Enhanced AI risk management system for importing medical products

Eduardo García Godos
GLR Abogados, Lima
egarciag@glrabogados.com

Introduction

Advanced or disruptive technology is now reconfiguring the paradigm of knowledge and its application by supplying innovative techniques that benefit the operations of border control agencies. This article outlines the benefits of leveraging artificial intelligence (AI) in risk management for the importation of medical products based on an IT system being implemented in Peru within the Single Window for Foreign Trade ('SW') platform.

The government has decided to gradually implement this system, starting with select controlled products, before expanding to encompass health regulatory bodies. Thus, this risk management system (RMS) is expected to incorporate premarket approvals for medical products at a later stage.

This analysis addresses the current Peruvian regulatory model for medical products and the legal framework for trade facilitation measures, as well as the available technological systems that would enable greater safety, effectiveness and quality in medical product import control.

Authorisation for trade and importation

Medical products require premarket approval for their import and commercialisation. Like most Latin American countries, in Peru, the medical products' regulatory model evaluates supply chain requirements and standards, as well as the product. Supply chain assessment involves the manufacture, storage and delivery processes, whereas the product's evaluation involves an analysis of supporting documents regarding its quality, safety and effectiveness. The approval timelines for the latter vary depending on the volume and complexity of information to be reviewed, ranging from 30 days to one year.

The National Regulatory Authority (NRA)^[1] verifies the product requirements' compliance; however, applications for premarket approval are processed through the SW managed by the Ministry of Foreign Trade. The SW has incorporated the product-approval procedures (Sanitary Registry) of the health authority, including those for domestic trade, whereas procedures related to the supply chain (eg, Good Manufacturing or Storage Practices standards) are processed through the NRA's system.

The inclusion of procedures for domestic trade has become a significant feature of the SW that distinguishes it from other models in the Latin American region, embracing a broader approach to the supply chain control of these products.

Unlike the authorisation of cosmetics or personal care products, which are automatically approved, the authorisation of medical products requires prior evaluation. However, the implementation of AI can now lead to the automatic approval of certain applications, effectively streamlining the process.

There is a significant delay in the submissions' approval and a lack of predictability regarding the NRA, which affects product availability in the market. For many years, the absence of structured data on applicants' risk profiles has limited the authority's ability to organise and expedite approval mechanisms based on the probabilities of whether the applicant could meet the requirements outlined in the legal framework.

Because the SW interoperates^[2] with Customs, details of the product offered for the import's authorisation are provided to the latter through data incorporated into the Customs Declaration. Leveraging this integration, risk patterns detected by the NRA are recorded by Customs, facilitating the synchronisation of the control operations of both authorities.

Risk management in international trade

Pre-market controls are expected to employ risk management criteria. Under a whole-of-government approach, the World Trade Organization (WTO) Trade Facilitation Agreement (2013) extends risk management to non-customs entities by stating that 'each Member shall concentrate customs control and, to the extent possible other relevant border controls, on high-risk consignments and expedite the release of low-risk consignments. A Member also may select, on a random basis, consignments for such controls as part of its risk management'.^[3]

There are multiple applications and manifestations of risk management applications arising from the agreement (eg, the Second Test and Authorised Economic Operator, which occur in the early stage of implementation). Furthermore, the necessity for coordinated management fosters the creation of a more integrated environment with enhanced controls.

Over the years, customs authorities have been refining their risk management techniques based on the Revised Kyoto Convention and ISO 31000:2018 guidelines, supported by increasingly sophisticated IT solutions. However, the growth of global trade, including in health products, and heightened oversight arising from a post-pandemic context, have led to improvements in risk management techniques. As such, trade facilitation measures have promoted the use of risk management in non-customs control entities.

Within the SW framework the NRA must deliver and apply risk management criteria in the evaluation of premarket approvals^[4] and inspections.^[5] In the case of the former, through these criteria, differentiated approval procedures are decided.

The so-called RMS for the evaluation of border controlling agencies procedures, as well as for inspection and surveillance in imports and exports,^[6] provides techniques to calibrate the type of assessments needed for each product; for example, a sanitary registration application deemed as low risk will have a less detailed evaluation, whereas a high-risk one will undergo a more thorough documentary review.

AI applied to risk management

Use of advanced technology for risk management

Over half of the world´s customs authorities employ advanced analytical technology, such as big data, data analytics, AI and machine learning,^[7] considering that one of its main benefits is the improvement of risk management. Such technology facilitates the detection of prohibited products or anomalies in medical products, as well as helping to identify low or high-risk applicants based on a risk score according to their commercial history. The technique employed by customs authorities is largely transferrable to health authorities' oversight, with the only necessary adjustment being the substitution of risk criteria.

AI in the Single Window

As mentioned, the SW legal framework not only enables the application of risk management for granting premarket approvals and (later on) conducting inspections but also establishes a system tailored for this purpose. In October 2023, a pilot project incorporating the RMS and an AI component was initiated. This IT solution allows, through machine learning techniques, the identification of risks during the SW procedures' evaluation. Consequently, authorities receive a risk-level notification advising them to adjust the level of scrutiny applied to the supporting documentation. The risk level, according to 'business rules', refers to the applicant's conditions and records, as well as the product's characteristics, such as its origin and source, and use. For this purpose, the RMS analyses multiple sources of information, applying deterministic, analytical and random models.

The construction of the system requires a deep understanding of the regulatory model and decisions from the NRA, who can either approve or deny an application. However, incidents (ie, observations) across application processing are likely to occur prior to such an approval or denial. The machine learning algorithm seeks to predict responses from the authority. The analytical rules provided by machine learning require data collection from the procedure. Various data groups categorised by the transaction stakeholder (ie, manufacturer, exporter and importer), product and mode of transportation are to be found across the import supply chain.^[8]

It is important to mention that the system can build analytical rules based on the supplied data. This leads to two important implications. First, it allows for the progressive improvement of analytical rules as new transactions are included. Second, it enables users knowledgeable about the business to define, build and test different analytical rules according to need and without further development effort.

This solution simplifies the internal tasks of the NRA. Assessment simplification implies that, based on sophisticated analytical rules, the system will choose applications requiring a thorough or in-depth evaluation, whereas other less complex applications can be processed more expeditiously. In some cases, premarket approvals may be granted automatically as AI-collected risk patterns render the review of submitted information unnecessary.

However, as selectivity in control is inherent to risk management, the application of the RMS does not weaken the effective controls exerted by the NRA; rather, it strengthens good regulatory oversight by providing a rigorous solution capable of analysing information and establishing relations that escape traditional methods. Nor does the NRA decline its oversight powers because it can conduct ex post controls, in which the AI provides rules to identify transactions with a higher probability of non-compliance.

Potential benefits

The RMS does not fully replace human judgement (official evaluation), but rather exponentially improves available tools for the better control of the medical product supply chain. AI provides tools for efficient risk management that must be implemented gradually. Analytical rules can be further enhanced by incorporating a greater number of transactions and collecting more cases to refine predictions. RMS will help to expedite processing, transaction transparency, improved targeting and easier access to importers. RMS can not only ease the granting of permissions but also the selection of merchandise to be inspected on entry into the country, and of course, pharmacovigilance when the product is on the market.

The sophistication of the RMS will foster the creation of a trade facilitation environment based on predictability and transparency, reinforcing health surveillance and control. It will also facilitate the identification of high-healthcare-standard compliant companies (authorised operators) and, as a result, streamline the efforts of the NRA for the effective control of the international trade of medical products.

Notes