JMIR Publications

ABSTRACT

Background:

Background Research waste remains a major challenge to the efficiency, reproducibility and societal value of biomedical research. Concurrently, the rapid integration of Artificial Intelligence (AI) into healthcare and research, coupled with persistent deficiencies in clinical documentation quality, is reshaping the generation and use of health data. Although AI has considerable potential to enhance clinical and research processes, concerns regarding health data quality, algorithmic bias, reproducibility and scientific integrity have intensified. The interaction between AI and clinical documentation quality as a source of research waste however remains insufficiently explored from a metascience perspective.

Objective:

Objective This viewpoint examines how deficiencies in clinical documentation quality and the inappropriate or poorly governed use of AI contribute to research waste and threaten scientific integrity and real-world healthcare outcomes. Furthermore, it proposes a metascience-informed framework for strengthening trustworthy evidence generation in the era of AI-enabled healthcare.

Methods:

Methods A metascience-informed narrative synthesis was undertaken using literature from research waste, health information management, clinical documentation quality, artificial intelligence, scientific integrity, reproducibility and health data governance. Evidence from seminal and contemporary studies was integrated to explore the interrelationships among these domains and to develop a metascience-informed conceptual framework for reducing research inefficiencies.

Results:

Results Clinical documentation deficiencies, including incompleteness, inaccuracies, inconsistencies and lack of standardization, undermine the quality of health data used for secondary research and AI applications. Artificial intelligence systems trained on poor-quality datasets may amplify biases, propagate errors and generate unreliable outputs, thereby compromising reproducibility and evidence validity. The convergence of these factors creates a self-reinforcing cycle of health data degradation that contributes to research waste and threatens scientific integrity. To address these challenges, a five-pillar metascience framework encompassing clinical documentation quality enhancement, transparent and validated AI systems, health data governance, human-AI collaborative intelligence and reproducibility and open science principles is proposed.

Conclusions:

Conclusions High-quality clinical documentation and responsible AI governance are complementary pillars of trustworthy evidence generation. From a metascience perspective, reducing research waste requires integrated approaches that address both upstream health data quality and downstream analytical processes. Strengthening clinical documentation practices, promoting transparent and explainable AI and fostering reproducibility are essential for safeguarding scientific integrity and maximizing the societal value of biomedical research in the digital era.