To evaluate the reporting quality and risk of bias of existing ML-based prognostic prediction models for spinal cord injury (SCI) and examine their clinical applicability, model characteristics, validation strategies, and barriers to clinical implementation.
Approach:
Study Design: Conducted a systematic review following PRISMA guidelines, with a registered protocol in PROSPERO.
Search Strategy: Performed systematic searches across multiple databases including CNKI, PubMed, and Scopus using MeSH and free-text terms related to SCI and machine learning.
Eligibility Criteria: Included studies on individuals aged 18+ with SCI that developed or validated prognostic models, excluding non-original publications and those without sufficient model details.
Data Extraction: Used a structured form based on CHARMS checklist to extract data on study characteristics, model types, performance metrics, and validation types.
Key Findings:
Existing ML models for SCI show variation in outcome measures and modeling approaches.
Challenges include variability in patient characteristics, treatment approaches, and lack of standardized evaluation methods.
Many studies did not adequately address reporting completeness, risk of bias, and model validation.
Interpretation:
The review highlights the need for improved reporting and standardization in ML-based prognostic models for SCI to enhance their clinical applicability.
Limitations:
Variability in study designs and outcome measures complicates comparison across models.
Many included studies inadequately assessed methodological quality.
Conclusion:
A targeted review of ML-based prognostic models for SCI is essential to address reporting quality and bias.
