Transforming Healthcare with Artificial Intelligence: Global Insights

Introduction

Artificial intelligence (AI) in healthcare holds immense potential to enhance patient outcomes, safety, and affordability of high-quality care. However, the journey to fully realise this is complex. This article published npj Digital Medicine explores the perspectives of the Future of Health (FOH) group, an international community of health leaders. These leaders highlight four key action areas to effectively channel AI’s capabilities.

Improving Data Quality to Power Artificial Intelligence

AI algorithms rely heavily on the accuracy and completeness of the data they process. High-quality data is essential for AI tools to function correctly and deliver reliable results. FOH members emphasise that identifying and ensuring the availability of high-priority data elements is vital. These elements include data with the most predictive value, high patient value, and importance for performance and bias analyses.

Policy incentives play a significant role in encouraging high-quality data collection. Aligned payment incentives for healthcare quality and safety, along with performance standards for data reliability, completeness, and timeliness, are essential. For example, the multinational project STANDING Together aims to improve data quality and representativeness of data for AI tools.

Building Trust and Verifying Artificial Intelligence Tools

Demonstrating that AI tools are both effective and safe within specific patient populations is paramount to building trust in AI. However, this is challenging as AI’s performance can vary significantly across different sites and over time due to changes in health data patterns and population characteristics.

Real-world evaluations are crucial for long-term predictions and preventing long-term complications. Health systems must prioritise implementing data standards and infrastructure that facilitate retraining or tuning of algorithms, test for performance and bias, and ensure scalability across the organisation. Meanwhile, efforts such as Health AI: The Global Agency for Responsible AI in Health and the Coalition for Health AI aim to build and certify validation mechanisms to ensure AI’s trustworthiness.

Sharing Data for Better Artificial Intelligence

High-quality internal data alone is insufficient to power and evaluate all AI applications. Effective AI-enabled predictive software for clinical care requires interoperable data across health systems to build a diverse picture of patient healthcare. FOH members recommend that healthcare leaders work with researchers and policymakers to connect detailed encounter data with longitudinal outcomes. This approach helps assure valid outcome evaluations and addresses potential confounding and population subgroup differences.

For example, the South African National Digital Health Strategy outlines interventions that improve adopting digital technologies while complying with the 2013 Protection of Personal Information Act. The country has made progress in building a Health Patient Registration System and releasing a Health Normative Standards Framework to improve data flow across institutional and geographic boundaries.

Incentivising Progress for Artificial Intelligence Impact

Financial incentives are necessary for accelerating the development, evaluation, and adoption of AI tools. There is a need for value-based payments, aligning financial incentives for high-quality data, effective AI tools, and improved patient outcomes and healthcare operations.

Value-based payments, which focus on quality, safety, better health, and lower costs for patients, are essential. These payments rely on high-quality, standardised, interoperable datasets from diverse sources. Within value-based payments, data are critical for measuring care quality and patient outcomes, adjusted for factors outside clinical control. This alignment of incentives promotes high-quality data collection and development of AI tools that genuinely benefit patients and healthcare systems.

Conclusion

Data has become the most valuable commodity in healthcare. The actions of healthcare leaders and policymakers in the coming years is the key to enhancing health outcomes, safety, affordability, and equity. By improving data quality, building trust, sharing data, and incentivising progress, the healthcare industry can truly take advantage of AI’s potential.