Artificial intelligence and machine learning-based technologies have the potential to transform healthcare by deriving new and important insights from the vast amount of data generated during the delivery of healthcare every day. The challenges for clinical pharmacy practice include discovering how to apply these technologies in ways that reveal new patterns in health data that make a real difference for patients.
Artificial Intelligence (AI) and machine learning-based technologies have the potential to transform healthcare because they offer new and important insights derived from the vast amount of data generated during the delivery of healthcare every day. The capacity of AI to learn from real-world feedback and improve its performance makes this technology uniquely suited as Software as a Medical Device (SaMD) and is responsible for it being a rapidly expanding area of research and development. Clinical pharmacy practice may undergo major change due to the implementation of this technology. The challenges facing clinical pharmacists include discovering how to apply AI technologies in ways that reveal new patterns in health data that make a real difference to clinical practice. AI is expected to assist healthcare professionals in enhancing patient experiences and health outcomes, augmenting the health of the population, reducing costs and improving the interventions that pharmacists and other providers instigate with patients. Thereby, the use of AI could help clinical staff to provide more informed medication-use decisions and improve outcomes.
However, healthcare professionals must ensure that there is evidence indicating that any new SaMD to be implemented and all AI implementations are safe and effective before they are put into use in practice. Whether or not a Health Technology Assessment (HTA) process is performed, pharmacists clearly play a critical role in helping to generate the evidence that is needed to inform decisions concerning how and when to implement AI on a widespread basis in routine clinical pharmacy practice.
Not all software used in the healthcare settings is considered to be a medical device. However, depending on its functionality and intended purpose, software may fall within the European Union (EU) definition of a “medical device”. The EU and the United States both have their own criteria for identifying healthcare and medical devices, although both definitions are the result of a purpose-based approach. AI software properly classified as a medical device must comply with the rules that aim to ensure its safety and level of performance. Given the capacity of AI to capture various forms of personal data, cybersecurity will also become very important to ensure the sustainability of this technology, including periodic reviews of the internal processes, to make sure it fulfils the requirements for the protection of privacy. In the EU, for example, the processing of personal data is governed by the General Data Protection Regulation (GDPR). Meanwhile, in the United States, regulatory issues may arise for AI developers based on the intended use of the product. Once a product is classified as a medical device, its class will define the regulatory requirements applicable for FDA clearance or approval. Regardless of the classification of the product, however, AI developers will need to assess whether the HIPAA (Health Insurance Portability and Accountability Act) rules apply, as well as any design controls and postmanufacture auditing that also apply in terms of cybersecurity.
The traditional paradigm of medical device regulation was not designed for adaptive AI technologies, which have the potential to adapt and optimise device performance in real-time to continuously improve healthcare for patients.
Online pharmacy activities and telepharmacy lead to a new style of relation between patients, doctors and other healthcare professionals. New opportunities and threats will arise due to the disruption of the digital revolution, and these affect medication safety with a specific focus on what pharmacy can contribute in a rapidly changing healthcare sector. Moreover, the application of AI will also be advanced by assisting and monitoring patients and their needs in the absence of clinical professionals (i.e., chatbots).
While in the ambit of diagnosis AI will provide considerable assistance with very rich databases, in therapy (and particularly in drug therapy) schemes and alternatives will be hierarchised. Medical treatment decisions may see less of an impact on them because they have to be very close to the patient, but AI and robots will evolve too. However, maybe the greatest potential for disruption from AI in clinical pharmacy will come from the discovery and application of patterns that matter in practice and can better inform pharmacist’s decisions and make a real difference to clinical practice.
By using a large number of Electronic Health Record (EHR) data and AI to learn patterns concerning appropriate use of medication, software could become able to detect and alert to instances when a prescribed drug seems to deviate from its pattern of appropriate use. Moreover, AI could help in drug selection decisions, indicating, through automated classification, which patients would not be likely to experience particular adverse effects from a specific drug.
Medication Therapy Management (MTM) and pharmacokinetic-guided dosing is standard practice in the clinical management of narrow therapeutic index drugs, and AI may eventually be used to help guide decisions on dosage in real time.
In recent years, clinical pharmacy practice has had to deal with the global problem of medicine shortages, which means taking inventory management decisions. Managing medicine shortages and ensuring continuity of supply can result in significant amounts of the time and attention of a clinical pharmacist being diverted from other important tasks in the provision of high quality, safe and efficacious care. AI could predict medication use in hospitals and health systems more accurately, as well as providing support for clinical decisions when exploring treatment alternatives if a drug is not available.
Polypharmacy is common in older adults and younger at-risk populations, and it increases the risk of adverse medical outcomes. The optimisation of the medicines clinical pharmacists deal with can provide a review of medications which thereby optimises cost-effectiveness and clinical use of medications, which when aligned with patient preferences should contribute to improved health outcomes. In this area, AI could provide new tools for understanding drug–drug interactions and associated mechanisms, as well as predicting alternative drugs for intended clinical use that avoid negative health effects.
The objective of pharmacovigilance is to detect, monitor, describe and prevent adverse drug events. AI is needed to analyse the large number of data collected through post-marketing studies, EMRs records and the Internet. Clinical pharmacists have an opportunity to lead the expansion of AI into pharmacovigilance, using entirely new skill sets in this discipline.
In addition, AI may facilitate communication between healthcare professionals and patients by decreasing processing times, thereby increasing the quality of patient care. Many healthcare procedures have a costly bureaucratic burden, which digitalisation by itself has not reduced. Moreover, scheduling conflicts or overbooking is another important issue that AI can solve. AI could prioritise appointment scheduling based on the risk of readmission and overall severity of an illness with the aim of reducing readmissions.
With AI, data analysis becomes more attractive and pragmatic. Reviewing drug healthcare outcomes in the real world could require permanent analysis of the application of existing therapeutic options to approved indications or to a new disease. In drug repurposing, this is advantageous because the new drug which is already approved can go directly to phase II trials for a different indication without having to pass through phase I clinical trials and toxicology testing again. While upstream from clinical practice, the progress of AI in the life sciences has been notably faster, with many peerreviewed publications. Drug discovery is being revamped through the use of AI at many levels, including sophisticated natural language processing searches of the biomedical literature, data mining of millions of molecular structures, the design and making of new molecules, predicting off-target effects and toxicity, predicting the right dose for experimental drugs, and developing cellular assays at a massive scale.
The work on developing AI algorithms to enable patients and care takers to take their healthcare into their own hands has lagged that for clinicians, pharmacists and health systems. Wearables and health apps can evolve from playing a passive role or recording patient data to giving standard advice to patients, from diagnosis to drug treatment adjustments.
Most common chronic conditions, such as hypertension, depression, and asthma could theoretically be managed with virtual coaching through AI devices such as chatbots. Crucially, if AI is going to make health professionals better at caring for patients, the datasets being used must be representative of the whole of society and not be biased in terms of sex, race, ethnicity, socioeconomic status, age, ability or geography. This need for representation is not only a data science issue, it is also an ethical one. In the absence of equal representation, discrimination and injustices have occurred.
However, although studies have demonstrated that AI can perform on a par with clinical experts in disease diagnosis, most of the tools involved have not been evaluated in controlled clinical studies to assess their effect on healthcare decisions and patient outcomes. Inconsistent data quality and a lack of clarity with regard to the effective integration of AI into clinical workflows are significant issues that threaten its application. Whether AI will ultimately improve the quality of care at a reasonable cost remains an unanswered, but critical, question. Therefore, AI tools have to be implemented with caution. Furthermore, not all challenges require AI solutions, as statistics and database research can often provide a perfect solution and may be easier and less expensive to implement.
While the application of AI in pharmacy practice faces several challenges, we should accept the fact that is needed. Challenges such as new pharmaceutical policy initiatives, the regulation of data protection and cyber security, the debate on unusual accountability and responsibility issues, questions about the fiduciary relationship between patients, and medical AI-based devices could all be approached in such a way as to reflect the ethical standards that have guided other actors in healthcare and solutions should be held to those same standards.