AI-powered health technology assessment Qorus-NTT DATA Innovation in Insurance Awards 2026

Medical schemes and health insurers in South Africa face escalating pressure from rising healthcare costs, rapid medical innovation, an increasing chronic disease burden, and heightened regulatory scrutiny. Health Technology Assessment (HTA) sits at the heart of responsible funding, impacting benefit design and reimbursement decisions. It determines whether new medical technologies, including pharmaceuticals, medical devices, diagnostics, and surgical procedures should be covered, and under what conditions. Yet the process remains highly manual, time-consuming, and resource-intensive, creating bottlenecks that delay member access to innovative therapies and strain operational capacities.

We developed an AI-powered HTA platform purpose-built for public health systems financing, payers and insurance environments that automates the full health economic assessment process – from evidence screening and appraisal, through report generation and decision scoring. The platform has reduced assessment timelines from weeks to minutes, while improving consistency, transparency, and scalability of funding decisions, accelerating member access to high-value care, and delivering material financial, clinical and governance impact.

THE BUSINESS CHALLENGE

The volume of applications from pharmaceutical manufacturers and device suppliers seeking funding approval continues to grow, both for new technologies and new indications for existing technologies, including pharmaceuticals, medical devices, diagnostics, and surgical procedures. At the same time, medical schemes (health insurers) operate under intense pressure to contain accelerating healthcare inflation, while needing to fund high-cost therapies responsibly, delivering value-based benefit design, and balancing member affordability with clinical excellence and positive outcomes. Each new technology application for funding must be rigorously assessed before a funding decision can be made. Discovery Health medical scheme receives on average around 140 applications for funding that need to be reviewed each year by highly specialized experts.

The Traditional HTA Process

The traditional HTA process places an enormous burden on health economic researchers. When a manufacturer (supplier) applies for a funding decision, they submit an extensive application form alongside supporting clinical evidence. In theory, this evidence should consist of high-quality clinical trials demonstrating the effectiveness, safety, and comparative value of the technology. In practice, the quality varies dramatically – submissions frequently include opinion pieces, conference abstracts, marketing materials, or documents not directly relevant to the technology under review.

A single application frequently includes over forty evidence documents, mostly in PDF format. The researcher must open, read, and critically appraise each of these before they can determine whether the evidence base is adequate to proceed, or whether they need to request higher quality evidence from the supplier. This process is inefficient and often creates a huge delay between receiving the application and the first contact back to the supplier. A researcher may spend days reviewing a large submission only to discover that most of the evidence is suboptimal, and the supplier needs to resubmit additional evidence. The full HTA process, which includes synthesising the evidence, populating a structured HTA report, cross-referencing regulatory databases, and generating a funding recommendation, can take several weeks for a single technology review. This means that if a single researcher were to review all 140 applications, it would take approximately eight years to complete the assessment of just one year’s submissions. This inevitably results in delays for members of society that would and could benefit from some of these innovations sooner had the assessments been reviewed quicker. Conversely it also allows for a comprehensive affordability assessment to meet budgets and contain cost pressures.

The Resulting Constraints

These limitations result in:

• Long assessment cycles (often stretching to several weeks per technology)

• High operational costs

• Inconsistent evidence appraisal and assessment outcomes, compounded by manual data extraction that is prone to human error

• Reduced responsiveness to clinical innovation

• Suboptimal benefit design agility

• Limited capacity to scale as submission volumes increase

• Delayed access to beneficial care and therapies for potential patients and societies

• Potential incomplete holistic economic evaluations for population and health systems

As the pace of medical innovation accelerates and submission volumes grow, this traditional model is structurally unsustainable.

THE SOLUTION – AN AI-POWERED HTA PLATFORM

The web-based AI platform was built and designed to support HTA researchers through the complete assessment workflow. The platform does not replace the researcher’s judgement; it augments their capacity by automating the most time-consuming and error-prone elements of the process, allowing them to focus on critical appraisal and informed decision-making. The platform has two core functionalities: the HTA Report Generator and the Decision Table Generator.

Functionality 1: HTA Report Generator

The health economic researcher begins by populating metadata fields and uploading the supplier’s application form and accompanying evidence.

Step 1: Automated Evidence Screening

The platform uses a Large Language Model (LLM) to intelligently analyse and extract key data from each evidence document. The extracted details are presented in a structured table within the interface, giving the researcher immediate visibility of the quality and relevance of submitted evidence.

Step 2: Report Generation

From the evidence screening table, the researcher selects documents that represent sufficiently high-quality evidence to inform the assessment. Only the selected evidence documents are carried forward into the report, ensuring the final output is not diluted by poor-quality submissions.

Traditionally, what follows is the most time-consuming phase of the entire HTA process. The researcher must manually extract and recreate structured information from the application form, an administrative burden that adds little analytical value but consumes a significant amount of time. They must then critically appraise each selected evidence document, synthesising findings across multiple clinical trials that may differ in design, population endpoints and reporting conventions. Registration and regulatory information must be manually verified by searching external databases, such as the FDA or EUDAMED registers, and additional knowledges bases, including the NICE guidelines and the Cochrane library of systematic reviews, must be searched to determine whether independent clinical guidance exists.

Finally, all this information must be compiled into a coherent, structured report following a predefined template. Depending on the volume and complexity of the submitted evidence, this process can take weeks, or even months, for a single technology.

The HTA AI platform automates this entire process by orchestrating multiple AI and data retrieval processes:

• Natural language processing (NLP): Extracts structured data directly from the application form (normally in PDF format), including costing information, comparator technologies, and registration details.

• Retrieval-augmented generation (RAG): Draws directly on the selected evidence documents to answer critical assessment questions, covering the technology details, safety profile, clinical benefits, and detailed overviews of each clinical trial.

• External verification: Employs real-time web scraping to cross-reference regulatory databases (e.g., FDA, EUDAMED) and consults external knowledge sources (e.g., NICE, Cochrane) to additionally inform the recommendation.

• Synthesised conclusion: Aggregates all findings to produce a structured conclusion and preliminary guidance on the funding recommendation.

The final output follows a predefined HTA template, providing a ready-to-use working document for the researcher to review, validate and finalise in a matter of minutes.

Functionality 2: Decision Table Generator

The second core functionality automates generation of the Decision Table, a structured framework used to systematically evaluate clinical trial evidence. The decision table requires the researcher to extract a set of predefined clinical endpoints from a trial – including primary and secondary endpoints, survival data, hazard ratios, mortality figures, and adverse events – for both the technology under review and the comparator. These data points are then used to calculate an internal score that informs the funding recommendation.

Manually populating these tables is labor-intensive and requires the researcher to extract and copy detailed statistical information from PDF documents into structured Excel spreadsheets. This process is prone to transcription errors, particularly when trials report complex, multi-arm, or subgroup analyses. The HTA AI platform transforms this process by using an LLM to extract these endpoints and feeding them directly into the internal scoring framework, removing the burden of manual entry and providing the researcher with an instant, calculated score for validation.

Where the traditional process required days of manual review before the researcher could determine the adequacy of the evidence, the platform delivers that insight in under one minute, regardless of whether the submission contains one or over a hundred documents.