Last spring a colleague sent me her study protocol for a quick sanity check. Twelve pages. Beautifully written. Two full sessions per participant, each scheduled for ninety minutes, plus an optional third visit. I read it and asked one question: "Have you ever tried to sit upright for ninety minutes on a day when your joints feel like they're full of crushed glass?" She went very quiet.
That's the whole problem in a sentence. Accessible research design for chronic illness participants isn't a checkbox at the end of the protocol. It's the difference between a study that produces data and one that produces twelve withdrawals and a note in your ethics report.
The number of people living with a chronic condition worldwide is enormous—the World Health Organization puts disability at roughly 16% of the global population, and chronic illness runs far wider than that. Ignoring this population isn't a niche omission. It's throwing away most of your potential sample.
Key Takeaways
- Design for the worst day, not the best day—because the worst day is when most participants will actually be trying to show up.
- Energy is the real recruitment budget. Every minute you spend is a minute taken from something else in their life.
- Asynchronous and remote options are not lesser data. They are the only data you'll get from a large slice of this population.
- Consent forms must be usable by someone who's been awake since 3 a.m. in pain.
- Compensate honestly. Nobody should go into energy debt to serve your study.
Designing for the bad day, not the good one
Here's a mistake I made and still wince about. Early on, I built a session schedule around what I called "typical capacity"—two hours of engagement, one hour break, one hour follow-up. I tested it on participants who happened to be doing well that month. It worked perfectly. Then autumn hit, two people flared simultaneously, and I had a data gap I couldn't close.
The problem wasn't the participants. It was that I had designed for the median and assumed the median was stable. Chronic illness is variable by definition. That variability isn't noise you control for statistically—it's a structural feature you plan around.
Treat participant energy as a finite budget
I now do this in every protocol: before I write a single session length, I write down what else the participant is likely doing that day. Work. Caregiving. Medical appointments. Basic human maintenance. Then I ask myself whether my study sits comfortably inside a slice of what's left, or whether it needs the whole slice.
In practice, this changed three things for me:
- I cut my main interview from 90 minutes to a 45-minute hard ceiling, and I offer a second 30-minute follow-up instead.
- Sessions are scheduled with a 48-hour reschedule window with no questions asked. I stopped tracking reschedules as a data quality issue. They're just calendar noise.
- I build in a mid-session pause of five minutes that participants can extend, shorten, or ignore. I don't ask them to justify it.
Did the shorter sessions hurt my data? Honestly, no. The 45-minute version produced richer transcripts because people weren't fighting exhaustion by minute 70.
Recruitment and consent that actually reach people
Consider the gap between how most studies recruit and how a person with a fluctuating condition hears about a study at all. If the only route in is a phone call during business hours and a screening questionnaire that assumes you're upright and available, you've filtered your sample before you started.
The screening form is your first accessibility test
I'll be blunt: most screening forms are hostile. They demand fine motor precision on mobile, they time out, and they offer no progress indicator. Fix the obvious things:
- Allow asynchronous completion. Let people start, save, and return. A flare can end a session mid-form, and the form shouldn't punish that.
- Offer text-only options. Not everyone wants a video call—and some cannot do one. Written interview responses, a typed chat, or a voice-note option cover far more ground than a single modality.
- Make the deadline generous. A two-week recruitment window is a filter that selects for people already in a good month.
Consent forms that are readable on a hard day
Consent is where accessibility failures hide in plain sight. Font size, contrast, screen-reader compatibility, and—crucially—length. I've seen consent documents run to nine pages because legal wanted every contingency enumerated. That's not informed consent; that's a barrier dressed up as rigor.
What worked for me: a two-page summary in plain language, with the full document available but not mandatory to read line by line, and a recorded verbal walkthrough option offered to anyone who wants it. Every participant who took the walkthrough told me they understood the study better afterwards. That's not a coincidence.
Remote, asynchronous, and hybrid options compared
Different modalities suit different conditions—and different days within the same condition. Here's what I've found across roughly four studies now:
| Modality | Best for | Main drawback | My verdict |
|---|---|---|---|
| In-person, single short session | Participants who can travel and want the human contact | Travel cost and fatigue; scheduling fragility | Valuable but never the only option |
| Synchronous video call | Rich qualitative interviews | Requires sustained attention and a good connection | Great default, offer alternatives |
| Asynchronous written responses | Brain fog days, night owls, people who need time to think | No interviewer probing in real time | Undervalued—my best data came from here |
| Voice notes or audio submissions | Low energy but fluent speech | Transcription overhead | Worth the extra work |
| Hybrid (mixed across sessions) | Almost everyone | Harder to keep analysis consistent | Where I've landed as standard |
The instinct is to call the in-person session the "gold standard" and the async option a compromise. I don't buy it. If a participant gives you three thoughtful paragraphs at midnight when they can't sleep, and the in-person interview got ninety minutes of them trying not to shift in their chair, which one is richer? In my experience the async material often is.
Compensation, timing, and follow-through
Which brings up the money question. People get awkward about this. Don't be awkward about it.
Paying participants is not a courtesy, and it's not a bribe. For someone managing a chronic condition, participation costs energy they may not get back. Some studies I've seen offer a token gift card at the end of a multi-visit study. That's a mismatch between the ask and the payment.
How much, and when
Two principles I hold firmly:
- Pay per session, not at the end. If someone withdraws after visit one, they've still spent energy. Pay them for what they gave you.
- Compensate for time and disruption, not just for "being in the study." Travel time counts. Waiting time counts. Recovery time—yes, that counts too, and I've had reviewers push back on it, and I've kept it in anyway.
On withdrawal: state clearly, in the consent form and verbally, that withdrawal carries no penalty and that participants keep their compensation. I've had one participant withdraw after a flare and then re-enroll six weeks later because withdrawing hadn't cost them anything. That only happened because the protocol didn't punish them.
What I got wrong (and what you can skip)
Three mistakes worth naming, because I see them in other protocols regularly.
One: I built a "flexible" schedule that wasn't flexible. It offered two time slots per week. That's not flexibility, that's a menu. Real flexibility means the participant sets the day where possible, and the researcher adapts.
Two: I relied on video for everything. I lost potential participants who had no camera setup, who shared a living space, or who simply didn't want to be seen on a hard day. Offering a fallback after participants complained was too late—they'd already left.
Three: I didn't ask. The best change I made was adding one line to every recruitment notice: "Tell us what would make participation possible for you, and we'll try to arrange it." Roughly a dozen people over two studies told me exactly what they needed—shorter forms, a phone call instead of a form, a follow-up reminder before deadlines. I couldn't accommodate all of it. I accommodated most of it.
The question I get most
"Won't all this flexibility ruin my analysis?"
No. Different participation modalities produce different data shapes, and you can document that. What actually ruins analysis is having a sample that's skewed toward people in remission, or losing a fifth of your participants to scheduling friction. A protocol that only works for participants having a good month is a protocol that generates a biased sample—and you'll spend weeks in analysis wondering why your findings feel thin.
Design for the bad day. You'll get the good ones for free.
And when you next write "session length: 90 minutes," ask yourself who that number is really for.
