Missing Data Contact Log and Request Templates for Systematic Reviews

A private email can change a public synthesis

Primary reports often omit statistics, definitions, denominators or methodological details required by a systematic review. Contacting study authors may clarify the report or provide additional data, but the exchange creates a new evidence source. If the request, response and interpretation are not recorded in your Systematic review data extraction form template, the analysis can contain information that readers and future reviewers cannot reconstruct. A value arriving by email should therefore be governed with the same attention to identity, provenance and transformation as a published value.

Contact is not always successful and should not be described as a guaranteed remedy. In one evaluation, additional data were obtained from 28 of 45 contacted authors while many requested items remained unresolved; the result is informative for that setting, not a universal response probability.¹² Methods for seeking unpublished information show variable yield and practical constraints.¹¹ The review should define why contact is needed, which sources were searched first, what will be requested, how follow-up will be handled and what analysis rule applies if the query remains unresolved.

PRISMA 2020 asks authors to report processes used to obtain or confirm data from investigators and how many reviewers collected information.^2,3 Transparent reporting begins with a complete internal event history. The published method may summarize the process, but the log should retain study ID, field, source checks, dates, channel, request, responses, files, reviewer interpretation and analytical status.

Contact only after the question becomes analytically precise

First establish whether the missing information is required for a planned analysis or a material interpretation. Search all linked reports, supplements, registries and repositories. Confirm that the apparent gap is not caused by using the wrong report, population or time point. Define the exact statistic or clarification needed, including arm, outcome, time and denominator. A broad request for “all missing data” increases burden and makes the response difficult to interpret.

Consider proportionality and consistency. Older studies may have inaccessible records; large numbers of contacts may exceed project resources; differential success can create selective availability. Apply the contact policy consistently to comparable studies and document exceptions. No universal follow-up interval or number of attempts fits every discipline, study age and channel. Predefine a reasonable review-specific approach and stopping rule.

Write a request that can be answered without reconstructing the review

Identify the review and study unambiguously. State the relevant report, describe the exact field and explain its analytical use. Provide the values already understood so the author can correct the premise. Ask one or a small set of structured questions, specify preferred units or format, and offer a secure method for transmitting files where needed. Avoid attaching copyrighted material unnecessarily or requesting participant-level data through ordinary email.

The message should not pressure authors to produce a preferred result. Neutral wording asks for clarification or underlying values and acknowledges that records may be unavailable. Provide a realistic response window appropriate to the project, but do not present it as a methodological standard. If a follow-up is planned, state the internal rule and log it as a separate event.

Keep personal data to the minimum necessary. Store contact details and correspondence in an approved institutional location with access controls. The public log or manuscript should report process and evidence status, not private addresses or irrelevant personal information.

Classify the response before using the value

A reply can clarify wording already in the report, correct a published value, supply previously unreported aggregate data, decline the request, report that records are unavailable, or provide information that remains insufficient. These states should not be collapsed into “author responded”. Record the message or file location, date, respondent’s relationship to the study where known, and the reviewer’s interpretation.

Check internal consistency and compatibility with the published study. Confirm units, arm, population, time point, analysis method and denominator. If the response conflicts with a report, preserve both and document the resolution; do not silently replace the publication. Where a new value requires calculation, retain the supplied inputs and the derivation. Independent review is particularly important when author-supplied outcome data enter a meta-analysis or are compressed into your Study characteristics table template.

Non-response and selective response remain part of the evidence problem

No response does not prove that data do not exist, that the study is flawed or that the missing value has a particular magnitude. Apply the prespecified missing-data or exclusion rule and retain the unresolved status. Report the number and purpose of contacts and the response outcome at an appropriate level. When only some studies provide additional data, consider whether availability may be associated with study age, result, sponsor or other characteristics.

Sensitivity analysis may compare results with and without author-supplied or imputed values when scientifically justified. The extraction dataset should contain a source-status field so such analyses are reproducible. Do not mix published and supplied values without marking their provenance. Contact can reduce uncertainty for a field while introducing a different uncertainty about selection or comparability.

Plan the contact portfolio, not only individual messages

When several studies have missing information, maintain a query register before correspondence begins. It should show the field, analytical priority, studies affected, alternative method, planned channel and current status. This portfolio reveals whether the team is applying the policy consistently and whether a large contact burden threatens the timeline. It also prevents different reviewers from sending overlapping requests to the same investigator.

Batching questions for one study can reduce burden, but only when the message remains navigable and each query has a stable identifier. If investigators are associated with several studies, distinguish them clearly. At reporting, summarize contacts using denominators that match the question: studies contacted, investigators approached, messages delivered and data items resolved are different measures. Retain both study-level and query-level records so that a partial response is not misclassified as complete resolution.

Use the log as the authoritative event index

The browser tool drafts one request and exports one record. The Word resource expands it into a contact-policy worksheet, event ledger, response-assessment form and adaptable message templates. Store actual correspondence and attachments in the approved repository and reference them by controlled location rather than embedding sensitive content in the log.

At analysis freeze, review all open queries, apply the prespecified rule, identify values changed by contact and record the dataset version. Later replies should create a new event and version rather than silently modifying the published analysis file. The event history allows an update team to determine whether a query was never attempted, remained unanswered or was resolved after the original reporting cut-off.

Prepare reporting directly from the event ledger

Use the log to describe the contact eligibility rule, number of studies and queries, delivery or response outcomes, information obtained and analytical treatment. Avoid a single response percentage without its denominator and purpose. If author contact altered inclusion, an outcome value or a sensitivity analysis, state that consequence. If no response was received, report the prespecified fallback rather than implying that missing information was absent.

The log should support concise reporting without exposing private correspondence. A manuscript can describe the method and aggregate results while the controlled repository retains messages, attachments and decisions under institutional policy. This separation preserves both transparency and privacy.

Distinguish three common query types

A clarification query asks what a published value means, for example, whether a denominator represents all randomised participants or complete cases. A numerical query asks for a specific missing input, such as a standard deviation for one arm and time point. A methodological query asks how a procedure was performed, such as the method used to allocate participants or define an outcome. Each requires different supporting context and may lead to a different destination in the review record.

For a clarification, quote or paraphrase the ambiguous passage and ask the author to confirm the interpretation. For a numerical request, list the values already available and specify the requested statistic, unit, population and time. For a methodological request, explain which assessment or eligibility decision depends on the answer without asking the author to make that review judgment. A response may resolve one query while leaving others open, so assign stable query IDs rather than one study-wide “responded” flag.

If authors provide a spreadsheet or analysis output, retain the original file, record its date and relationship to the request, and extract from a controlled copy. Do not modify the source file to fit the review. Document calculations in the review dataset and preserve the correspondence source separately.