The Casebook

The idea: attrition, and why it is invisible while it is happening. Every instrument a study has reports on the people who responded, so the person who stopped responding drops out of the very numbers meant to catch them. And unlike most research problems this one cannot be repaired after the fact: data not collected in week four does not exist in week twelve.

The trope: the missing person.

Based on: Retention & Attrition / Missing Data

The idea: recruitment as a queue rather than a total. Enrollment is usually measured in counts — how many at each stage — and a count cannot distinguish someone who applied this morning from someone who has been waiting a month. Most people lost before baseline are not refusing; they are waiting, and nobody has told them how long.

The trope: intake, and a line of walk-ins.

Based on: Recruitment & Enrollment / CONSORT Reporting

The idea: the fragility of the front door. Contact details reach a study on paper, in handwriting, and get typed in by a person, so at any real volume a transcription error is a certainty rather than a risk. Delivery confirmations compound it: they prove a message was transmitted, never that it reached the right human being. A system can detect the absence of an answer — it can never detect a wrong recipient.

The trope: the message that goes out and never comes back.

Based on: Data Integrity

The idea: how hard it is to actually conduct a properly randomized trial, and the safeguards the field has accumulated so the benefits of randomization aren't quietly lost — concealing the allocation order, stratifying where sites genuinely differ, varying block sizes so the next assignment can't be deduced, and being able to establish afterwards who saw what. Randomization is not a step performed once at the start; it is a property that has to survive contact with two clinics, six months, and everybody's good intentions.

The trope: a deck she is sure is rigged.

Based on: Randomization / Blinded Studies

The idea: wear time, and the difference between data a device produced and data a study can use. A wearable records nothing distinguishable from perfect stillness when it is sitting on a nightstand, so validity depends on a threshold somebody chose — and a participant who takes it off is not refusing the study, they are living in it.

The trope: he was wearing a wire and he took it off.

Based on: Digital Measurement / Protocol Adherence

The idea: the distance between holding data and knowing what is in it. Most study data is never interrogated beyond the questions somebody thought to build a report for in month one, which means the finding that matters is routinely present and unasked.

The trope: the lamp, the chair, and a subject who finally talks.

Based on: Exploratory & Secondary Analysis

The idea: study design as a schedule. Every decision about when something is delivered or measured is made months before anyone can tell whether it was right, and those decisions silently determine which analyses are available at the end.

The trope: planning the job — red string across a wall of index cards.

Based on: Study Design / Measurement Schedules

The idea: protocol change, which is normal, versus protocol change nobody recorded, which is fatal. Studies are amended constantly and for good reasons; what separates a defensible result from an indefensible one is whether anyone can reconstruct what the protocol was on a given date.

The trope: somebody changed the books in month two.

Based on: Protocol Amendments / Auditability

The idea: knowing that the person you are messaging is the person who consented. Contact details are self-reported, phones change hands, and consent to participate is not the same act as confirming a channel.

The trope: making sure you're talking to who you think you are.

Based on: Informed Consent / Participant Identity Verification

The idea: the boundaries inside a study — baseline, intervention, follow-up — and the fact that they are enforced by people remembering them unless something enforces them.

The trope: everyone stops at the border and gets their papers reviewed.

Based on: Protocol Compliance / Visit Windows

The idea: delivery in a language the participant actually uses. An intervention translated but never reviewed by a speaker is not an intervention that was delivered, and the resulting non-response looks identical to disinterest.

The trope: the intervention was speaking the wrong language.

Based on: Health Equity / Cultural Adaptation

Handle with care. The one premise that can turn preachy or condescending. Hold it until the voice is fully proven, and make the miss hers to catch, never a community's deficiency.

The idea: delegated care, and the information boundaries that come with it. More than one person works a roster, and who is entitled to read whose messages is a study design question before it is a permissions question.

The trope: somebody else working the case with you.

Based on: Delegation of Duties / Research Team Roles

The idea: what leaves the study and who is entitled to see it. An export is the moment data stops being governed by the protocol, and blinding that holds inside the platform has to hold in the file somebody downloads.

The trope: the evidence locker.

Based on: Data Governance / De-identification

The idea: how a study tells you it needs attention, and what it costs to be told too much. An alert nobody can act on is noise, and noise is how the one that mattered gets missed.

The trope: the voice on the phone at 3am.

Based on: Study Monitoring / Alert Fatigue

Possibly spent. Largely covered by The Quiet Participant, where the alert is the break in the case — may not need a strip of its own.