**Transforming Real World Data into Insights**
Numerous discussions and writings have highlighted the capacity of artificial intelligence (AI) to propel and enhance health care. Although there remain several hurdles to address and inquiries to resolve, a specific domain stands out with remarkable positive outcomes.
Employing AI to decode the unstructured data derived from electronic health care records (EHR) is revealing formerly obscured insights and facilitating significant advancements in health care and clinical research.
EHR data, capturing patients’ medical histories, has long been acknowledged as an exceptional reservoir of information. De-identified to ensure patient confidentiality, this data can be mined for insights regarding disease trends, treatment efficacy, and usage patterns.
The challenge, however, lies in the fact that most information within EHRs isn’t formatted for easy querying, comparison, or analysis. Structured data—such as patient vitals, medication histories, diagnostic codes, procedure details, and lab results—constitutes only a fraction of the EHR.
The rest is unstructured, including clinician annotations, radiology reports, and imaging. This unstructured data is highly valuable; however, comprehending it on a large scale has been resource-demanding, if not unfeasible, until the rise of AI.
**Transforming Real World Data into Insights**
Advanced machine learning (ML) and natural language processing (NLP) tools can swiftly extract vital details from free-text clinical notes and images. With proper oversight and validation, this real-world data (RWD) becomes an indispensable resource for the medical field.
There is an increasing momentum towards utilizing real-world data in patient management and drug development. Recent FDA guidance on incorporating RWD from EHRs and medical claims in regulatory decisions marks a pivotal moment in the acceptance of AI-validated real-world datasets.
Clinical researchers are already harnessing AI technologies to uncover crucial insights into disease development, treatment choices, and long-term patient outcomes.
For instance, an AI-driven analysis of language patterns in EHR unstructured clinical notes detected metastatic progression that was not recorded in the patient file. This effort revealed a five-fold surge in metastatic patients for a prostate cancer investigation.
In another instance, researchers examining geographic atrophy (GA) employed AI to extract disease signals from unstructured clinical notes and eye images. This strategy increased a GA cohort by almost half a million patients and enhanced understanding of disease progression and treatment responses.
**Expanding Potential in Research and Health Care**
Through AI-enabled tools, life sciences organizations are uncovering crucial insights hidden in unstructured, real-world data, enhancing trial efficiency and speeding up drug development.
– Researchers can optimize site selection by analyzing past recruitment, patient demographics, and disease prevalence.
– They can boost patient recruitment by aligning eligible patients with trials using predictive analytics.
– They can refine study design and minimize complexity.
– They can enhance patient safety by pinpointing safety indications and adverse event trends across varying populations.
Simultaneously, AI’s capability to analyze vast data volumes has indicated the dawn of a new phase in personalized health care, featuring improved diagnostics, algorithmic predictions of patient reactions, and customized treatment strategies.
With such immense potential, it’s vital to remember that ML and NLP models are only as effective as the data that drives them. The realization of AI-enhanced health care promises relies on high-quality, thoroughly validated data curation processes.
**Essential Considerations for Success**
Guaranteeing the correctness of AI outputs begins with a thorough and stringent approach to the foundational data.
This necessitates a neutral stance regarding vendors and health care systems, concentrating on patients and the clinicians overseeing their care, rather than the EHR technology utilized to manage medical records or the facilities providing treatment.
De-identified RWD, including ongoing longitudinal records (data collected repeatedly over time from the same patients), obtained from a variety of healthcare environments and technologies, is crucial. The creation of AI-curated datasets that can drive transformative alterations in health care and research results requires a thoughtful approach and a clinically informed cooperation at every phase—from ingestion to curation to implementation.
While the “machines” are often in the spotlight in AI-driven health care innovations, effective decisions on how to curate and standardize data while preserving its original clinical context demand collaboration among clinicians, nurses, informaticians, data scientists, epidemiologists, biostatisticians, and engineers. This data must be harmonized (integrating both structured and unstructured data), and models frequently revised to avoid bias and ensure accuracy.
The effort and organizational expertise required to implement and deliver at this elevated level are considerable. However, the benefits make it worthwhile.
With appropriate measures and supervision, AI is aiding in converting disordered RWD into extraordinary insights and enhancements that can benefit everyone.
*Note: This article is authored by Sujay Jadhav, a health care executive.*