Acceptability of and Willingness to Take Digital Pills by Patients, the Public, and Health Care Professionals: Qualitative Content Analysis of a Large Online Survey

Background: Digital pills are pills combined with a a a patch connected a when pills ingested. care professionals patient data from digital pills online via their own interface, them to check whether a patient took the drug. Digital pills were developed for the stated goal of improving treatment adherence. the first digital pills 2017, the manufacturer 2020 of insufficient evaluation. Objective: As recommended for the evaluation of health technologies, this study assesses the prospective acceptability of and willingness to take digital pills among patients, the public, and health care professionals. Methods: Participants were patients who were receiving long-term for a chronic condition, public participants (both groups recruited from a representative sample), and health care professionals. Participants answered 5 open-ended questions regarding the acceptability of digital pills and 1 close-ended question regarding the willingness to take digital pills, which were developed in a preliminary qualitative study. We explored the 5 theoretical dimensions of acceptability by performing an abductive qualitative content analysis of all free-text responses. We assessed data saturation with mathematical models. We fitted a multivariate logistic regression model to identify the sociodemographic and health characteristics associated with the willingness to take digital pills. Results: Between January 29, 2020, and April 18, 2020, 767 patients, 1238 public participants, and 246 health care professionals provided 11,451 free-text responses. We identified 98 codes related to the acceptability of digital pills: 29 codes on perceived clinical effectiveness (eg, sensor safety cited by 66/2251 participants, 29.5%), 6 on perceived burden (eg, increased doctors’ workload, 164/2251 participants, 7.3%), 25 on perceived ethicality (eg, policing, 345/2251 participants, 15.3%), 30 codes on perceived opportunity (eg, exclusively negative perception, 690/2251 participants, 30.7%), and 8 on affective attitude (eg, anger, 541/2251, 24%). Overall, 271/767 (35.3%) patients, 376/1238 (30.4%) public participants, and 39/246 (15.8%) health care professionals reported willingness to take digital pills. This willingness was associated with male sex (odds ratio 1.98, 95% CI 1.62-2.43) and current use of a connected device to record health settings (with a dose–response relationship). Conclusions: The prospective acceptability of and willingness to take digital pills were limited by clinical and ethical concerns both at the individual and societal level. Our results suggest that digital pills should not be considered a mere change in the form of drug administration but a complex intervention requiring specific evaluation before extended use in clinical routine practice as well as an ethical and legal framework to ensure safe and ethical collection and use of health data through a patient-centered approach.


Introduction
More than one-third of people worldwide live with at least 2 chronic conditions; hence, an increasing number of people are receiving long-term treatment [1]. Control of many of these conditions depends on adherence to treatment, which can be defined as the degree to which the use of medication by the patient corresponds with the prescribed regimen [2]. However, about half of patients with chronic conditions do not take their medications as prescribed. Poor treatment adherence leads to a range of negative consequences such as poor control of disease, life-threatening events, treatment resistance, and increasing costs [3][4][5][6][7][8]. In 2003, the World Health Organization suggested that improving treatment adherence would have a greater impact than any improvement in specific medical treatments, and this has led to a continuous search for effective interventions to monitor and improve medication adherence (eg, blood drug dosage, self-reported patient diaries, smart pill dispensers) [2,9,10].
The most recent monitoring tool, approved in 2017 by the US Food and Drug Administration, is called a "digital pill." Digital pills are pills combined with a sensor, which sends a signal to a patch connected to the smartphone of the patient when the pill is ingested. Health care professionals and caregivers can access patient data (eg, adherence, activity, heart rate) from digital pills online via their own interface, thus allowing them to check whether a patient took the drug. The sensor is an ingestible event marker made from a copper-magnesium pair of electrodes within a silicon insulating skirt disk 5 mm in diameter and 0.3 mm thick. On contact with gastric fluid, the sensor sends a unique digital code to allow identification and timestamping of the medication and dose form [11]. The first digital pill was an antipsychotic (aripiprazole) used for schizophrenia or bipolar disorder treatment, which was rapidly followed by a range of other digital pills for multiple medical conditions (eg, antihypertension, diabetes, viral diseases) [12].
Digital pills have been developed for the stated goal of improving medication adherence and optimizing treatment management; these are used to identify nonadherence to first-line therapies before more expensive second-line therapies can be considered [10,13]. The European Medicines Agency declared that the evaluation for the digital pill submitted was inadequate and withdrew the application on July 26, 2020; the quality of the available studies (observational studies or small trials) and the lack of clinical relevance of the outcomes (short-term outcomes related to adherence to digital pills, safety outcomes restricted to tolerance of the patch) were points of criticism that led to this withdrawal [12,[14][15][16][17].
Moreover, guidelines for the evaluation of health technologies recommend taking into account patient and public perspectives, which has not been done yet for digital pills [18]. Acceptability is a multifaceted construct that reflects the extent to which people delivering or receiving a health care intervention consider it appropriate based on anticipated, present, or retrospective responses to the intervention [19]. It includes the extent to which the intervention is perceived likely to achieve its purpose (perceived clinical effectiveness), the amount of effort required to participate in the intervention (perceived burden), the extent to which the intervention aligns with individual values (perceived ethicality), the extent to which benefits/values must be compromised to engage in the intervention (perceived opportunity), and how people feel about the intervention (affective attitude) [19].
As part of new health technology assessment, this study aims at evaluating the prospective acceptability of and willingness to take digital pills by 3 subpopulations (patients, public participants, and health care professionals).

Methods
We conducted an online survey by recruiting a representative sample of the French general population, dichotomized into patients and public participants, and an additional convenience sample of health care professionals. The survey included 5 open-ended questions related to the acceptability of digital pills and 1 close-ended question regarding the willingness to take digital pills (Multimedia Appendices 1 and 2).

Participants and Recruitment
Patients were adults (≥18 years old) with at least one self-reported chronic condition for >6 months and who were prescribed long-term treatment for >1 month. Public participants included adults (≥18 years old) who did not meet the eligibility criteria to be classified as patients. Health care professionals were professionals prescribing, delivering, or monitoring pharmaceutical treatments (physicians, nurses, pharmacists, and midwives). All participants were living in France, spoke French, and had access to the internet.
We used the 2 distinct methods to recruit participants: • The representative sample was recruited using a quota sampling method in the IPSOS Access Panel [20]. Quotas were based on the known profile of the French population in terms of sex, age, socioprofessional category, and population density of the residential area. This sample was secondarily classified into patients and public participants according to the eligibility criteria. We calculated weights with the rim weighting method-raking (Multimedia Appendices 3) [21].
• The additional convenience sample of health care professionals was recruited via an online campaign by sending emails to professional associations and posting advertisements on social networks (Multimedia Appendices 4).
All participants provided informed electronic consent before enrollment. The Institutional Review Board CERAPHP.5, Paris, France, gave ethical approval (IRB0001198).
To recruit patients with various combinations of chronic conditions and socioeconomic status, perform quantitative analysis using logistic regression, and ensure that data saturation was reached based on previous experience with an online survey and considering the fact that 34.5% of the French population lives with a chronic condition, we targeted a representative sample of 2000 participants [22,23]. Regarding the convenience sample of health care professionals, we decided to stop recruitment after reaching 95% data saturation [22].

Survey Development and Data Collection
We conducted a preliminary qualitative study to draft the questions (wording, formatting, and content) of the survey and provide the investigators with a contextual framework for analyzing and interpreting textual data. A researcher trained in qualitative methods (AC) conducted face-to-face collaborative interviews with 5 patients, 5 public participants, and 3 health care professionals (Multimedia Appendix 5). From the results of this preliminary study, we developed an illustrated explanation of digital pills, 5 open-ended questions on the acceptability (Textbox 1) and 1 close-ended question on the willingness to take digital pills ("Would you agree to use this device for yourself?") (Multimedia Appendices 1 and 2).
Textbox 1. Open-ended questions to all participants on the acceptability of digital pills.

•
What do you think of this device (transmitter + Bluetooth patch + smartphone app + access to data)? We want to know your immediate reaction: try to write down all the ideas that came to you when you first saw this device.
• What do you envisage are the positive aspects of this device (transmitter + Bluetooth patch + smartphone app + access to data)? For example, in what situation(s)/for whom could it be useful, what kind of benefit could it bring?
• What do you envisage are the negative aspects of this device (transmitter + Bluetooth patch + smartphone app + access to data)? For example, what drawbacks or risks do you foresee?
• If your doctor suggested a treatment using this device (transmitter + Bluetooth patch + smartphone app + access to data), what would your reaction be? What would you think of his/her approach?
• The manufacturer says that this device will allow for greater consistency between the prescription and the actual taking of medications. In your opinion, why would a person take their treatment more consistently if they were equipped with this device?
In addition, we collected sociodemographic data (eg, sex, age, level of education) for patients and public participants, professional status data (profession and duration of professional experience) for health care professionals, and health characteristics for all participants (eg, chronic condition, current use of a connected device to collect health data) (Multimedia Appendices 1 and 2).
The survey for patients and public participants was available online in a dedicated IPSOS platform. The survey for health care professionals was available in a LimeSurvey form. Both surveys were tested online for usability, clarity, and wording in a convenience sample of 10 patients, 20 public participants, and 5 health care professionals, whose feedback helped improve the final version.

Acceptability of Digital Pills
We conducted a multiple-round qualitative content analysis of each free-text response by following a 3-step approach: inductive open coding of all units of meaning related to the acceptability of digital pills, inductive condensation of the codes, and inductive development of the themes and deductive classification of the themes in the 5 dimensions of acceptability [19,24].
In the first step, the team of coders (AC, AF, AJ, FD) read and reread the responses at the individual level (all responses to all questions individual by individual) and at the question level (all responses for all individuals question by question). The process of coding identified in each response units of meaning and assigned a code. A unit of meaning was defined as any expression in the text related to the acceptability of digital pills, as explained by Sekhon et al [19]. The coding was limited to the manifest content of the responses, except for the first open-ended question on the emotional reaction to the digital pill, for which the latent emotional content was interpreted when possible [25]. During a face-to-face meeting, all researchers coded together all responses for 100/2005 (5%) individuals of the representative sample and 50/246 (20%) individuals in the sample of health care professionals. Then, 1 researcher coded all responses (AC) and 3 others double-checked the codes (AF, AJ, FD). In case of discrepancies, consensus was reached by discussion with another researcher (SS). To reduce interpretation bias, coders had different backgrounds and training (psychiatry, family medicine, training in clinical research, social science, cognitive psychology, and medical ethics).
The second and third steps occurred in a meeting in which the 4 coders and SS inductively condensed all the codes using semantic, psychological, and anthropological considerations. Then, they categorized the codes inductively into themes and classified them deductively into the 5 theoretical dimensions of acceptability [19].
We then calculated the number of citations for each code identified with the qualitative content analysis in each group using the weighted data for patients and public participants.
Data saturation was assessed for patients, public participants, and health care professionals with a mathematical model [22].

Willingness to Take Digital Pills
We calculated the number and proportion of participants who agreed to take digital pills. We estimated odds ratios (ORs) and 95% CIs using the weighted data of the representative sample to calculate the association of the willingness to take digital pills with all socioeconomic and health characteristics. We used the same data for a multivariate logistic regression model, including sex, age, socioprofessional status, residential area, chronic condition, current use of a connected device to collect health data, and ease to talk about treatment with the doctor. For the regression, we considered the response "I do not wish to answer" to be a refusal to take digital pills. The significance threshold for P values was calculated with Bonferroni correction to correct for multiple testing.
All quantitative analyses involved using the free software R v3.3 (R Foundation for Statistical Computing), and the package "survey."

Data Sharing Statement
The deidentified quantitative data set will be shared on request (astrid.chevance@gmail.com).The textual data set will not be shared because of ethical restrictions.

Participants
Among patients, 396/767 (51.6%) were women and the mean age was 54.6 (SD 15.6) years (Table 1). Hypertension was the most reported chronic condition (169/767, 22%), followed by diabetes (135/767, 17.6%) and chronic pain (81/767, 10.6%) (Multimedia Appendix 6). Do not wish to answer a Chronic condition was defined as at least 1 long-term treatment and 1 self-reported condition. The most common chronic conditions were hypertension, diabetes, chronic pain, thyroid disease, heart disease, asthma, and dyslipidemia. All chronic conditions are presented in Multimedia Appendix 6. b N/A: not applicable.
A total of 99/767 (12.9%) patients and 120/1238 (9.7%) public participants reported checking health data recorded by a connected device daily, and 524/767 (68.3%) patients and 902/1238 (72.9%) public participants reported rarely checking or never checking data (Table 1)   b For clarity, we report the results for only the 3 most prevalent medical specialties. The other medical specialists were anesthetists, intensive care unit physicians, emergency physicians, child psychiatrists, surgeons, obstetricians-gynecologists, hematologists, neurologist, gastroenterologists, dermatologists, endocrinologists, oncologists, pneumologists, infectious diseases specialists, rheumatologists, nephrologists, pediatric care specialists, palliative care specialists. c Chronic condition was defined as at least 1 long-term treatment and 1 self-reported condition.
The qualitative content analysis of the 11,451 open-text responses identified 98 codes related to acceptability. Data saturation was reached for patients (99.4%), public participants (100%), and health care professionals (97.7%) (Multimedia Appendix 8). We classified the 98 codes into the 5 theoretical dimensions of acceptability: perceived clinical effectiveness (29 codes, 29%), perceived burden (6 codes, 6%), perceived ethicality (25 codes, 25%), perceived opportunity (30 codes, 30%), and affective attitude (8 codes, 8%) [19]. We report the most-cited codes for each dimension with the proportion of citation and participants' quotes in Table 3. Figure 2 and Multimedia Appendix 9 report all 98 codes.  In Figure 2, which shows the acceptability of connected drugs, the dots represent codes regarding acceptability of treatments identified by the qualitative content analysis of the responses for 767 patients, 1238 public participants, and 246 health care professionals. The width of the line is proportional to the number of spontaneous citations of the codes by the representative sample of patients and public participants (N=2005). The overarching categories in bold correspond to the 5 dimensions of the theoretical model of acceptability by Sekhon et al [19].
Perceived clinical effectiveness is the extent to which the intervention is likely to achieve its purpose. Perceived burden is the amount of effort required to participate in the intervention. Perceived ethicality is the extent to which the intervention fits well with individual values. Perceived opportunity is the extent to which benefits/values must be given up to engage in the intervention. Affective attitude indicates how people feel about the intervention.

Perceived Clinical Effectiveness
Participants reported perceived benefits ( Participants described the pressure of surveillance as a disciplinary mechanism based on social desirability, fear of being reprimanded, fear of social disqualification as a "bad patient," and fear of shame or guilt. Therefore, participants raised concerns about a policing patient-doctor relationship and the meaning of care in the future.  Figure 3). Complete analysis of other characteristics associated with the willingness to take a digital pill is presented with crude and adjusted ORs and 95% CIs in Multimedia Appendix 10. The forest plot (Figure 3) presents the ORs with their 95% CIs and the P value obtained with the logistic regression analysis of the association of the willingness to take connected drugs by sex, age, profession, population density of the residential area ("residential area"), chronic condition, frequency of health data recording with a connected device, and ease to talk about treatments with the doctor. The plot shows the number of participants per category (n) willing to take a connected drug among the total number of participants in the given category (N). The corrected threshold for statistical significance is P=.003. For the continuous variable "age," we present the mean age of the whole sample and the standard deviation. Figure 3. Determinants of the willingness to take connected drugs among the representative sample of patients and public participants (N=2005). The second column present the number of people willing to take connected drugs (n) over the total number in the patient and public sample (N), except fot the variable "Age", presenting the mean and standard deviation (sd) of people willing to take connected drugs. inh.: inhabitants.

Principal Findings
Overall, two-third of patients and public participants in this study refused to take digital pills. Their arguments were grounded in the fear of possible serious clinical and ethical harms. Health care professionals largely refused (177/246, 72%) to take digital pills citing disagreement with their professional deontology and ethics. Participants who accepted digital pills had positive a priori for technology considered as progress: people agreed to take it in the framework of research for a limited investigation time to help medical progress.
Participants who refused to take connected drugs, reported that these would be useful for other people, such as individuals with cognitive impairment, individuals with chronic conditions or mental disorders, and those incarcerated, highlighting the view that potential clinical or ethical harms would be acceptable for these already vulnerable people. Given the condition of the designated populations, the intervention may be stigmatizing and may increase vulnerability.
To our knowledge, this study is the first to explore the acceptability of connected drugs in line with the recommendation that health technology assessment take into account patient and public perspectives [18]. We used a representative sample of the French population, thus allowing for generalizable results in the subsamples of patients and the public. In France, medication for chronic conditions is fully reimbursed, so when participants answered the questionnaire, they were not prematurely restricted by individual economic concerns but could detail their preferences. To limit the risk of capturing only a "hot reaction" and to mimic the process of a rational and balanced deliberation, we used 5 open-ended questions and asked the close-ended question about willingness to use connected drugs at the end of the survey. Only 16.7% of patients and public participants found no cons and 31% found no pros.

Limitations
The first limitation is the restriction to only 1 country as a consequence of the choice to recruit a representative sample. Because culture and health care system may affect the acceptability of digital pills, further studies are required in different settings.
A common limitation of using an online survey is the risk of selection bias because of the need for internet access. However, the target population for the digital pill must have access to smartphones and the internet. While the use of an online survey may have overestimated the acceptability of digital pills in the general population, it may not have done so in the target population. Moreover, to limit potential selection bias due to reading level, we created the questionnaire during a qualitative preliminary study and tested it for clarity.
We could have increased the reliability of the results by verifying the final codes and themes with the participants, but we minimized the risk of misinterpretation by limiting the condensation of codes to synonyms only, halting the inductive interpretation at a very basic level-which led to 98 codes in the final results-and using the participants' wording.
Another limitation is that the sample of health care professionals was not representative of the larger population of health care professionals.
Finally, we only investigated the prospective acceptability because digital pills were not available in France at the time of our study. Studies about concurrent acceptability (while taking connected drugs) and retrospective acceptability (after taking connected drugs) may be complementary and lead to different results.

Comparison With Prior Work
Our results are consistent with the concurrent and retrospective acceptability discussed in 18 observational studies and trials of digital pills, which involved small samples of 5 to 129 patients [26]. Moreover, some participants who refused digital pills for themselves found them useful for other people considered vulnerable (eg, people with cognitive or mental disorders or those incarcerated), which would lead to a risk of stigmatizing the use of digital pills.
This study added further ethical concerns to the well-known topics raised in the scientific literature (eg, consent, confidentiality, privacy) [12,[26][27][28][29][30]. This shows the substantial importance of including patient and public perspectives in health technology assessment, as is already recommended but marginally done, with the risk of neglecting social, ethical, and political aspects of these technologies that are critical determinants of treatment effectiveness in real life settings [18]. A further strength is the use of a representative sample, which allows for generalizable results.
This survey contributes to advanced scientific and clinical reflections about medication adherence. As underlined by participants, digital pills seem to be a naïve solution considering the complexity of what determines medication adherence and nonadherence [12,31,32]. Digital pills appear to be burdensome and expensive reminders with an apparent contradiction-whether unintentional or intentional-between their complexity and the ability of those needing support for adherence. Particularly in the case of intentional nonadherence, digital pills would be either useless or efficient at the cost of intrusiveness, policing, and negative effects on the patient-doctor relationship.
Finally, digital pills could be useful to assess treatment efficacy in explanatory trials, but there are 2 prerequisites for extending their use in clinical practice. First, digital pills should be evaluated as a complex intervention according to specific standards and not as if they are a mere change in the form of drug administration [17,33]. Second, there is urgent need to develop an ethical and legal framework to ensure the safe and ethical collection and use of health data through a patient-centered approach [12,27,28].

Conclusion
Our results suggest that patients, the public, and health care professionals view connected drugs not as a promising new medical device to better monitor medication adherence but as a complex intervention with a possible impact on patient-doctor relationship and patient autonomy. The participants also raised additional concerns about burden of treatment, cost-effectiveness, and privacy, which need to be addressed in further investigations. Future studies should take into account the views of all stakeholders, ie, patients, potential prescribers, as well as health regulatory authorities and researchers, to ensure the safe and ethical collection and use of health data through a patient-centered approach.