Do you have baseline data? Testing your problem statement
A literature review can show that a problem matters. It cannot establish how often that problem occurs at your project site. Before writing that your clinic has “poor follow-up” or your staff have “low knowledge,” look for evidence that supports the local claim.
SQUIRE separates the local problem from the broader knowledge available about it. [1] That distinction is useful before you draft the proposal: published evidence explains why a problem deserves attention, while local evidence describes the situation you plan to address.
Ask three questions
Who is affected? How much of the problem is present? Compared with what expectation or reference? A baseline does not have to be a large dataset, but its source, period, and limitations should be clear.
| Potential source | What it may tell you | What to check |
|---|---|---|
| Existing dashboard | The current reported pattern | Does its definition match your proposed measure? |
| Approved record review | How often a defined event is documented | Are eligible records consistently identifiable? |
| Preintervention assessment | Starting knowledge, skill, or experience | Can you use the same method later? |
| Staff or patient feedback | The nature of a perceived problem | Is it descriptive feedback or a measured frequency? |
Replace a broad claim with a bounded one
Hypothetical example. A manager says referral requests are frequently incomplete. An approved review identifies 60 eligible requests submitted during four defined weeks. Eighteen lack at least one required field. A defensible statement is: “In a review of 60 eligible referral requests, 18, or 30%, lacked at least one required field.”
That statement is more useful than “referral quality is poor.” It specifies the records, period, definition, and denominator. It does not establish why information was missing or claim that all requests throughout the year have the same rate.
You might compare the result with a local expectation, a previous period, or an external benchmark. Only make a direct numerical comparison when the definitions and populations are sufficiently similar. A published national figure is not automatically a suitable local target.
When no baseline exists
Do not invent one, replace it with a national statistic, or rely on staff memory as though it were measured data. Discuss whether a short prospective baseline or an approved historical review is possible. Retrospective data are useful only when the earlier records can support the same outcome definition.
When baseline recovery is not feasible, revise the question transparently. You may be able to evaluate implementation, participation, or current performance without claiming improvement from an unknown starting point. Confirm that this scope still meets program and site requirements.
Your next decision
Draft your problem statement using one local fact and one sentence about its limitations. Confirm the required permission before accessing records or collecting baseline data. See Valid, feasible, relevant: Choosing outcome measures for definitions and QI or research? How project review decisions work for review decisions.