MyRx | Devpost

The Problem

Today, the US increasingly finds itself lagging behind the healthcare systems of its wealthy peers. Among high-income countries, the US spends the most on healthcare as a percentage of GDP, but finds itself performing among the lowest outcomes. Yet, one important source of these costs is misprescription of drugs, which can cause dangerous, and at times life-threatening, adverse inter-drug reactions. In fact, 6.7% of hospitalized patients have a serious adverse drug reaction, resulting in a fatality rate of 0.32%, or about 106,000 deaths annually. Compounding this trend is the fact that more drugs are being used to treat patients than ever before, given that the rate of adverse drug reactions increases exponentially after a patient is on four or more medications (Woosley et. al.). It is therefore an ethical imperative to pursue solutions to this significant—yet largely preventable—source of death.

Paradoxically, it is also easier than ever for doctors to access information about potentially dangerous drug interactions. The digitization and centralization of vast amounts of drug information into government databases has made this knowledge publicly accessible to healthcare officials nationwide. Yet, Woosley et. al. found that even in cases where computerized drug interaction software was in place, pharmacists still filled prescriptions for drug combinations with known interaction risks (Woosley et. al.). The problem is thus not the availability of information surrounding drug interactions in general, but rather the uniform incorporation of this information in conjunction with specific patient information across the healthcare continuum. Relatedly, Jane Barnsteiner writes of the importance of “medication reconciliation,” or the process of comparing patient prescriptions to their actual medication intake, in preventing prescription-related errors. Crucially, Barnsteiner observes that a multitude of factors, including patients’ lack of knowledge of their own medications and a lack of integration of patient health records, all hinder the achievement of medication reconciliation, increasing the probability of error (Barsteiner). As a result, we believe that solutions to the problem of adverse reactions must also attend to the needs of patients to ensure that they have the agency to make safe and informed medical decisions.

Status Quo

Though drug interaction information has long been digitized, we find that existing tools intended to assist in preventing adverse drug reactions are inadequate to address the reconciliation gap. In particular, many software applications are designed for usage by doctors and healthcare workers, which renders them inaccessible to the general public. Aside from the prohibitively high cost that these software tools charge, the information returned from these tools is often difficult to interpret because it requires advanced medical knowledge and familiarity with technical terms. This context becomes especially problematic when considering that the elderly are among the most vulnerable to adverse drug reactions, due to the high degree of polypharmacy and advanced medical conditions among older populations (Zerah et. al.). In short, existing tools largely exclude patients from their purview, leaving them dependent on an error-prone medical system to meet their prescription needs.

Moreover, even when tools designed for patients do exist, they replicate many of the same problems as doctor-facing tools or return untrustworthy results. Websites such as WebMD’s drug interaction checker, for instance, require users to select from a long list of variations of brand-name drugs, which could prove difficult given that many patients lack detailed knowledge of their current medications. On the other hand, AI tools like OpenAI’s ChatGPT can generate more comprehensible answers to simple user queries, but have no way to guarantee that such information will be medically accurate. Furthermore, AI-driven solutions are not yet accessible to users on a mass-market basis and therefore also require some level of technological proficiency to use.

The Solution

It is at the intersection of doctor and patient-facing solutions to the harm of drug interactions that MyRx sits. MyRx aims to mitigate the risks of serious drug interactions by providing patients with accessible information about medications they are already taking. We utilize a dual function pipeline that combines the medical accuracy of doctor-facing software with the accessibility of AI-driven models. Based on user input, MyRx first queries a government NIH drug information database to identify any potential drug interactions. Relying first on this NIH database ensures that the drug interactions returned to the user are accurate and consistent with the most up-to-date government data possible. However, this list of drug interactions may still contain confusing medical terminology for many users. Therefore, MyRx then augments this initial data with an AI-driven language model to offer clarifying definitions and interpretations to the user where needed. Narrowly tailoring the AI model to information returned by the database constrains the AI to strictly interpreting reliable data, thereby minimizing the risk of misinformation. The app also incorporates other useful features such as OCR image-to-text upload, a printable medication report, and accessible website design, all of which decrease prerequisite knowledge and increase accessibility to everyday users.

Additionally, we envision a wide variety of beneficial use cases that bridge gaps in the current healthcare system. As previously mentioned, Woosley et. al. document that pharmacists may fill combinations of drugs even, even when it is known that those drugs may lead to potentially dangerous interactions. In these and similar situations, MyRx bridges this crucial gap by allowing patients the independence to investigate these interactions and take their health into their own hands. However, MyRx will be especially impactful in contexts outside the pharmacy and doctor’s office, including in cases of over-the-counter medication. Research routinely finds that people often take over-the-counter products with potentially dangerous interactions with other drugs and lack awareness of these serious risks (Tarn et. al.). Alternatively, patients may also elect not to disclose certain medications to their doctor due to stigma associated with types of medication, such as birth control. Such users may face economic, social, or personal barriers to gaining quick medical access, making accessibility incredibly important for extra-medical contexts. By supplementing more costly medical advice with quick yet reliable information, MyRx enables potentially life-saving interventions against otherwise common drug reactions.

Ethics

As with any project that has to do with patient safety, technology, and healthcare information, there are many associated ethical concerns. We consider the most important issues below and discuss how we worked through each of them.

Data Privacy

Given the potentially sensitive nature of the healthcare data that MyRx processes, it was important to us to protect the privacy of patient data in-app. The app comes at a particularly appropriate time in the wake of the Supreme Court’s decision overturning Roe v. Wade, after which many scholars warned that digital healthcare information could be used to prosecute women for healthcare-related crimes (Citron). It thus becomes especially likely that patients would decline to share reproductive medication history with their doctor, making MyRx an important tool for enabling reproductive safety and autonomy so long as data is kept secure. In the app, we took several steps to maximize user agency and safety with respect to data. First, MyRx collects minimals amounts of data, limited only to the medications the user has entered. Second, MyRx never saves patient data to a database; all data is saved on users’ own devices via local cookies. While we recognize that it may still be possible for law enforcement to obtain warrants for users’ browsing history, but local data storage still presents a significant security improvement and higher legal threshold for such access cases compared to data that is stored server-side. Furthermore, to that end, we have also built in full user control over local data storage by giving users the option to immediately halt cookie storage and delete any current cookies.

Medical Skepticism

Another potential ethical concern is that MyRx may inadvertently increase the patient-doctor barrier by fostering medical skepticism of prescription instructions. Yet, it may also be the case that such medical skepticism falls well-within patient rights to make decisions about their own care. The above reproductive data privacy case, for instance, provides one such example where patients might justifiably withhold and refuse medical treatment for legal and personal reasons. We might thus frame such a concern as a fundamental tension between care and consent: on the one hand, doctors have an ethical duty to “do no harm” and provide care to the best of their ability, but on the other hand, patients equally have an ethical right to refuse such care.

In approaching this dilemma, we drew upon the current AMA standard of informed consent, which embodies the idea that patients have the right to receive information to make informed decisions about their own medical care ("Informed Consent"). Under this model of patient care, the role of the doctor is not to impose a standard of care onto patients, but rather to provide the requisite medical information to make informed decisions, guiding them through the decision-making process. Importantly however, the ultimate decision as to whether to accept or reject care rests with the patient.

Under a model of informed consent, MyRx serves as a useful corrective to medical advice by giving patients access to reliable information in the event of doctor or pharmacy errors. For this reason, we notify every user that MyRx is not intended to serve as a replacement for medical advice, but rather as a potential starting point for further conversations with verified healthcare professionals. At the very least, however, we believe that MyRx will not contribute uniquely to existing medical skepticism. Claggett et. al., for instance, find that “[p]atients with a greater degree of medical skepticism are more likely to follow web-based medical advice, regardless of the website traits" (Claggett et. al.). In other words, though not widely accessible, tools already exist that enable highly skeptical individuals to seek out information to support their skepticism, irrespective of MyRx. It therefore seems that our app will not uniquely enable such skepticism, and in our opinion, its life-saving benefits provided to traditionally excluded users outweigh such potential costs. In this view, MyRx might even help to reduce the prevalence of healthcare misinformation compared to tools like ChatGPT by providing more accurate information that has been verified with government sources.

Funding

Furthermore, funding for our app would also be a potential source of ethical concerns. Once MyRx scales to a large number of users, traditional funding sources would each present their own set of ethical concerns for patient healthcare. For instance, running ads on the platform would undermine the model of informed consent by allowing pharmaceutical companies to unduly influence the medical decisions of patients. Alternatively, charging patients to use our app would contravene the accessibility of the app and make it difficult for low-income populations. In keeping with our mission of accessibility, we thus plan to initially launch MyRx as a free app, and continue to keep costs minimal as the app scales. Unfortunately, it is clearly not viable to forgo receiving funds due to the necessary costs of engineering, server upkeep, and other maintenance costs. Therefore, we will primarily explore methods of applying for public grants as a nonprofit organization to fulfill our mission of serving the public. If insufficient, we may then begin to charge larger organizations, such as hospitals and pharmacies, to integrate app functionalities into their record systems. Our overarching goal in charging users will not be to generate profit from the app, but instead only to maintain a quality level of service for patients. Such decisions will be subject to conflict-of-interest scrutiny as well as user feedback from hospitals, patients, and other stakeholders.

Linguistic Exclusion

Finally, we recognize that due to our dependence on the NIH database, MyRx is inherently limited by the limitations of the database itself. For example, the database itself is only written in American English, which excludes not only different drug names from other countries, but also non-native English speakers more generally. Our own group members had experienced situations where family members failed to take medication as prescribed due to unknown fears about interactions with drugs from other countries. On top of this fact, since databases largely only contain information about prescription drugs, there is the risk that many traditional herbal remedies could interact with prescription drugs in unknown ways (Woosley et. al.). As a group whose members grew up in non-English speaking households, we were especially concerned about issues of linguistic accessibility to non-English speaking populations. While we were not able to adequately address this issue within the two-day timeframe, we agreed that future features of the app would include support for foreign drugs and non-English speakers. In particular, we hope to expand the databases queried to include those from foreign countries, include herbal remedies, and provide in-app translation services for such users.

Works Cited

Barnsteiner, Jane H. “Medical Reconiliation.” Patient Safety and Quality: An Evidence-Based Handbook for Nurses. Agency for Healthcare Research and Quality, Apr. 2008, https://www.ncbi.nlm.nih.gov/books/NBK2648/.

Citron, Danielle Keats. “The End of Roe Means We Need a New Civil Right to Privacy.” Slate, 27 June 2022, https://slate.com/technology/2022/06/end-roe-civil-right-intimate-privacy-data.html?via=rss.

Claggett, Jennifer et. al. “The Effects of Website Traits and Medical Skepticism on Patients’ Willingness to Follow Web-Based Medical Advice: Web-Based Experiment.” Journal of Internet Medical Research, vol. 24, no. 2, 24 Feb. 2022, doi:10.2196/29275.

“Informed Consent.” AMA Code of Medical Ethics, n.d., https://code-medical-ethics.ama-assn.org/ethics-opinions/informed-consent.

Tarn, Derjung M. et. al. “Prevalence and Knowledge of Potential Interactions Between Over-the-Counter Products and Apixaban.” Journal of the American Geriatrics Society, vol. 68, issue 1, Jan 2020, pp. 155-162, doi: 10.1111/jgs.16193.

Woosley, Raymond L. et. al. “Preventable Adverse Drug Reactions: A Focus on Drug Interactions.” U.S. Food and Drug Administration, 6 Mar. 2018, https://www.fda.gov/drugs/drug-interactions-labeling/preventable-adverse-drug-reactions-focus-drug-interactions.

Zerah, Lorène et. al. “Prevalence of drug-drug interactions in older people before and after hospital admission: analysis from the OPERAM trial.” BCM Geriatrics, vol. 21, no. 571, 18 Oct. 2021, doi: 10.1186/s12877-021-02532-z.