Most source documentation problems are design problems that were solved too late. The template gets built after three participants have already been seen, assembled from whatever notes the coordinator was taking in the meantime, and by then the first three records will never quite match the rest.

Coffeen, who is based in Lakewood Ranch, Florida, came to clinical research from patient care. He holds a Doctor of Chiropractic Medicine and spent years charting in settings where an incomplete note had immediate consequences for the person in the room. That background shapes a fairly blunt position: if you have not built the source document, you are not ready to enroll.

A worksheet is a clinical instrument

Sites sometimes describe source worksheets as administrative tools, something the data team wants. That framing is wrong and it produces worse documents.

A worksheet determines what the coordinator asks, in what order, and with what prompts. It shapes the encounter. A poorly sequenced worksheet makes the visit feel like an interrogation, and a worksheet missing a field means the coordinator has to remember an item that nobody wrote down. Design it the way you would design an intake form for a clinic, because functionally that is what it is.

There is a second-order effect that sites underestimate. A well-built worksheet makes a new coordinator competent faster, because the form carries the protocol knowledge that would otherwise live in someone’s head. Staffing changes stop being a data quality event.

Build it from the schedule of assessments, then check it against the protocol body

The schedule of assessments tells you what happens at each visit. It does not always tell you how, or under what conditions, or what to do when a window is missed.

Coffeen builds the skeleton from the schedule and then reads the protocol body specifically hunting for requirements the grid abbreviates. Fasting status. Position for a measurement.

Whether a questionnaire must be completed before the investigator examines the participant, because the order affects the answer. Those details belong on the worksheet as prompts, not in a coordinator’s memory.

The same review catches requirements that live in the safety section rather than the visit grid. Reporting timelines, definitions of what counts as an event for this protocol, and instructions for what to do when a participant reports something between visits all belong somewhere the coordinator will see them.

Leave room for what the protocol did not anticipate

A worksheet built only for expected answers forces people to write outside the lines, and writing outside the lines is where source documentation gets messy.

Every worksheet needs space for a participant comment that does not fit a field, a note explaining why something was done out of order, and a place to record that an assessment was not performed and why. If the form has nowhere to put an inconvenient fact, the inconvenient fact tends not to get recorded. That is a design failure, and it shows up later as a query nobody can resolve.

Coffeen builds in a comment field on every page rather than one at the end. The difference sounds trivial and it is not. A note written next to the assessment it concerns stays connected to that assessment, while a comment collected at the back of a packet loses its anchor.

Version the template like you version the protocol

When a protocol amendment changes an assessment, the worksheet has to change with it, and the site has to be able to demonstrate which version was in use on which date.

Sites that edit a worksheet in place, without a version number and without an effective date, create a documentation trail that cannot be reconstructed. Coffeen treats worksheet versions the same way he treats protocol versions, with a number, a date, and superseded copies retained rather than deleted. It costs nothing while the study is running and it saves a great deal during an inspection.

A source document is supposed to tell the story of what happened to a participant on a given day, completely enough that someone else could reconstruct it. You cannot retrofit that. You either built the instrument that captures it, or you spend the rest of the study working around the one you did not build.