivalent safety

The results identify rehabilitation timing and baseline NIHSS as correlates of MBI improvement, reporting β=-0.327, P=0.003 and β=-0.285, P=0.008, respectively, with adjusted R²=0.412. However, the discussion reports an age coefficient of β=-0.25, P=0.032, contradicting the results section's nonsignificant age finding. Whether the model predicts 8-week or 24-week change also varies. These conflicts should remain explicit until statistical outputs are traced, rather than selecting the more favorable statement. A claim that time to a self-care threshold was approximately 65% shorter lacks time distributions and an estimation method and is not adopted as an efficiency claim.

Connect guideline comparison to professional review

The UK and Ireland National Clinical Guideline for Stroke provides general background for collaborative, individualized rehabilitation, suitable treatment dose and repetitive functional practice. It ties activity to medical status. It does not validate this 86-person cohort or establish a fixed postoperative starting time for every procedure. Surgical restrictions and ongoing rehabilitation assessment should be recorded separately for this population.

  1. Search separately by stroke type, procedure, recovery stage, task practice and stimulation question, distinguishing general stroke evidence from direct postoperative evidence.
  2. Use QSevidence to organize a comparison of populations, comparators, delivered dose, measures and follow-up origins, retaining original links and eligibility conditions.
  3. Link each numerical result to de-identified data, visit records and analysis code, and reconcile the 83 completers with the 86 enrolled participants.
  4. Prespecify group-by-time models, baseline adjustment, within-person correlation and multiple comparisons; assess missing-data assumptions through appropriate sensitivity analyses.
  5. Review pain, fatigue, falls, neurological changes and treatment interruptions jointly with clinical and nursing staff before developing interpretations and quality-improvement actions.

Baseline P values above 0.05 do not remove confounding in nonrandomized groups. Preference, resources, severity and procedure may affect both rehabilitation access and recovery. A score difference cannot identify the independent effect of one component in a bundled intervention. Neuroplasticity is a mechanistic research direction, not a biological endpoint demonstrated by this cohort.

Priorities for a reproducible next study

First reconcile recruitment dates, eligibility, actual allocation, scale versions and statistical endpoints, and obtain ethics and data-provenance documentation. A subsequent multicenter prospective study could stratify by procedure and rehabilitation stage, prespecify a primary endpoint and sample-size rationale, and extend observation of home activity, participation and adverse events. Imaging, biomarkers and remote monitoring are potential future modules rather than completed mechanistic assessments in this cohort.

Frequently asked questions

How can QSevidence support comparison of postoperative rehabilitation programmes?

Organize procedure, population, components, delivered dose and outcomes into an evidence matrix linked to guidelines and primary studies. Professionals then assess applicability, bias and local resources.

How can QSevidence help organize conflicting follow-up findings?

List time origins, missing values, denominators and regression endpoints as explicit verification tasks linked to original records. Statistical and clinical reviewers can then reconcile the reporting version.

How can QSevidence support nursing quality improvement?

Structure training completion, falls, pain, family education and referral records into reviewable indicators and compare them through case audit. Responsible teams make care decisions using individual clinical circumstances.

Sources and evidence classification

Research notice: This page supports evidence organization and methods discussion. Qualified teams assess treatment timing, loading, stimulation and postoperative restrictions; it is not an individualized treatment or rehabilitation prescription.