⚡ electrical engineering

The Pill Left Behind: A Passive Sensing Approach to Medication Adherence Monitoring

This paper proposes and validates a low-cost, load cell-based passive sensing system that achieves 96% accuracy in detecting medication removal events, offering a statistically significant, deployable solution to address the institutional accountability gap driving the global medication non-adherence and waste crisis.

Original authors: Shishir Dhanapunen, Aayush Nerkar, Alijawad Peera

Published 2026-07-22

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Shishir Dhanapunen, Aayush Nerkar, Alijawad Peera

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the world of healthcare as a giant, bustling kitchen where millions of meals are prepared every day. Sometimes, the chefs (doctors) cook up a perfect recipe, and the diners (patients) eat every bite, feeling stronger and healthier. But often, the food sits untouched on the plate, eventually getting thrown in the trash. This isn't just about wasted food; it's about wasted medicine. When people don't take their pills as prescribed, it's like throwing away a golden ticket that could save lives, while the rest of the kitchen burns money trying to fix the problems that arise from the hunger.

To understand how scientists are trying to fix this, we need to look at three simple ideas. First, there's medication adherence, which is just a fancy way of saying "did you actually eat the food?" Second, there's passive sensing, which is like having a smart scale under the plate that knows you took a bite without you having to press a button or tell anyone. Finally, there's institutional accountability, which means the people who pay for the food (like insurance companies) should know if the food is being wasted so they can help fix the problem, rather than just watching the trash pile up. This paper dives into the "smart scale" idea, asking if a simple, low-cost device can tell us when a pill is missing from a bottle, and if that information can help stop the massive waste of money and medicine.

The Pill Left Behind: A Detective in a Bottle

Meet the "Pill Left Behind." It's not a superhero, but it is a tiny, low-cost detective designed to solve a massive mystery: Where did the pills go?

Every year, American households throw away more than 1 billion unused drug units. That's a billion pills sitting in medicine cabinets, gathering dust, or ending up in the trash and our waterways. This waste costs the healthcare system over $500 billion a year and is linked to about 125,000 preventable deaths. It's a tragedy of epic proportions, but the usual solutions—like phone apps that beep to remind you to take your pill—often fail because they rely on you to remember to tap the screen. They also keep the data to themselves, meaning the insurance companies and pharmacies never know the pills are actually sitting in your cabinet, untouched.

The authors of this paper, a team of high school researchers, wanted to build a system that doesn't nag you but simply watches the bottle. They call it a passive sensing approach.

The Magic Scale

Imagine your pill bottle is sitting on a super-sensitive bathroom scale. This isn't a normal scale, though; it's a load cell, a device that can feel the weight of a single pill. The researchers built a system using three main parts:

The Load Cell: A tiny sensor that feels the weight of the bottle.
The Amplifier (HX711): A helper that takes the tiny signal from the scale and makes it loud enough for a computer to hear.
The Brain (Seeed XIAO nRF52840): A small microcontroller that acts as the detective, checking the weight every few days and sending the news wirelessly.

The system works like this: You fill the bottle with pills. The device weighs it and remembers the starting weight. A few days later, it wakes up, weighs the bottle again, and does some quick math. If the bottle is lighter, the detective concludes, "Hey, a pill is missing!" It then sends this information wirelessly to a computer where doctors, pharmacists, or insurance companies can see it.

The Big Test: Did It Work?

To see if their "smart scale" was any good, the team ran a test. They filled a bottle with 90 cholesterol pills (Atorvastatin). They then played a game of chance: they flipped a coin. If it landed on heads, they took one pill out. If it landed on tails, they left it alone. They did this 100 times, making sure to put the pill back after each turn so the bottle was ready for the next round.

The results were impressive. Out of 100 trials, the system correctly guessed whether a pill was removed or not 96 times. It only made a mistake 4 times.

But the researchers didn't just stop at "it looked good." They ran serious math tests (statistical tests) to be absolutely sure this wasn't just luck. They asked: "Is the success rate actually better than 90%?" The answer was a resounding yes. The math showed that the chance of getting these results if the system was actually bad was incredibly low (a probability of 0.024). In plain English, the system is statistically proven to be reliable enough to detect pill removal.

Why This Changes the Game

Here is the most exciting part. Most "smart" pill bottles today just send a reminder to you when you miss a dose. They are like a nagging parent. This new system is different. It sends the data to the institutions—the insurance companies and pharmacies that are losing billions of dollars because the pills aren't being taken.

Think of it like a usage-based car insurance plan. If you drive safely, you get a discount. If you drive recklessly, you pay more. The researchers suggest that if insurance companies could see this data, they could create programs where patients who take their meds get rewards (like lower costs), and the system can finally start fixing the root cause of the waste. It shifts the focus from blaming the patient to fixing the system.

The Catch (And the Future)

The paper is very honest about what it doesn't know yet. The test was done in a perfect lab setting on a flat table. In the real world, a bottle might sit on a wobbly nightstand, or a fan might blow on it, or the pills might be different sizes. The system also can't tell if you took the pill out and then put it back in, or if you took it out and didn't swallow it. It only knows the pill is gone from the bottle.

The researchers admit that while their device works great in the lab, it needs more testing in real homes with real people and different types of pills. They also note that the current version uses a wired setup for the test, but the final goal is a wireless device that can run on a battery for months.

The Bottom Line

This paper doesn't claim to have solved the entire problem of medication waste overnight. Instead, it proves that a low-cost, simple device can accurately tell when a pill is missing from a bottle. It shows that we can build a "smart scale" that works well enough to be trusted. By connecting this data to the people who pay for the medicine, the authors argue we can finally close the gap between a prescription being written and a pill actually being taken. It's a small step in the lab, but a giant leap toward a healthcare system that stops throwing away billions of dollars and, more importantly, saves lives.

Technical Summary: The Pill Left Behind: A Passive Sensing Approach to Medication Adherence Monitoring

Problem Statement
The paper identifies a critical disconnect in the healthcare system regarding medication non-adherence. While existing solutions (such as reminder apps or smart caps) focus on consumer-facing notifications, they fail to transmit adherence data back to the institutions that bear the financial and environmental costs of wasted medication: pharmacies, insurance companies, and hospitals. This lack of visibility prevents these entities from implementing targeted interventions. The authors argue that medication waste is not merely a behavioral issue but an institutional accountability gap, resulting in over $500 billion in avoidable healthcare costs annually, 125,000 preventable deaths, and significant environmental contamination from discarded pharmaceuticals. Current policies, such as the Secure and Responsible Drug Disposal Act of 2010 and the Inflation Reduction Act of 2022, address disposal or pricing but fail to bridge the gap between a prescription being dispensed and a patient actually taking the medication.

Methodology
To address this gap, the authors propose and validate a low-cost, passive sensing system designed to estimate medication adherence by detecting pill removal events.

Hardware Architecture: The system utilizes a load cell-based embedded design. The core components include a 1kg strain gauge load cell, an HX711 amplifier for signal conditioning and analog-to-digital conversion, and a Seeed XIAO nRF52840 microcontroller for processing and wireless transmission.
Circuit Design: To ensure energy efficiency, the circuit employs a P-channel MOSFET as a high-side switch. This allows the microcontroller to power the sensing electronics only during periodic measurement windows, significantly reducing battery drain compared to continuous monitoring. The load cell is connected in a differential configuration to the HX711 to reject noise and detect minute mass changes.
Software Logic: The system operates on a periodic check interval. It records the initial mass of the pill bottle and compares subsequent readings against this baseline.
- Detection Algorithm: If the current mass is less than the previous mass, the system infers a pill removal event. The change in mass ( $\Delta m$ ) is divided by the known mass of a single pill to estimate the quantity removed.
- Adherence Estimation: The estimated number of pills removed is compared against the expected number of doses based on the medication schedule. Discrepancies trigger non-adherence alerts.
- Data Transmission: Processed adherence metrics are transmitted via Bluetooth Low Energy (BLE) to external software accessible by insurers and healthcare providers.
Experimental Design: The authors conducted a controlled proof-of-concept experiment using a 30-dram bottle containing 90 Atorvastatin pills (approx. 500 mg each). The device was interfaced with an Arduino Uno for the test. A total of 100 trials were conducted where a coin flip determined whether a single pill was removed or not. After each trial, the pill was returned to the bottle to ensure independence. The system's output was compared against the actual event to determine success or failure.

Key Results

Detection Accuracy: In the 100 controlled trials, the system correctly identified pill removal and non-removal events 96 times, resulting in an observed success rate of 96% (4 failures).
Statistical Significance: The authors performed three statistical analyses to validate the results:
1. One-proportion z-test: With a null hypothesis ( $H_0$ ) of a 90% success rate, the test yielded a p-value of 0.02275, leading to the rejection of the null hypothesis.
2. One-proportion z-interval: A 95% confidence interval was calculated as (0.92159, 0.99841), lying entirely above the 90% threshold.
3. Exact Binomial Test: Due to the low number of failures (4), which technically violated the success-failure condition for normal approximation, an exact binomial test was conducted. This yielded a p-value of 0.024, providing convincing evidence that the true success rate exceeds 90%.

Significance and Claims
The paper claims that this system represents a viable, low-cost infrastructure solution to the medication waste crisis. Its primary significance lies in shifting the paradigm from consumer-centric reminders to institutional accountability. By passively capturing behavioral data and transmitting it to payers and providers, the system enables:

Data-Driven Incentives: Insurers and pharmacies can utilize adherence metrics to design incentive structures (e.g., rebates, reduced copays) rather than punitive measures.
Targeted Intervention: Institutions can identify non-adherence patterns and intervene before costly hospitalizations occur.
Environmental Impact: By addressing the root cause of waste (non-consumption) rather than just disposal, the system aims to reduce the billions of drug units entering waterways annually.

The authors maintain a modest tone regarding the system's current capabilities. They explicitly state that the device estimates pill removal as a proxy for consumption, acknowledging that it cannot verify ingestion. They note that the current study was conducted under controlled conditions and that future work is required to validate performance in real-world environments, account for pill variability, and assess long-term sensor drift. The paper concludes that while the technology is a necessary step, the ultimate solution requires a structural policy shift where adherence data informs institutional decision-making.