Papers
Topics
Authors
Recent
Search
2000 character limit reached

Evaluating the Impact of Using GRASP Framework on Clinicians and Healthcare Professionals Decisions in Selecting Clinical Predictive Tools

Published 25 Jul 2019 in cs.CY | (1907.11523v1)

Abstract: Background. When selecting predictive tools, clinicians and healthcare professionals are challenged with an overwhelming number of tools, most of which have never been evaluated for comparative effectiveness. To overcome this challenge, the authors developed and validated an evidence-based framework for grading and assessment of predictive tools (GRASP), based on the critical appraisal of published evidence. Methods. To examine GRASP impact on professionals decisions, a controlled experiment was conducted through an online survey. Randomising two groups of tools and two scenarios; participants were asked to select the best tools; most validated or implemented, with and without GRASP. A wide group of international participants were invited. Task completion time, rate of correct decisions, rate of objective vs subjective decisions, and level of decisional conflict were measured. Results. Valid responses received were 194. Compared to not using the framework, GRASP significantly increased correct decisions by 64% (T=8.53, p<0.001), increased objective decision making by 32% (T=9.24, p<0.001), and decreased subjective decision making; based on guessing and based on prior knowledge or experience by 20% (T=-5.47, p<0.001) and 8% (T=-2.99, p=0.003) respectively. GRASP significantly decreased decisional conflict; increasing confidence and satisfaction of participants with their decisions by 11% (T=4.27, p<0.001) and 13% (T=4.89, p<0.001) respectively. GRASP decreased task completion time by 52% (T=-0.87, p=0.384). The average system usability scale of GRASP was very good; 72.5%, and 88% of participants found GRASP useful. Discussion and Conclusions. Using GRASP has positively supported and significantly improved evidence-based decision making and increased accuracy and efficiency of selecting predictive tools.

Summary

  • The paper demonstrates that the GRASP framework improves tool selection accuracy from 53.7% to 88.1%, standardizing evidence appraisal for clinical predictive tools.
  • The study shows that GRASP reduces decisional conflict while enhancing user confidence and satisfaction across diverse clinician groups.
  • The research outlines practical efficiency gains and future directions for automating evidence synthesis and ensuring guideline consistency.

Evaluation of the GRASP Framework's Impact on Clinical Predictive Tool Selection

Introduction

The selection of clinical predictive tools for implementation at the point-of-care or for inclusion in guidelines presents significant challenges to healthcare professionals, given the surfeit of available tools with highly variable evidence quality. The paper "Evaluating the Impact of Using GRASP Framework on Clinicians and Healthcare Professionals Decisions in Selecting Clinical Predictive Tools" (1907.11523) addresses this complexity through the development and assessment of the GRASP (Grading and Assessment of Predictive Tools) framework, an evidence-based schema aimed at standardizing and objectifying the evaluation and selection of clinical predictive models.

GRASP Framework Overview

GRASP is a multidimensional evidence appraisal system that assigns a composite grade to predictive tools based on three orthogonal features: Phase of Evaluation (pre-implementation, planning for implementation, and post-implementation), Level of Evidence (internal validity, external validity, usability, potential/realized effects, with quantification by number and quality of studies), and Direction of Evidence (positive, negative, or equivocal, determined via a protocol privileging high-quality and context-matched studies). The output is a granular yet interpretable summary that can distinguish between superficially similar tools by integrating robustness, applicability, and demonstrated effectiveness.

Methods: Experimental Design

The impact of GRASP on professional decision-making was evaluated through a controlled online experiment targeting clinicians and healthcare professionals with varying roles and expertise. Respondents were randomized into blocks representing selection tasks for well-known head injury predictive tools in both pediatric and adult emergency contexts, with and without access to GRASP evidence summaries. Decision accuracy, efficiency, subjectivity/objectivity of decision process, decisional conflict, confidence, satisfaction, and perceived usability were rigorously quantified. Statistical inference incorporated paired-samples t-tests with Bonferroni correction and effect size characterization.

Core Findings

Decision Quality Enhancement

Availability and usage of GRASP led to a substantial improvement in correct tool selection, with accuracy increasing from 53.7% to 88.1% (Δ = +64%, p<0.001p < 0.001). This effect was consistent across physician and non-physician groups, specialists and non-specialists, and participants with varying levels of familiarity, experience, and demographics.

The deployment of GRASP increased evidence-based, objective decision-making (Δ = +32%, p<0.001p < 0.001) and reduced decisions driven by guessing (Δ = -20%, p<0.001p < 0.001) or reliance on prior experience (Δ = -8%, p=0.0035p = 0.0035). Additionally, the framework significantly augmented users' confidence (+11%, p<0.001p < 0.001) and satisfaction (+13%, p<0.001p < 0.001) with their selections, reflecting lower decisional conflict.

Efficiency and Usability

While the reduction in raw task completion time (from 14.5 to 7.0 minutes on average, -52%) did not achieve nominal statistical significance due to high variance, percentile analysis (up to -48% at the 90th percentile) indicates a practical acceleration of decision-making in real-world scenarios. The System Usability Scale (SUS) score for GRASP was 72.5%, above standard usability thresholds, with especially high acceptance among users unfamiliar with predictive tools and among female participants.

Comparative and Distributive Effects

Notably, when GRASP was utilized, non-physicians outperformed physicians making unaided decisions, and non-specialists outperformed specialists. Familiarity with head injury tools, which predicted higher accuracy in the unaided scenario, ceased to be a determining factor once GRASP was applied, demonstrating the framework’s potential to democratize expertise.

Practical and Theoretical Implications

GRASP addresses a critical translational gap in the deployment of clinical predictive models by providing a standardized, interpretable, and scalable evidence grading system. This framework offers an alternative to the ambiguous and often inconsistent expert selection process, notably paralleling the operational logic of GRADE for therapeutic interventions, but tailored to model assessment and selection.

Practically, integrating GRASP into guideline development and operational workflows could result in rapid, reproducible, and audit-ready tool selection processes, reducing cognitive load and selection bias. The framework also provides a substrate for automated or semi-automated computational implementation, essential for maintaining continuous evidence synthesis given the pace of new model publication.

From a theoretical perspective, GRASP demonstrates that structured evidence triage frameworks can not only harmonize tool selection but make accuracy attainable by those without deep methodological expertise—addressing both the efficacy and equity of clinical decision support system (CDSS) implementation.

Future Directions

A primary challenge identified is the sustainability and scalability of GRASP as the evidence base expands. Maintaining an up-to-date grading database necessitates the application of natural language processing, information extraction, and possibly meta-analytic pipelines for automated evidence accrual and grade updating. Expansion of GRASP to domains beyond emergency head injury may highlight domain-specific adjustments to grading schema. Broad adoption by professional organizations would facilitate consistency across consensus guidelines and health systems.

User feedback highlighted the need for expanded reporting on tool limitations, context-specific applicability, and detailed methodological parameters; addressing these will further enhance GRASP’s utility.

Conclusion

The findings establish that the GRASP framework significantly enhances the accuracy, objectivity, confidence, and efficiency of clinicians and healthcare professionals in selecting clinical predictive tools, eliminating gaps induced by experience or familiarity differentials. GRASP operationalizes evidence-based model selection for real-world decision-making and provides a blueprint for scalable deployment of evidence triage in health informatics. Its adoption has direct implications for both the reliability of clinical guideline development and the optimization of CDS deployment in practice, though ongoing maintenance, context extension, and interface refinement remain imperative for maximal impact (1907.11523).

Paper to Video (Beta)

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Collections

Sign up for free to add this paper to one or more collections.