As health care embraces digital transformation, one truth is clear: AI’s success depends on data quality. No matter how advanced the algorithms, poor clinical data limits their effectiveness.

AI in health care: Promise meets reality

The health care technology landscape is abuzz with large language models (LLMs) and conversational AI applications that show remarkable capabilities in synthesizing patient information and streamlining clinical workflows. Clinicians are hopeful that these tools can reduce documentation burden and enhance decision support.

However, as implementations multiply across health systems, a fundamental challenge is becoming increasingly apparent: AI-assisted documentation tools can produce impressive outputs, but these are only as reliable as the data fed into these systems. This reality is becoming a central concern for health care technology leaders.

The data quality challenge

Several recurring concerns are emerging from technology leaders and clinicians alike:

Foundation of flawed data: Many health care organizations are building AI initiatives on data repositories filled with inconsistencies, errors, and gaps.

Increased workload paradox: Rather than reducing clinical work, some AI implementations are inadvertently creating additional tasks as clinicians spend time verifying and correcting AI-generated content.

Structured data deficiencies: While conversational AI excels at generating narrative text, it often fails to create the structured clinical data necessary for regulatory compliance, quality metrics, and analytics.

Interoperability obstacles: Despite advances in data exchange standards like FHIR, the “garbage in, garbage out” problem persists when systems share problematic data.

The cost of poor data quality

The financial and clinical implications of data quality issues are substantial. Poor clinical data quality can lead to significant challenges, including:

• Medical errors and financial risk
• Denied claims from coding errors
• Barriers to interoperability and care continuity
• Inefficient clinical workflows
• Reduced ability to capture proper reimbursement
• Limited effectiveness of clinical decision support

Toward solutions: Addressing the data quality problem

Health care organizations are beginning to address these data quality challenges with focused approaches. Based on industry discussions, several key strategies are emerging:

Data validation and normalization

Organizations need solutions that can validate clinical data and resolve issues stemming from bad mappings, duplicate or incorrect items, and inadequate codes. This involves using advanced technologies to process structured, semi-structured, and unstructured data.

Clinical terminology enhancement

Solutions are needed to address inconsistencies in local codes and legacy systems, including:

• Inconsistent mappings across systems and localities
• Custom terms that are not properly validated or maintained
• Historical concepts insufficient for current care coordination
• Terminology standards that are not adequately maintained and updated

AI and evidence-based algorithms

By combining AI technologies with evidence-based algorithms, health care organizations can work toward normalizing historical data, matching related diagnoses, recategorizing inappropriate items, and fixing inadequate or missing codes.

Strategic implementation roadmap

The health care industry stands at a critical inflection point. As organizations race to implement AI solutions, those who address the fundamental data quality issues first will see the greatest returns on their investments.

What is needed is a focus on transforming problematic clinical data into a trusted source of truth—creating the foundation needed for AI to truly transform health care delivery.

Before health care organizations can fully realize the potential of AI, they must solve their data quality challenges—generating text alone is not enough. The industry needs solutions that transform conversational interactions into high-quality, structured clinical data that can be trusted for patient care, compliance, and analytics.

By addressing these foundational issues, the industry can build AI systems that clinicians can trust, that improve patient outcomes, and that deliver on the promised efficiency gains. The future of health care AI depends not just on sophisticated algorithms but on the quality of data those algorithms use.

Jay Anders is a physician executive.