ADVANCING THE ASSESSMENT OF DIGITAL HEALTH INNOVATION - LESSONS DRAWN FROM A CATALOGUE OF AI USE CASES
Author(s)
Teresa Barcina, MSc, PharmD1, Afua van Haasteren, PhD2, Christine Muzel, MSc3, Melike Deger Wehr, BSc, MSc4, Miray Aibibula, Senior Market Access Manager5, Nancy van Lent, Director Health Economics6, Vito Parago, MSc7.
1Health Technology Assessment, Industrial Affairs, MedTech EUrope, Bussels, Belgium, 2Roche Diagnostics International, Rotkreuz, Switzerland, 3Philips, Eindhoven, Netherlands, 4ResMed, Münich, Germany, 5Siemens Healthineers, London, United Kingdom, 6Medtronic, Brussels, Belgium, 7Johnson & Johnson MedTech, Milan, Italy.
1Health Technology Assessment, Industrial Affairs, MedTech EUrope, Bussels, Belgium, 2Roche Diagnostics International, Rotkreuz, Switzerland, 3Philips, Eindhoven, Netherlands, 4ResMed, Münich, Germany, 5Siemens Healthineers, London, United Kingdom, 6Medtronic, Brussels, Belgium, 7Johnson & Johnson MedTech, Milan, Italy.
OBJECTIVES: The rapid advancement of digital medical devices and artificial intelligence (AI)-based technologies poses significant challenges within existing health technology assessment (HTA) frameworks. This research has two objectives: to identify limitations in current frameworks when applied to AI, and foster understanding of evidence gaps highlighted by assessors.
METHODS: A catalogue of AI use cases was developed building on a literature search conducted by The Austrian Institute for HTA (AIHTA), with a focus on CE-marked AI applications developed as clinical decision support systems (CDS) to inform screening and diagnosis. To establish relevant comparisons, cases were selected if more than one assessment had been conducted by different HTA agencies in a 5-year timeframe (2020-2025). The scope of the catalogue was informed by the volume of physician-facing innovations developed in this field, expected to have organizational impact and lacking access pathways.
RESULTS: Our analysis of assessments conducted for over 15 CDS systems across six jurisdictions has revealed variability in assessment outcomes despite similar data packages, which may reflect differences in interpretation and evolving evaluation criteria. The uncertainty regarding evidence standards and how to address assessors’ expectations extends beyond the first assessment, as requests for further evidence generation are not accompanied by guidance. Moreover, positive assessment outcomes do not clearly translate into adoption. Limitations in HTA processes, such as the difficulty keeping pace with technological advances and the reliance on measures of clinical benefit, are especially apparent in assessments of evolving AI-based technologies designed to deliver value and through improvements in operational efficiency.
CONCLUSIONS: Current approaches to assess AI-based technologies struggle to capture their value proposition and adaptive nature. To drive the path forward, this research underscores the importance of enabling dialogue between industry, assessors and end-users at all stages of the assessment process, including the prioritisation of assessment topics and the definition of the assessment scope.
METHODS: A catalogue of AI use cases was developed building on a literature search conducted by The Austrian Institute for HTA (AIHTA), with a focus on CE-marked AI applications developed as clinical decision support systems (CDS) to inform screening and diagnosis. To establish relevant comparisons, cases were selected if more than one assessment had been conducted by different HTA agencies in a 5-year timeframe (2020-2025). The scope of the catalogue was informed by the volume of physician-facing innovations developed in this field, expected to have organizational impact and lacking access pathways.
RESULTS: Our analysis of assessments conducted for over 15 CDS systems across six jurisdictions has revealed variability in assessment outcomes despite similar data packages, which may reflect differences in interpretation and evolving evaluation criteria. The uncertainty regarding evidence standards and how to address assessors’ expectations extends beyond the first assessment, as requests for further evidence generation are not accompanied by guidance. Moreover, positive assessment outcomes do not clearly translate into adoption. Limitations in HTA processes, such as the difficulty keeping pace with technological advances and the reliance on measures of clinical benefit, are especially apparent in assessments of evolving AI-based technologies designed to deliver value and through improvements in operational efficiency.
CONCLUSIONS: Current approaches to assess AI-based technologies struggle to capture their value proposition and adaptive nature. To drive the path forward, this research underscores the importance of enabling dialogue between industry, assessors and end-users at all stages of the assessment process, including the prioritisation of assessment topics and the definition of the assessment scope.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MT47
Topic
Health Technology Assessment, Medical Technologies, Organizational Practices
Topic Subcategory
Digital Health
Disease
Cardiovascular Disorders (including MI, Stroke, Circulatory), Gastrointestinal Disorders, No Additional Disease & Conditions/Specialized Treatment Areas, Oncology