Supporting Investigator-Initiated Studies When Site Data Lives in PDFs, Scans and Mixed Formats

Investigator-initiated studies are one of the most meaningful ways medical affairs teams support scientific inquiry. They let clinicians pursue the questions they see in their own practice, and they often produce the local, real-world insights that company-sponsored studies cannot. In much of Asia, they also depend heavily on routine clinical records: discharge summaries, laboratory reports, imaging reports and clinical notes that sit in PDFs, scans, paper files and several systems, sometimes in more than one language.

When a study stalls, it is rarely because the question was wrong. More often, the data turned out to be harder to find, read or combine than the proposal assumed, and the investigator's team ran out of time or budget to close the gap. Medical affairs teams who understand this can review proposals more realistically, and investigators who plan for it can protect their study. This guide walks through what good data practice looks like at each stage.

A simplified view. Company policies, agreements and local rules set the exact boundaries.

Stage 1

Feasibility and proposal review

This is the stage where a realistic view of the data costs least and changes most. A proposal can describe an elegant question and a sensible sample size, yet say little about whether the records needed to answer it exist in usable form. For medical affairs reviewers, asking about data feasibility is part of assessing whether a proposal can deliver, and it does not cross into directing the study.

Where problems usually appear

1. The sample size assumes every eligible patient has complete records, when key variables are often missing for a large share of patients.
2. Critical variables such as disease stage, treatment response or reasons for switching live only in clinical notes or scanned reports.
3. The study team has not yet looked at real records, so the effort needed to extract data is a guess.

Practical options

Investigators can strengthen a proposal by reviewing a small number of eligible records before submission, noting where each key variable is found and how complete it is. Where most variables sit in unstructured documents, the proposal should say how they will be extracted and by whom, and the budget should reflect it.

Questions to discuss during proposal review
1. Where are the key study variables recorded today, and in what formats and languages?
2. Has the team reviewed real records to confirm completeness, and what did they find?
3. Is the planned sample size realistic once incomplete records are excluded?

Stage 2

Set-up

Once a study is approved, the set-up stage decides whether the data will hold together. This is when the investigator's team turns intentions into a data management plan, a case report form that matches how information is actually recorded, and the approvals that allow the data to be used.

Where problems usually appear

1. The case report form asks for information in a form the source records do not hold, such as a numeric score that clinicians describe only in words.
2. Definitions are left open, so different abstractors interpret the same record differently.
3. Pseudonymisation, storage and processing location are decided late, delaying ethics or data protection approval.
4. The budget covers the analysis but not the work of preparing the data.

Practical options

A short data management plan helps even a small study: it defines each variable and where it comes from, how records will be pseudonymised at source, where data will be stored and processed, and how quality will be checked. Piloting the case report form on a handful of real records before full collection starts is one of the most effective ways to catch mismatches early.

Questions for the investigator's team to settle
1. Is every variable defined clearly enough that two people reading the same record would record the same value?
2. Where will data be pseudonymised, stored and processed, and does that meet local requirements?
3. Is data preparation included in the study budget and timeline?

Stage 3

Conduct

During conduct, the practical work of turning records into study data happens, and it is usually where time runs short. Research coordinators, residents and nurses often abstract data by hand alongside clinical duties. When records are long, scanned or multilingual, progress can slow sharply, and consistency between abstractors becomes hard to maintain.

Where problems usually appear

1. Abstraction takes far longer per patient than planned, and recruitment targets slip.
2. Different people interpret the same types of records differently, and nobody notices until analysis.
3. There is no clear record of where each value came from, which makes queries hard to resolve.

Practical options

Study teams broadly have three approaches to extraction. Each can work, and the right choice depends on volume, complexity and resources. What matters most is that clinical judgement is applied where interpretation is needed, and that every value can be traced back to its source.

A general comparison. Results depend on how each approach is set up and checked.

Questions for the investigator's team to track
1. How long is each patient record taking, and is that in line with the plan?
2. Who reviews values that need clinical interpretation, and how are disagreements resolved?
3. Can every value in the dataset be traced back to its source document?

Stage 4

Analysis and publication

By the analysis stage, the quality of earlier decisions becomes visible. A dataset that was defined clearly, extracted consistently and checked clinically can be analysed with confidence. One that was not will generate queries, delays and, sometimes, conclusions that reviewers challenge.

Where problems usually appear

1. Data from different sites or periods cannot be combined without further cleaning.
2. Reviewers or journals ask how variables were derived, and the answer is not documented.
3. The prepared dataset is discarded after publication, so the next study starts from scratch.

Practical options

Keeping a simple record of how each variable was extracted, cleaned and transformed makes analysis faster and publication smoother, and it supports the transparency that journals and data sharing statements increasingly expect. Where permissions allow, keeping the prepared dataset and its definitions for future approved research means the institution's effort continues to create value after the first paper.

Questions for the investigator's team to plan for
1. Is there a documented record of how each variable was derived?
2. What data sharing or reproducibility commitments does the target journal expect?
3. Can the prepared dataset be retained for future approved research, and under what permissions?

The stages at a glance

Use this table when reviewing a proposal or planning a study. It summarises what good looks like at each stage and the signals that a study may struggle.

How Jonda Health can help

Routine clinical records hold the answers to many important questions, and turning them into research-ready data is the work we focus on. We can support investigators and institutions directly, and we can work with medical affairs teams on how data readiness is considered across their programmes, always within the boundaries that keep investigator-initiated research independent.