Running a PMCF survey
Running a PMCF survey turns its plan into fielded data and a report. It involves programming and testing the survey, choosing software that can show where its data came from, justifying the survey's position on ethics review, recruiting respondents, managing fieldwork and analysis, and carrying each round's learning into the next.
Programming and testing
A PMCF survey is most commonly programmed online and hosted on survey software. The correct programming notes, routing and logic notes need to be added to the questionnaire design, because whether a question is shown can depend on the answer to an earlier one. The survey links need to be tested thoroughly, because a link that does not work correctly risks missing data at analysis, and missing data could severely affect the results, especially for the endpoints.
Survey software
A PMCF survey needs the correct hosting software. PMCF surveys sit in what one practitioner sees as a grey area, and many hold that their software needs to follow a clinical trial's electronic data capture system, which would fall under some requirements of ISO, US regulation and GCP. In that view, not all of that guidance needs to be followed, but parts of it do: validation, meaning full testing that shows the questions asked return the data wanted; routing and logic capability; control over who has access to the collected data to amend it; and an audit log and change log, so that where the data in the report came from can be shown later. SurveyMonkey and Qualtrics are not fully aligned with some of these requirements. Other tools are available, and the validation, the testing and the flow of data from the respondent through the software to the analysis all need to be documented.
Ethics review
A case-specific survey collects data about individual patients, which raises the question of where the line falls between needing ethics committee approval or not. Not seeking ethics committee or institutional review board approval can be justified on several grounds. The respondent may go back to patient charts or records for some of the data, but enters no patient-identifiable information into the survey. The analysis is often of aggregated data, which again mitigates the risk of patient-identifiable information being shared. And no physician is asked to go out of their way to use the device on a patient for the survey. The collection stays in line with their current practice, and it is retrospective.
The physician has access to the identifying information but does not provide it. So the survey is designed not to ask for it, for example by asking for the patient's age range rather than their age.
Recruitment and fieldwork
Recruitment settles how the survey reaches its respondents. A manufacturer that distributes the link to its existing customers may add bias to the results, because the survey asks about adverse events and complications and the request comes from the manufacturer itself. A third-party healthcare-professional panel could be used instead, and whether respondents will be compensated for their time is a further question. Managing the fieldwork is quite resource-heavy in staying on quotas, timelines and quality. As the data comes in, it is important to ensure that it gets a quality control check.
Analysis and deliverables
The data then needs to be tabulated, collected and processed so that it can be reported. As the data comes in, further questions arise: which tabulation software is used, how the data is split, how it is analysed statistically and who is responsible for that. It is really important that the analysis aligns the data with the acceptance criteria and the statistical analysis. The data is analysed in line with the statistical analysis in the survey protocol, and the deliverables are presented back to the manufacturer, so that it can pull the data into any relevant documentation, such as its risk management documentation.
Deliverables come in different forms: a summary of the findings as counts, percentages and charts in a report such as a PowerPoint deck; the raw individual-level data, alongside which it is important to have a formal summary; and the tabulated data with the statistical analysis, which can be imported into the final report.
The next survey round
PMCF is a continuous process, and it is really important to feed what a survey yields into the next round of PMCF data collection. The content built for a survey should be reused in later cycles and updated with what was learned. On a first round especially, nobody expects everything to be right, and what is learned from the data should be fed into the next round.