Use of appropriate cardioprotective medication is a cornerstone of cardiovascular disease (CVD) prevention, but less-than-optimal patient adherence is common. A meta-analysis and systematic review of 44 prospective studies assessing nearly 2 million participants found that only 60% of patients had good adherence (defined as ≥80%) to CVD medications, and lack of adherence was strongly linked with adverse clinical outcomes [1]. Reasons for nonadherence are multilayered and include patient-, health care provider-, and system-related factors, and thus strategies for improving adherence are recommended to adopt a multifaceted approach [2].

The crucial first step in this process is an accurate assessment of the patient’s medication-taking pattern, not just to make the diagnosis of partial adherence or nonadherence, but also to help resolve ambiguities surrounding drug action (or lack thereof) and side effects, the latter being a key determinant of adherence.

Assessing adherence in daily clinical practice can be challenging. While several different methods exist, ranging from the relatively simple (eg, self-reported adherence derived from patient questionnaires, or assessment of prescription refills or pill counts) to more sophisticated techniques (eg, directly observed therapy, urinary metabolite assays, and electronic monitoring devices that record the frequency and time of the opening of a pill bottle), these measures all have limitations [3]. Self-report in particular is subject to recall bias and social desirability, while indirect methods such as pill counts or analysis of prescription refill data are not synonymous with actual medication taking. Urinary drug metabolite assays have shown some promise in hypertension management but can be confounded by “white coat adherence” and do not necessarily reflect longitudinal medication taking [4,5]. This last limitation similarly applies in directly observed therapy, which is also limited by staff time costs and the potential for tablet concealment [3].

Strategies for reliably measuring and promoting medication adherence in daily clinical practice are thus urgently required. Concomitantly, there is widespread evidence for the increasing use of mobile health technologies in CVD risk reduction and patient self-management [6].

Here we describe our experience of using an innovative telehealth system consisting of an ingestible pill-sensor combination to record an accurate dosing history in patients at elevated risk of CVD attending a CVD prevention and rehabilitation program at our institution. We sought to compare this objective, real-time measurement of adherence with that collected by self-report while also determining the feasibility and acceptability of this technology in everyday clinical practice.

We invited 166 consecutive patients who met study eligibility criteria to participate (Figure 3), of whom 38 (22.9%) agreed and provided written, informed consent. Of those, 10 patients withdrew prior to starting to use the system, leaving a remaining 28 participants with use experience. A total of 7 participants ended their use period prematurely for reasons including the following: “Went away on holiday,” “preferred app on phone.” “didn’t want to use the device,” “device wasn’t for me,” and “off-boarded due to adverse event.”

Thus, 21 patients with a minimum of 7 days of use experience were the focus of our analysis. Table 1 outlines their baseline characteristics. Median patch wear sensor time for the group, expressed in days of collected data as a percentage of the total number of days where data were expected, was 91%, (range 49%-100%). Patients had a mean age of 62 (SD 12) years, and the majority were male (n=15, 71%) with a mean body mass index of 30.7 (SD 5.0) kg/m². A total of 14 (67%) were primary prevention patients (increased CVD risk, type 2 diabetes mellitus), and the remainder were enrolled for secondary prevention (n=7, 33%). Most patients were taking 2 or more CVD drugs at their initial assessment (Table 1); 3 patients were taking a single CVD drug, and 2 patients enrolled with a view to commencing statin therapy during the program but did not start and instead remained on at least one non-CVD drug throughout.

To the best of our knowledge, this is the first inpatient study using a system featuring an ingestible, biodegradable sensor for the objective assessment of adherence in patients attending a cardiac prevention and rehabilitation program. Self-reported adherence was high at the baseline assessment, with 90% reporting that they never missed medication. However, adherence measured by sensing of events demonstrated that only 57% achieved good adherence (defined as ≥80%) over the course of the program. This is consistent with the literature, where it is well recognized that measurement of adherence by self-report leads to overreporting, in turn due to a combination of factors, including recall bias and social desirability.

What is intriguing is that, despite the fact that participants knew their ingestion event record would be scrutinized, only 1 of the 19 who reported good adherence at baseline changed their reporting of adherence from “good” to “bad” at the end-of-program assessment, while those (n=3) with the worst adherence (<40%) continued to report good adherence at both time points. These data, therefore, underscore not only the lack of reliability of self-report as a measure of adherence, but also the psychological complexities of medication-taking behavior.

The only patient who admitted to poor adherence at the initial assessment reported never missing medication at the end-of-program assessment and achieved a sensed adherence of 74%. Although somewhat subjective, this may perhaps represent a positive change in behavior. Indeed, the majority of participants also felt that the system helped them to avoid missing doses. While some data were missing at the end-of-program assessment due to loss to follow-up, the rate of follow-up attendance in this study compares favorably with the expected typical follow-up attendance rate of approximately 60% for such a program.

The device was well tolerated, with only 1 adverse event (an episode of patch-related contact dermatitis), which is a well-recognized side effect. Despite this being an older population, the majority found the system easy to use. Recent studies in patients with hypertension or diabetes of similar average age also found high levels of acceptability and low levels of adverse events [10,11].

This high level of acceptance likely reflects the permeation of mobile phone or tablet technology into society as a whole. The evidence for mobile health interventions (mHealth) overall is steadily increasing, and one of the most prolific areas of development is CVD risk reduction [6]. Simple interventions such as text messaging have proven effective in improving clinical outcomes in a secondary prevention population, and the same research group is now studying the effect of text messages on medication adherence [12,13].

The main potential of this technology, therefore, relates to not only measurement of adherence reliably but also a strategy to increase adherence. Access to real-time adherence data and progress logs could also be extended to a patient’s family or medical team, allowing them to play a much more active role in the patient’s medication management and potentially to overcome other common barriers to health care such as limited mobility or distance from care givers or the health care setting [14]. These concepts, however, would need to be tested in the context of well-designed, controlled studies including a determination of cost effectiveness before being put into widespread use.

There are also key limitations to our study that need to be addressed. First, patients choosing to participate in the study were clearly a highly selected group (approximately 20% of those invited), a majority of whom were white, and the results, therefore, may not be reflective of the general population. We expect, however, that the high degree of self-selection would have resulted in higher measured adherence, as those who were more motivated and had good medication behavior were more likely to enter the study.

Second, the duration of the period studied was relatively short (about 3 months) and may not necessarily reflect medication adherence in the longer term.

Third, to register a dosing event, the ingestible sensors had to be coingested with the patient’s own medications, and it is entirely possible that participants could have ingested the medication alone, without ingesting the sensor (leading to underestimation of ingested events), or the sensor alone, without the accompanying medication (leading to an overestimation of adherence). Sensors that are coformulated or overencapsulated with prescribed medicines would overcome this issue, but this would also add to the complexity and cost of the intervention.

Fourth, we did not have a control group for comparison and therefore any benefits seen cannot be separated from those resulting from participation in cardiac prevention and rehabilitation alone.

In conclusion, this integrated telehealth system using an ingestible pill sensor demonstrated lower levels of adherence to CVD medications than that indicated by self-report. The technology was both safe and acceptable to patients. Larger studies are needed to determine the system’s potential for measuring and promoting adherence on a wider scale.