In the EU, software is defined as a medical device, while in the US, it is not. In vitro diagnostic devices are a separate entity in both the EU and US. The definition of a medical device in the EU focuses on human use, while in the US, it covers both human and veterinary use. The intended purpose of the device also plays a role in its definition.
In-House exemptions in the UK can only be applied to diagnostics, not devices. It allows for the provision of a service within an NHS trust to private patients or other trusts without the need for certification. However, if the service is provided to the trust's own patients or if it is sold to a company, certification is required. The definition of a manufacturer of medical devices includes those who place something on the market, and the NHS trust is considered a separate legal entity.
Software can be classified as a medical device if it is specifically listed in the MDR or if it influences the use of a device. The level of classification is based on whether it informs, drives, or treats clinical management. Informing clinical management involves simple calculations, while driving clinical management involves real-time actions based on the software's calculations. Treating clinical management involves diagnosing conditions and applying therapy.
According to the EU, if a product is intended to be used for informing, driving, treating, or diagnosing a medical condition, and it could be any type of similarity act or anything where you're actually claiming to treat and improve someone's mental health condition, it is considered a medical device.
Formative usability studies involve gathering user feedback on various aspects of the software, such as font size, language, and flow, while summative usability studies involve testing the software with a representative cohort of intended users to evaluate its effectiveness in terms of risk mitigation and user training.
The EU classifies medical devices into three categories (Class I, Class IIa, and Class IIb) based on their intended use and the level of risk they pose. The US system is slightly different, using Class I, Class II, and Class III instead of Roman numerals. The classification determines the level of regulatory control required for the device.
Such software is not considered a medical device, but it has its own set of rules and regulations, including the European Union Medical Device Regulation, USFDA, and MHRA in the UK.
The UK regulations are expected to turn more into a classification similar to the US and EU, with class one being less common. This change is due to the difficulty in developing and deploying medical devices in the UK, considering the higher class requirements for trading in the EU and US.
The classification depends on the intended use, the user, and the potential for foreseeable misuse. For example, if the application gathers symptoms from patients remotely for GP receptionist appointment booking purposes, it might fall under a higher class due to the potential impact on patient prioritization.
The fundamentals in the ulatory pathway include defining the intended use of the device, documenting a regulatory strategy, determining the device class, establishing a quality management system, and creating technical documentation.
Verification ensures that the product meets the specified requirements, while validation ensures that the product is fit for its intended use. Verification checks if the jacket has two holes for arms and buttons, while validation checks if the jacket fits and is suitable for the purpose it was intended for.
Post-market surveillance is important to monitor the performance of the device in the market, gather feedback from users, identify bugs or fixes, and ensure that the device does not fall victim to the failures of competitors. It involves proactive and reactive vigilance, problem resolution processes, and impact analysis.
TRL is a concept developed by NASA to understand how mature technology was for the space program. It is a scale from one to nine, with one being very basic proof of principle and nine being full production use.
The UK responsible person is a concept introduced by the UK MDR, and is the MHRA's interface for contacting a manufacturer abroad. It is not necessarily the importer or distributor.
Corrective action is taken after a problem has occurred, while preventive action is taken before the problem even arises.
A notified body is responsible for giving CE marking or UKCA respectively, an approved body can only grant ISO certification but cannot do CE or UKCA, and a certification body can only grant ISO certification. Notified and approved bodies are typically used for medical devices, while certification bodies are typically used for component manufacturers.
A stage one audit is an adequacy check of a company's quality management system, resources, and people, while a stage two audit is a final official audit where a company's records are closely examined to ensure compliance with their procedures. Major and minor non-conformities can be issued during both stage one and stage two audits.
Under rule 11 of the EU MDR, nearly every software medical device is classified as a class 2A device, which is a higher risk classification than under the old EU MDD or UK MDR. This means that software medical devices will require a full quality management system and will be subject to more stringent market surveillance and periodic safety update reports.
Under the EU MDD, there are very few class 1 software as a medical device due to the strict rules regarding making claims or providing information used to make decisions with diagnostic or therapeutic purposes. However, under the UK MDR, software as a medical device may be classified as class 1 if it does not provide such information. Mental health apps that deliver CBT are an example of this.
Under the UK MDR, companies are responsible for post-market surveillance and clinical follow-up, which comes at a cost. This is to ensure the safety and effectiveness of medical devices on the market.
The classification of a mental health app under the EU MDR depends on several factors, including the intended use and patient population. For example, if the app suggests actions to a patient with suicidal ideation that could cause them to attempt to take their own life, it could be classified as class 3.
Parental consent is important for medical devices if the user is a child. The age limit for when parental consent is needed may vary, but it should be clearly stated in the instructions for use. This is crucial for the backbone of the medical device's documentation.
The intended use profile for a software medical device should be defined clearly and specifically, taking into account that healthcare professionals have varying jobs and qualifications. If the device requires specific training or technical knowledge, it should be clearly stated. The audience and required training should be clearly identified in the instructions for use.
Software medical devices can be classified into three buckets in terms of AI and decision support: assistive, augmentative, and autonomous. Assistive tools help a clinician make a decision, augmentative tools add to the clinician's existing information, and autonomous tools can diagnose on their own. It is important to be careful with claims and understanding where the product is going currently and where it is intended to go in the future.
A product backlog contains all problem reports and feature requests, while a release backlog contains the items that will be included in a specific version of the product, such as version 1.1.
The different tiers of changes in agile language are tier three (small changes to the platform), tier two (changes to the requirements), and tier one (potential revalidation of features or user interface). Tier zero is a significant change that may require a new submission to regulatory bodies.
The impact of changing the clinical language and performance of AR models underlying a chatbot on the tier of change depends on the specifics of the change. A change impact analysis should be conducted to determine if it is a tier zero, tier one, or tier two change.
The speaker is referring to technology readiness levels one to three.
The bigger issue is how generalizable the model is from the demographic it was collected from to the intended patient population, which may require using another database. Additionally, the speaker mentions the need to consider consent and the intended use statement.
Data should be collected from the place of intended use, according to the speaker.