The Pondering Professor – Reflections on digital data and trust from PI Ipek Ensari
As a researcher whose investigative focus revolves around conditions like endometriosis, adenomyosis, fibroids, and other similar chronic pelvic pain related disorders (CPPDs), I kept hearing the same phrase for years: “These conditions are just mysterious.” In practice, what people usually mean is that we don’t understand these conditions well enough to diagnose them early or treat them reliably. But when you listen to patients with these conditions, “mysterious” is not how they describe their lives. They describe recognizable patterns that I see among our participants regularly: the day every month they can’t get out of bed, the week their sleep collapses, the morning they wake with racing heart and bone‑deep fatigue. What we call “elusive” disorders are often not rare at all. They are data‑poor; we lack structured, longitudinal, high‑resolution data that our systems know how to use. So I believe the women’s health problem that we allude to is a measurement problem.
What daily digital data makes visible
My lab works at the intersection of women’s health, digital technologies, and data science. We work with three main types of information:
-
Mobile health (mHealth) apps where people record symptoms, pain flares, mood, menstrual cycles, and daily functioning in real time.
-
Wearables that track steps, heart rate, sleep, and other physiological signals across everyday life.
-
Clinical narratives in the electronic health record, where years of experiences are compressed into a few lines of text.
If you look only at a traditional medical chart, you see what the system is designed to capture: diagnoses, procedures, imaging, labs, occasional pain scores. It’s a sparse series of dots on a timeline. When you add daily digital traces, the curve connecting those frequently sampled data points starts to come into view. And these curves (“functional data“) are what we focus on that allow us to see patterns we would not see otherwise:
-
Within‑person cycles of pain, fatigue, and mood across the menstrual cycle that are invisible in annual snapshots.
-
Distinct behavioral “phenotypes”: people who keep moving despite significant pain, people who sharply restrict activity, and people whose patterns change with work, caregiving, or stress.
-
Early warning signatures of symptom escalation, like changes in sleep, movement, or stress that often precede a “bad week.”
When our data begin to match the temporal resolution of people’s lives, what used to look like “unpredictable pain” starts to look like a pattern we can understand and sometimes anticipate. The conversation can turn from “nothing is wrong; your tests are normal” to “there is a pattern here; let’s talk about what might be driving it and how to respond.”
AI as microscope, not oracle
Digital traces on their own are messy: irregular, noisy, and profoundly individual. This is where artificial intelligence when used judiciously can help us make sense of them. In our work, we use:
-
Unsupervised learning to discover symptom and behavior patterns without assuming in advance what “types” of patients exist.
-
Reinforcement learning to personalize exercise recommendations over time for someone with chronic pelvic pain, learning from their feedback and context.
-
Large language models to synthesize unstructured clinical notes and identify sub-types of endometriosis or patients who might have conditions like endometriosis years before a formal diagnosis.
I think of these tools as a microscope, not an oracle. A microscope lets you see subtle structures that are invisible to the naked eye. It does not tell you what to value or how to care. In the same way, AI can surface patterns in years of symptom tracking and clinical notes; but it should not overrule clinical judgment or lived experience. There are also real risks if we are not careful. Models trained on historically incomplete data will reproduce those gaps. Over‑confident outputs can harden genuine uncertainty into premature “answers,” especially in areas where women’s pain has already been minimized. Algorithms optimized purely for prediction can miss what matters most to patients: functioning, identity, relationships, dignity. The question is whether we will use AI and digital technologies to deepen listening, or old patterns.
From data to justice
Underneath all of this is a question of epistemic justice: Who is recognized as a knower; whose experiences are treated as credible; whose words and signals count as evidence; who gets to shape the questions we ask. In women’s reproductive health, epistemic injustice has been the norm. Pain and bleeding that disrupt work, school, and family life are framed as “normal.” When tests come back “reassuring,” patients are left feeling gaslit: their lived reality is in conflict with the medical record. Digital health can either reinforce these dynamics or start to correct them. When people use mHealth tools and wearables, they become co‑producers of data about their own bodies. But if those data disappear into opaque silos or are fed into models that no one can explain, we risk a new kind of dispossession: intimate traces of people’s lives used, with little meaningful care in return.
Closing the data gap in women’s health therefore requires more than better algorithms. It requires a different relationship to data, grounded in three commitments:
- Measure what matters to patients, not only what is convenient for billing or devices.
- Design with the most overlooked at the center, not as an afterthought.
- Treat data as building blocks of relationship, not commodity. This goes back to my previous post on clear consent, shared interpretation, and real options to say no.
A future for women’s health
I imagine a near‑future visit where a patient and clinician sit together to review several months of symptoms, activity, sleep, and mood, summarized in intuitive ways, not just in dense graphs. This means that regardless of how much AI integration there might be introduced to the process, making sense of those patterns in the context of a real life requires a conversation between the two humans.
The goal is not to turn everyone into a quantified‑self enthusiast, or to automate away expertise. It is to build a system in which women’s experiences are legible, believed, and actionable. We already have many of the tools we need. The open question, one that belongs at the center of any conversation about the future of health, is whether we will use those tools to listen better to women and to anyone whose suffering has been poorly captured by our current systems. That is the future my lab is working toward, one dataset and one conversation at a time.
Next Steps for CPP Tracker
All of these make me very excited for what is next for CPP Tracker, our study that aims to develop validated digital patient-reported outcome measures (dPROMs) specifically designed for those with CPPDs. Ever since we began enrollment back in 2023, I have been getting questions on when it will be scaled and available for everyone to download. We will soon make this research-based App available for everyone interested to download and start using to track their symptoms, health, and have summaries that they can use with their clinicians. It will continue to evolve and improve based on our participants’ feedback and needs, so the collaboration we have with our participants will continue. This is what makes this tool relevant and useful. Because digital health research depends on participant trust, and trust depends on clear communication.
Closing: The Pondering Professor Corner
This post is the fourth in our series The Pondering Professor Corner about how we think about your data, your trust, and your participation in digital health research. You can read other posts in this series on informed consent on research data use, and on patient trust with digital data
—
About the author: Ipek Ensari, PhD, is an Assistant Professor at the Icahn School of Medicine at Mount Sinai Department of AI and Human Health and principal investigator of the CPP Tracker study on chronic pelvic pain. She is a digital health researcher focused on gynecological pain disorders like endometriosis, adenomyosis, and fibroids, wearable and app-based data, and AI-driven methods, with a particular interest in health data privacy, informed consent, and participant-centered study design.