Forget everything you think you know about fitness trackers. That sleek wristband counting your steps and monitoring your sleep? It was supposed to democratize health research, bringing low-cost data collection to places that desperately need it. Instead, new research from Uganda reveals a rather inconvenient truth: when we deploy these gadgets in low-resource settings where they could do the most good, roughly a third of the data simply vanishes into the ether. And not randomly, either - which makes everything considerably more complicated.

The Promise vs. The Reality

Here's the pitch we've all heard: wearable fitness trackers can revolutionize public health research by collecting massive amounts of individual-level data cheaply. No more expensive clinical visits. No more relying on people to accurately remember how many hours they slept last Tuesday (spoiler: they can't). Just strap a Garmin on someone's wrist and let the data flow.

The problem? This utopian vision has largely played out in high-income countries where participants have reliable electricity, stable housing, and the luxury of not worrying about whether their fancy watch will get stolen. When researchers from The Onward Project on Well-being and Adversity tried to bring this approach to informal settlements in Kampala, Uganda, they discovered that reality had other plans.

When 30% of Your Data Goes AWOL

The study equipped 300 women living in slum communities with Garmin smartwatches to track their sleep over five days. Simple enough, right? Except approximately 30% of the nighttime data went missing. For a study aiming to understand sleep patterns and well-being, that's not a minor hiccup - it's a gaping hole in the dataset.

But here's where it gets interesting for the statistics nerds among us (and I count myself proudly among that peculiar tribe). Not all missing data is created equal. In the world of biostatistics, we obsess over something called the "missingness mechanism." It sounds like a rejected title for a spy thriller, but it's actually fundamental to whether your research conclusions hold water.

A Brief Tour of Missing Data Mechanisms

There are three flavors of missing data:

Missing Completely at Random (MCAR): The data fairy randomly plucked out values with no rhyme or reason. If you're lucky enough to have MCAR data, your analyses are relatively straightforward. You can essentially ignore the gaps.

Missing at Random (MAR): The missingness depends on other observed variables but not on the missing values themselves. Trickier, but manageable with the right statistical gymnastics.

Missing Not at Random (MNAR): The data is missing precisely because of what that data would have shown. This is the nightmare scenario. Think of people not reporting their weight because they've gained weight. The missingness itself contains information, and ignoring it biases everything.

Playing Detective with Disappearing Data

The researchers employed an impressive arsenal of statistical methods to figure out what was happening. They used pattern analysis, the Little test, random forest classification, and logistic regression - essentially throwing every reasonable approach at the problem to see what stuck.

Three main patterns emerged from the chaos. Some participants had complete data (the gold standard we dream about). Some had only the fifth night missing (curious, that). And some had no data at all across the entire study period (which raises obvious questions about device compliance, device failure, or both).

The verdict? The data was definitively not missing completely at random. The Little test came back with a p-value less than 0.001, which in scientific terms means "we're very, very sure this isn't random." Both the random forest and logistic regression models agreed - something systematic was causing these gaps.

Why This Actually Matters

You might be thinking: "Okay, so some data is missing. Just work with what you have." And this is where I get to be the bearer of uncomfortable news.

When data goes missing non-randomly, simply analyzing what remains can lead to spectacularly wrong conclusions. Imagine studying whether sleep quality affects depression in this population. If the women with the worst sleep are also the ones whose data is missing - perhaps because chaotic living conditions both disrupt sleep AND make it harder to keep a device charged and on your wrist - then your "complete" dataset systematically excludes exactly the people you most need to understand.

The researchers' careful characterization of missingness patterns isn't just academic navel-gazing. It's essential groundwork for developing statistical corrections that can salvage meaningful insights from imperfect data. And in low-resource settings, the data will almost always be imperfect.

The Bigger Picture

This study highlights a tension at the heart of global health research. We have these remarkable tools that could theoretically bring precision health monitoring to underserved populations. But the same factors that make these populations underserved - unreliable infrastructure, economic instability, competing survival priorities - also make the data collection fundamentally harder.

It's a bit like developing a fuel-efficient car and then discovering it only works well on smooth highways, not the unpaved roads where fuel efficiency would matter most.

The path forward isn't to abandon wearable research in these settings. Rather, it's to go in with eyes open about the challenges, design studies that anticipate high missingness rates, use appropriate statistical methods that account for non-random gaps, and perhaps develop devices and protocols specifically suited to challenging environments.

What Comes Next

This research from Kampala is part of a growing body of work trying to make digital health tools actually useful in the contexts where they're most needed. Future studies will need to dig deeper into why the missingness occurs - is it device removal, battery failure, sync problems, or something else entirely? - and develop interventions to reduce it.

Until then, the next time someone tells you that fitness trackers will revolutionize health research in developing countries, you can nod politely while mentally noting that roughly a third of that revolution might be missing, and not randomly.

This blog post discusses research findings and should not be taken as medical advice. If you have concerns about sleep disorders or health monitoring, please consult a healthcare provider. Research discussed here represents ongoing scientific investigation and clinical validation is still in progress.

All images used in this post are decorative illustrations only and do not represent or reflect the accuracy, reality, or correctness of the referenced research.

Primary Source: Determining a Likely Mechanism of Missingness in Repeated Measures Sleep Data From Wearable Fitness Trackers: Longitudinal Analysis. JMIR. 2025. PMID: 41915671