Vardaan Shah

Measuring Medication Adherence

Introduction

The measurement of adherence to a medication regimen is a surprisingly challenging task. Many different solutions exist, each with their own trade-offs. There are particularities with ADHD and ADHD medication that need to be considered to produce an appropriate measurement scheme, and this project has several salient features that may also impact that choice.

Measures

Lam and Fresco's overview of medication adherence measures is my main source for this section. (Lam & Fresco, 2015)

Medication adherence measurements can be divided into two major categories, namely, subjective measurements and objective measurements. Subjective measures involve a provider or a patient evaluating their patient's or their own behavior without external verification. Objective measures use some kind of measurement of either the medication or the patient to sidestep the issues of subjective measurements, but these methods come with issues of their own.

Another way to divide up measurement methods is into direct and indirect methods. Direct methods involve either an investigator directly observing a patient take a medication or the measurement of the drug or one of its metabolites in the patient's bodily fluids. Indirect measures measure some proxy for ingestion of the medication. Direct measurements will not be useful in the context of this project---they are too expensive and are not appropriate for long-term monitoring of patients.

Objective Measurements

Lam and Fresco divide the category of objective measurements into measurements that involve:

  1. secondary database analysis,
  2. electronic medication packaging (EMP) devices, and
  3. counting pills.

Secondary Database Analysis

Secondary database analysis involves deriving patterns through analyzing electronic prescription service or pharmacy insurance databases. The primary assumption underlying the accuracy of this method is that prescription-refilling patterns correspond to the patient's medication-taking behaviors.

EMP Devices

EMP devices are devices incorporated into the packaging of prescriptions that monitor and sometimes provide feedback on adherence performance. MedPet is an EMP device. As a result, as long as the data is accessible to the team doing the hypothetical study on MedPet's efficacy, we will get the data from it for "free," although that data has limitations.

Pill Count

Pill count, as the name suggests, counts the number of pills left in the bottle after a specified time has elapsed, and compares it to the number of pills that should have been left in the bottle.

​Lam and Fresco's review is from 2015, and in the 11 years since, technology has changed significantly. As such, several adherence-measurement technologies that Lam and Fresco do not address have emerged in recent years. The narrative review published by Mason et al includes evaluations of many of these technologies, including blister pack technologies, ingestible sensors, video-based monitoring technologies, and motion sensor technologies. (Mason et al., 2022)

Blister Pack Technologies

This category of technology uses stickers with a pattern of wire that attach to blister packs. When the pack is broken to access the medication inside, a circuit is broken in the wire and a microchip records the date and time of that breakage event.

Ingestible Sensors

Ingestible sensors involve modifications to the pill itself and the wearing of an external monitor. The pill is equipped with a sensor that, upon contact with gastric acid, transmits a signal to the external monitor. The external monitor records the date and time of the ingestion.

Video-Based Monitoring

This technology involves a patient either sending a video to the investigator in which they take the medication, or going on a video call with the investigator live where they take the medication.

Motion Sensor Technology

Motion sensor technologies generally equip a patient with a bracelet-mounted accelerometers. The system attempts to identify the movements that a patient makes when administering medication, and when such a movement is detected, logs an administration event.

Subjective Measurements

Lam and Fresco divide the subjective measurement methods into three subcategories: patient-kept diaries, patient interviews, and questionnaires and scales.

Patient-Kept Diaries

The patient is to keep a diary in which they document when they do and do not take their medication. At the end of the diary period, the diary is collected and analyzed.

Patient Interviews

Interviews can come in a broad range of formats, but the general outline involves an investigator asking the patient to estimate their adherence behavior retrospectively.

Questionnaires and Scales

Many questionnaires and scales exist for measuring medication adherence. A notable entry is the Medication Adherence Questionnaire (MAQ), a questionnaire upon which several others are built and one that has been tested on the widest range of diseases. Apparently, it is the most widely used scale in research. However, a derivative scale called the Eight-Item Morisky Medication Adherence Scale (MMAS-8) had become more popular by 2015 due to its better psychometric properties.

Evaluation

The breadth of options available to measure and the lack of a "gold standard" measure lead to the recommendation to use multiple methods simultaneously in order to capture different aspects of the data. That being said, each of the above methods has trade-offs, and the ideal selection would be a set of measures that complement each other's problems.

ADHD has a symptom profile that makes the measurement of any related outcomes more challenging than they otherwise would be. These symptoms include failing to pay attention to detail, difficulty organizing tasks and activities, excessive talking or fidgeting, difficulty relaxing, overworking, forgetfulness, and distractibility. (Katzman et al., 2017) They have a measurable impact on the assumptions underlying several measurement methods, and should inform how we judge the success of our intervention.

In particular, EMR records should be taken with a grain of salt---a 2020 study examining them found that only 42% of adult patients prescribed stimulants renewed their initial prescription. The authors postulate that the symptoms of ADHD themselves could play an important role in this observation. (Biederman et al., 2020)

As a result, secondary database analysis should not be taken as a ground source of truth for ADHD patients' desire to take their medication. These records can, however, corroborate the extent to which our intervention over-estimates patient adherence to the medication regimen.

Another consequence of this symptom profile is that the methods that rely on the patient to independently keep a record of their adherence or to check in every day for something are prone to error. These include patient diaries and asynchronous video-based monitoring.

Because ADHD is associated with forgetfulness, methods that rely on retrospective estimation are also likely more prone to error than the same method on a general population. Questionnaires and scales either rely on frequent check-ins or retrospective estimation, and so should be ruled out for empirical information about adherence to treatment. That being said, retrospective questionnaires and scales can still be valuable sources of qualitative data, and thus should still be considered, just not as sources of empirical truth.

In addition, ADHD medication is always at least once-daily, i.e. a once-weekly ADHD pill does not exist. As a result, any measurement method that requires the investigator to be present, whether physically or virtually, for each dose simply cannot scale enough to produce statistically meaningful results, and thus must be discarded.

There is a wide variety of ADHD medication available. At least 20 different oral medications are approved by the FDA, and each of those comes in a wide variety of dosages. ("ADHD Medications Approved by the US FDA (Infographic)," n.d.) Any intervention that requires a change in the packaging of the medication or its formulation would either incur enormous costs in order to accommodate several types of medication or exclude an enormous portion of the ADHD medication-taking population and severely bias the sample towards those who take the supported medications. This rules out (as cool and high-tech as they are) the blister pack technologies and the ingestible sensors.

Motion capture technology, although undeniably fascinating, would be an additional source of uncertainty in a project that will already be testing a hardware device. It would be more appealing if the accuracy was higher, but the correct ingestion rate Wang et al reported was 84.17%. (Wang et al., 2014) This rate is not nearly high enough to use as a reference to compare the MedPet's accuracy against. It would also come with a significant added cost and technical maintenance requirement.

The MedPet project has some advantages: the MedPet itself gives us a source of EMP-based measurement. Thus, another EMP solution would be redundant.

Pill counting seems like a relatively inexpensive way to sanity-check the MedPet measurements---however, traditional pill counting does not have a particularly high sample rate. I think it would be helpful to, along with the MedPet device, give each participant in the hypothetical study a digital scale. They would take a video of themselves measuring the mass of the remaining pills in the bottle on some regular interval (a week feels right?). This way, we can verify whether the patient correctly tared the scale before placing their bottle on the scale.

Prescription bottles are pretty standardized, and we would only need to obtain baseline measurements for as many bottles as there are unique pharmacies in the study. Then, when the patients get a new 30-day prescription, we ask them to send the video to establish a baseline for the full bottle. From there, we have all the information we need to estimate the number of pills left in the bottle. That procedure relies much less on manual counting and thus is less prone to error, and can be done at an arbitrarily high sample rate---4 measurement samples per month seems more than feasible.

In addition to this, if we can get access to EMR or other secondary database records for prescription refills, we can further contextualize the performance of MedPet month-to-month and control for the factors that make it difficult for ADHD patients to refill their prescriptions.

Conclusion

After reviewing the landscape of medication-adherence-measuring techniques, I am confident that statistically meaningful data can be derived out of the stack of measurement techniques I have constructed. The MedPet will be used as the primary source of data. Mass-based estimation of pill-counting will be used as a way to place a bound on the MedPet data and to validate it. If possible, EMR/insurance data will be used to place refills in time to better evaluate MedPet's performance in context of when the user actually has access to their medication.

References

ADHD Medications Approved by the US FDA (infographic). (n.d.). CHADD. Retrieved February 27, 2026, from https://chadd.org/about-adhd/adhd-medications-approved-by-the-us-fda/

Biederman, J., Fried, R., DiSalvo, M., Woodworth, K. Y., Biederman, I., Driscoll, H., Noyes, E., Faraone, S. V., & Perlis, R. H. (2020). Further evidence of low adherence to stimulant treatment in adult ADHD: An electronic medical record study examining timely renewal of a stimulant prescription. Psychopharmacology, 237(9), 2835--2843. https://doi.org/10.1007/s00213-020-05576-y

Katzman, M. A., Bilkey, T. S., Chokka, P. R., Fallu, A., & Klassen, L. J. (2017). Adult ADHD and comorbid disorders: Clinical implications of a dimensional approach. BMC Psychiatry, 17(1), 302. https://doi.org/10.1186/s12888-017-1463-3

Lam, W. Y., & Fresco, P. (2015). Medication Adherence Measures: An Overview. BioMed Research International, 2015(1), 217047. https://doi.org/10.1155/2015/217047

Mason, M., Cho, Y., Rayo, J., Gong, Y., Harris, M., & Jiang, Y. (2022). Technologies for Medication Adherence Monitoring and Technology Assessment Criteria: Narrative Review. JMIR mHealth and uHealth, 10(3), e35157. https://doi.org/10.2196/35157

Wang, R., Sitová, Z., Jia, X., He, X., Abramson, T., Gasti, P., Balagani, K. S., & Farajidavar, A. (2014). Automatic identification of solid-phase medication intake using wireless wearable accelerometers. 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, 4168--4171. https://doi.org/10.1109/EMBC.2014.6944542