Writing a clear study configuration

Write criteria for a stranger

Your inclusion and exclusion criteria are applied by the platform when it suggests screening labels, and by every person on the project. Write them so that someone who has never met you would apply them the same way: name the protein sources, material forms, measurement types, publication types and languages.

Avoid criteria that rely on unwritten knowledge. "Exclude poor-quality studies" is not a criterion; "exclude studies that do not report the protein concentration of the tested sample" is.

What the configuration holds

Each part of the configuration is used by a particular stage. A copy is kept with every run that used it, so earlier decisions stay readable after you revise the criteria.

• Protein scope and synonyms: the names screening and term recognition look for.
• Property scope and synonyms: which functional properties are collected.
• Search terms, date range and databases.
• Inclusion and exclusion criteria, and conditions that must be reported.
• Special rules: study-specific instructions shown again at review.
• The target unit for each property.
• Scoring rules and thresholds for ordering papers.
• Fields that a person must always check.

Choose target units before extraction

Each property has a target unit. The value is kept exactly as the paper reports it, and the converted value is stored beside it with the conversion that was used.

A conversion is made only when it is safe. Per cent w/v is not converted to per cent w/w without a density, and a wet-basis value is not converted to dry basis without the moisture content. Where a safe conversion is not possible, the value is flagged for a person.

Record what is out of scope

Also record what you decided not to collect. Months later, a note such as "foaming stability was not extracted" shows that a gap in the dataset was a decision and not a mistake, and saves a colleague from looking for values that were never collected.