The World Health Organization refers to non-adherence as a worldwide problem of striking magnitude. Non-adherence leads to worsened medical conditions, increase comorbidity of disease, or death. Clinical studies have shown the most efective strategies for adherence are those that are individualized to the context that each patient and situation require. Yet, existing aids are relatively ridged and poorly support adaptation to individual behaviors and lifestyles. We believe our work in this space is relevant to the HEALTHI workshop. Broadly, we are eager to engage in scholarly conversations and collaborations which seek to improve the research community's ability to move beyond one-size-fits-all approaches to the design and construction of technologies to improve human health.
vate behaviors [8]. In particular, past work has shown utility is greatest in interventions Prescription pharmaceuticals are an essential that are highly customized to address the component of modern healthcare. Yet, glob- specific barriers preventing adherence. It ally, it is estimated that nearly 50% of pa- argues there needs to be a broad shift totients do not properly adhere to their medi- wards highly individualized, consumer-level cations [1]. Despite the scale of the problem, healthcare tools. the number of efective technology aids and These lessons of past technical and clinical interventions to improve adherence are quite work must guide the design and development small compared to the size of the healthcare of future health technologies; specifically the technology market. The lack of successful need to employ methods that seek solutions technology is not for lack of research, as our that easily integrate and support highly specommunity has deep contributions to study cialized, personal behaviors. Towards this and technology development [2, 3, 4, 5, 6, goal, we believe the maturity and accep7]. All prior technology investigations agree tance of in-home sensing, machine learning, with the clinical research, which show adher- and lightweight input/output devices (such ence is highly dependent on individual situ- as smart speakers) present contemporary opations and sentiments that define and moti- portunities to develop novel approaches to personalized adherence aids.
gated patients’ comprehension of medications’ purpose and dosing, as well as how their care providers participate and influence this comprehension. As expected, interviewees provided a strong understanding of There are already many examples of machine why they were taking their prescribed medilearning and sensing technologies that sup- cations and what the resulting health benefits port personalized interventions. For exam- should be; topics that were clearly discussed ple, wearables can provide notifications and as part of the patient-provider interaction. reminders, smartphone apps can help users Less clear was when medications should be catalog and organize medication dosing in- taken, how they should be monitoring interstructions, and smart pill bottles can accu- actions and side efects, and how to improve rately monitor for when medications were their adherence to the medications; topics taken. These existing solutions, though, do typically not germane to patient-provider not ofer the promise of true personalized conversations or to the role of providers. behaviors change. Sending a notification, Through these interviews, we also preor capturing time of doses, are not sufi- sented interviewees with several hypothetcient for sustained behavior change. These ical technologies, including those that used technologies must integrate across dimen- sensing and machine learning to share inforsions of technology innovation; which in- mation about patients with providers. In parclude cost, scalability, inclusion, accessibility, ticular, we described a technology that would and eficacy while also spanning stakehold- analyze behaviors pertaining to when mediers; which include individuals, family and cations were taken, where they were stored, friends, community, health systems, and gov- and a persons activities within their home. ernment policy. We explained this data would be collected,
Over the past year, we have been inves- summarized, and presented to providers to tigating opportunities for technology and seed discussions on more eficient strategies community integration using sensing, ma- for adherence, including changing dosing, chine learning, and lightweight I/O. We medication selection, or behavior modificabriefly describe these eforts as evidence of tions. Reaction to these technologies was our ability to contribute to a rich discus- mixed, but many respondents indicated a desion of advancing health-based interfaces sire to use the technology, as they found it and technologies at the workshop. would help providers understand that they are trying and to accelerate recommenda2.1. Provider Involvement tions and changes that were specific to their individual struggles. In essence, they saw the technology as useful in cutting through the boilerplate recommendations and feedback to engage with providers on personal terms.
Truly hyper-personalized technologies are those that support and strengthen the connection between a patient and his/her/their providers. We have conducted over twenty in-depth, multi-hour interviews with hypertensive patients to better understand adherence practices and how those practices inlfuence the interaction with the healthcare providers. In our interviews we investi2.2. Proactive, Not Reactive Aids A significant proportion of the personalized health technologies are designed to provide reactive opportunities for users. For instance, be limited in reach to achieve user trust. health tracking apps can keep track of steps taken or calories burned, and provide those 2.3. Aids Which Promote summaries to user for them, in turn, to trans- Independence late the knowledge into action. Similarly, medication tracking applications and smart Increased technology interventions, espepill bottles can inform users when they miss cially those that use pervasive sensing or a dose; again, providing knowledge that must those that communicate and report to care be turned into action. Future technologies, providers travel a thin line between enwith the capabilities of pervasive sensing and abling and dictating a patient’s indepenadvances in behavior modeling, should lever- dence. Our investigations have proved that age the constant stream of information about personal health maintenance is an extraordithe user to be proactive, rather than reactive narily private activity for most people. When in design. asked in our interviews if other members
In our interviews, described above, we also of their home help them to remember to explored user reaction to technologies that take their medications, we often received rewere proactive in design. Some were very sponses to indicate such dependence was not simple; for instance, using proximal location preferred, as it was perceived as an unnecand knowledge of a pending dose to nudge a essary burden on the other household memuser to take their medication when they hap- ber and a admission of not being able to hanpen to be near the bottles, rather than at the dle the responsibility on one’s own. This was scheduled time when it may be physically in- true for living situations from roommates to convenient. Other hypothetical technologies intimate partners in decades-long marriages. were more complex, suggesting locations for As we look forward to design technologies medication storage and times of dosing based that provide significant levels of personalizon physical, historical behaviors in the home. ing through both technology pervasiveness For instance, suggesting medication be kept and increase integration with daily activities in the kitchen and taken in the morning be- and health provider interactions, we must be cause the user consistently makes cofee in keen to do so from a perspective of preservthe morning. In this case, the technology ing individual independence. can proactively help users form medication habits consistent with their activities of daily 2.4. Leveraging Existing, living. Pervasive Technology
Feedback on these approaches were positive, but with concerns about overreach. In Important to the vision of proactive, indeparticular, integrity of the data and how and pendence promoting aids, is the ability to who it would be shared with was a consistent leverage new forms of pervasive technoloconcern. Interviewees expressed this con- gies for patient interaction. We believe the cern in way that were less about outsiders application of these technologies remains unwatching their activities, but about whether derstudied. One such example is the success this data would be used to negatively impact of short message service reminders in delivtheir care. For instance, sharing information ering health-based reminders. While more with insurance providers or employers. The basic than current technology, this method results of these early insights point to the of notification nearly always finds improved value of proactive, AI-driven tools; but must patient medical compliance [9]. While this method is successful in engage patients, its national Conference on Systems, Man, traditionally static deployment can be made and Cybernetics, 2011, pp. 3213–3218. more dynamic to proactively time notifica- doi:10.1109/ICSMC.2011.6084164, tions relevant to behaviors related to medical zSCC: 0000013 ISSN: 1062-922X. adherence. [4] L. G. Dalgaard, E. Grönvall, N. Verde
In addition, we believe these existing ap- zoto, Accounting for medication proaches can be used in combination with particularities: designing for everyday other emerging technologies, such as voice medication management, in: Proceedactivated technology assistants and smart ings of the 7th International Conference speakers. Industry reports find that nearly on Pervasive Computing Technolo20% of households own at least one smart gies for Healthcare, PervasiveHealth speaker1. The use of these technology to ’13, ICST (Institute for Computer Scimonitor and improve health behavior is lim- ences, Social-Informatics and Telecomited compared to the vast literature explor- munications Engineering), Venice, ing other internet of things devices, partic- Italy, 2013, pp. 137–144. URL: http: ularly within the design space of proactive, //doi.org/10.4108/icst.pervasivehealth. dynamic engagements with patients; either 2013.252058. doi:10.4108/icst. alone, or with their providers. In particular, pervasivehealth.2013.252058, we believe this approach can create engage- zSCC: 0000026. ments with aids that can improve and sup- [5] F. Nunes, G. Fitzpatrick, Understandport patient health behaviors over the long ing the Mundane Nature of Self-care: term. Ethnographic Accounts of People Living with Parkinson’s, in: Proceedings of the 2018 CHI Conference on Human References Factors in Computing Systems - CHI ’18, ACM Press, Montreal QC, Canada, [1] W. H. organization, Adherence to long- 2018, pp. 1–15. URL: http://dl.acm.org/ term therapies: evidence for action, citation.cfm?doid=3173574.3173976.
Geneva, 2003. doi:10.1145/3173574.3173976, [2] L. Palen, S. Aaløkke, Of pill boxes and pi- zSCC: 0000024.
ano benches: "home-made" methods for [6] J. Pater, S. Owens, S. Farmer, E. Mymanaging medication, in: Proceedings of natt, B. Fain, Addressing medication the 2006 20th anniversary conference on adherence technology needs in an aging Computer supported cooperative work, population, in: Proceedings of the CSCW ’06, Association for Comput- 11th EAI International Conference on ing Machinery, Banf, Alberta, Canada, Pervasive Computing Technologies 2006, pp. 79–88. URL: http://doi.org/ for Healthcare - PervasiveHealth ’17, 10.1145/1180875.1180888. doi:10.1145/ ACM Press, Barcelona, Spain, 2017, 1180875.1180888, zSCC: 0000136. pp. 58–67. URL: http://dl.acm.org/ [3] D. Asai, J. Orszulak, R. Myrick, C. Lee, citation.cfm?doid=3154862.3154872.
J. F. Coughlin, O. L. d. Weck, Context- doi:10.1145/3154862.3154872, aware reminder system to support med- zSCC: 0000006. ication compliance, in: 2011 IEEE Inter- [7] K. Stawarz, A. L. Cox, A. Blandford, 1https://techcrunch.com/2020/03/24/infinite-dial- Don’t forget your pill!: designing 2020/ efective medication reminder apps that support users’ daily routines, in: Proceedings of the 32nd annual ACM conference on Human factors in computing systems - CHI ’14, ACM Press, Toronto, Ontario, Canada, 2014, pp. 2269–2278. URL: http://dl.acm.org/ citation.cfm?doid=2556288.2557079. doi:10.1145/2556288.2557079, zSCC: 0000132. [8] J. G. Hugtenburg, L. Timmers, P. J. Elders, M. Vervloet, L. van Dijk, Definitions, variants, and causes of nonadherence with medication: a challenge for tailored interventions, Patient Preference and Adherence 7 (2013) 675–682. [9] F. J. Schwebel, M. E. Larimer, Using text message reminders in health care services: A narrative literature review, Internet Interventions 13 (2018) 82–104.
URL: http://www.sciencedirect.com/ science/article/pii/S2214782918300022. doi:10.1016/j.invent.2018.06. 002, zSCC: 0000055.