Identifying Quality and Safety Learning Opportunities Through EHR Data Analytics in Musculoskeletal Practice

Technical Summary: Identifying Quality and Safety Learning Opportunities Through EHR Data Analytics in Musculoskeletal Practice

Problem Statement
Musculoskeletal (MSK) clinicians currently lack objective mechanisms to assess their performance post-graduation. Continuing professional development relies heavily on subjective self-assessment rather than objective data, creating a gap between perceived and actual competence. While Electronic Health Records (EHRs) collect vast amounts of real-world clinical data, there is limited evidence that routine EHR analysis is utilized to identify clinical practice variations or inform clinician development. Current safety paradigms often focus on reactive adverse-event monitoring rather than proactive analysis of routine practice data, which is essential for a "Learning Health System" (LHS) and Safety-II principles (where safety is found in everyday work that goes well). The core problem is the inability to systematically translate routine EHR data into actionable learning opportunities to improve patient outcomes and clinician skills.

Methodology
This study employed a retrospective observational design analyzing de-identified EHR data from an Australian occupational healthcare organization.

• Dataset: The analysis covered 46,539 patient complaints (care plans) from 52 clinicians (chiropractors, physiotherapists, and osteopaths) between July 2014 and November 2021.
• Constraints: To control for case-mix variability, data was restricted to the food manufacturing industry. Clinicians with fewer than 100 complaints were excluded.
• Primary Outcome: "Treatments to Release" (TTR), defined as the number of visits until a patient is released from care (resolution or self-management), including any reinjury or recurrence within six months.
• Variables: Selected based on literature reviews of predictors for work-related MSK outcomes and aligned with the Donabedian structure-process-outcome framework. Key variables included:
  • TTR by Body Region: Grouped into 7 larger areas.
  • Recurrence: Proportion of reopened care plans.
  • Mechanism of Injury: Coded as traumatic, repetitive overuse, static/postural, exertional, or non-specific.
  • Obstacles to Recovery: Documented barriers across nine categories (e.g., psychosocial, workplace, compliance).
  • Prognosis Accuracy: Comparison of initial treatment estimates against actual TTR.
• Analysis Tools: Data analysis and visualization were conducted using Excel and Tableau to prioritize accessibility for clinical leaders over complex statistical software. The study focused on descriptive statistics and confidence intervals to identify patterns without making inferential claims about individual clinician performance.

Key Results
The analysis revealed significant variability across all examined variables, indicating the presence of learning opportunities within routine data:

• Treatment Variability: Mean TTR varied significantly across clinicians (ranging from 1.02 to 5.16). Body region and injury mechanism were associated with recovery length; for instance, shoulder and elbow complaints required an average of 3.89 treatments, while upper back complaints required 2.92.
• Recurrence Rates: There was wide variation in care plan reopenings, ranging from 1.52% to 23.89% across clinicians. Overall, 11.41% of care plans were reopened.
• Mechanism of Injury: "Repetitive overuse" injuries had the highest average TTR (3.69) compared to traumatic injuries (3.18). However, the documentation of "repetitive overuse" varied drastically between clinicians (2.55% to 84.21%), whereas "traumatic" documentation was more consistent.
• Obstacles to Recovery: Documentation rates for obstacles varied widely (1.86% to 85.39%). When obstacles were documented, the mean TTR increased to 4.87 compared to the overall mean of 3.43. Psychosocial factors were recorded in only 7.26% of complaints, with significant under-reporting of factors like fear avoidance and poor coping mechanisms.
• Prognosis Accuracy: Clinicians consistently overestimated recovery length. Only 16.89% of care plans matched the initial prognosis, while 63.36% resulted in faster recovery than predicted. This tendency to overestimate was stable across all experience levels, suggesting prognosis accuracy does not inherently improve with tenure.

Key Contributions

• Proof-of-Concept Methodology: The study demonstrates a practical, non-inferential methodology for using routine EHR data to identify "learning opportunities"—defined as patterns prompting reflection on skills, biases, or training needs—without labeling clinicians as providing substandard care.
• Identification of Specific Gaps: The analysis highlights specific areas for potential intervention, including the inconsistent documentation of injury mechanisms (specifically repetitive overuse), the low capture rate of psychosocial obstacles, and the systematic overestimation of prognosis.
• LHS Integration: The paper provides a concrete example of the "Performance-to-Data" (P2D) and "Data-to-Knowledge" (D2K) phases of the Learning Health System cycle, showing how aggregated data can feed back into practice improvement.
• Quality Indicator Development: The study suggests that recurrence rates and prognosis accuracy are viable candidates for future MSK quality indicators, as current frameworks lack validated metrics for these specific outcomes.

Significance and Claims
The authors claim that clinical EHR data contains meaningful information that can support the identification of learning opportunities for clinicians. The significance of this work lies in shifting the focus from reactive safety monitoring (adverse events) to proactive performance analysis.

• Objective Feedback: The study argues that clinicians deserve access to objective EHR data to complement self-assessment, which is often biased.
• Safety Improvement: Integrating clinical quality measurement into routine governance is presented as a significant safety improvement for patients, ensuring care is delivered based on objective data rather than solely self-assessed competence.
• Cultural Shift: The findings support the World Federation of Chiropractic's call for the profession to embrace transparent safety data and function as a "learning profession."
• Modest Scope: The authors explicitly state that the study does not definitively determine whether findings represent true training needs, skill gaps, or organizational factors. Instead, the data serves as a signal to prompt discussions between educators, leaders, and clinicians to investigate the underlying causes of variability. The study is a step toward building systems that support clinicians in improving outcomes, rather than a final evaluation of individual performance.