Turning a public-health problem into an answerable research question

#Question

How do you turn a public-health problem into a question a study can actually answer?

#Answer in brief

Treat the initial concern and the research question as different objects. A problem such as "vaccination uptake seems low" identifies a reason to investigate, but it does not yet define what evidence would count as an answer. A quantitative comparative question normally needs an explicit population, intervention or exposure, comparator, outcome, and time frame. PICO was developed to structure clinical questions; PECO makes the exposure explicit for observational questions, and PICOT/PECOT variants add time when it is essential to the estimand.^[1][2][3]

Structure is necessary but not sufficient. The resulting question must also be feasible, ethically answerable, relevant to a decision, and narrow enough that one study design can address it without silently changing its meaning.

#Key points

• The problem motivates; the question constrains. A problem statement can remain broad. A research question must specify what will be compared and measured.
• Choose the frame that matches the relation of interest. PICO suits many intervention questions; PECO is designed for exposure–outcome questions.^[2]
• Time and setting are often causal or interpretive, not decorative. "Did uptake change?" cannot be interpreted without a period, population, and comparison.
• Question structure should precede design selection. Choosing a familiar design first invites the design to redefine the question around what it can conveniently measure.
• A well-structured question may still be a poor research priority. It should be tested for feasibility, relevance, novelty, ethics, stakeholder value, and whether the likely answer could change a decision.^[3]
• Different questions require different frameworks. PICO is not a universal grammar for qualitative inquiry, descriptive surveillance, implementation questions, or complex systems; alternative structures may be more appropriate.

#Scope and exclusions

This guide focuses on comparative quantitative questions in public health, epidemiology, and health-services research. It does not prescribe a framework for every qualitative or mixed-methods question, choose a final study design, calculate sample size, determine regulatory classification, or replace stakeholder and ethics review.

#Step 1 — Separate the concern from the claim

A public-health concern often arrives as an observation:

• cases appear to be increasing;
• uptake seems lower in one area;
• an intervention may not be reaching the intended population;
• response appears slower than before;
• one group seems to experience worse outcomes.

Each statement contains an implicit comparison but leaves important choices unstated. Writing the implicit claim in full exposes those choices before they become hidden assumptions in the dataset or analysis.

#Step 2 — Identify the relationship being asked about

#Intervention question

Use an intervention-oriented frame when the question concerns something assigned, offered, implemented, or changed:

Population
+ Intervention
+ Comparator
+ Outcome
+ Time

Cochrane uses PICO as a standard way to define intervention-review questions while noting that implementation, equity, context, and complex systems may require additional dimensions.^[4]

#Exposure question

Use PECO when the question concerns an exposure rather than an assigned intervention:

Population
+ Exposure
+ Comparator
+ Outcome

Morgan and colleagues described PECO specifically for questions about associations between environmental or other exposures and health outcomes.^[2] Adding time produces PECOT when the temporal relation needs to be explicit.

#Descriptive question

Not every useful public-health question needs an intervention or exposure. Questions about burden, distribution, service coverage, timeliness, or data quality may be descriptive. Forcing a comparator into a purely descriptive question can create a false causal implication. The question should state the quantity to estimate, population, place, period, and measurement definition.

#Step 3 — Specify each component

#Population

Define who or what the inference concerns, not merely where the data came from. A database of clinic attendees is not automatically the same population as all residents who could have attended.

Ask:

• eligibility and exclusion criteria;
• geography and service setting;
• age or risk group;
• whether the dataset covers the target population;
• whether conclusions concern individuals, facilities, areas, or time periods.

#Intervention or exposure

Define what varies and how it is measured. "Access," "programme exposure," or "high pollution" is not specific until the operational definition, dose, assignment process, or measurement period is clear.

#Comparator

State the counterfactual or reference explicitly:

• no intervention;
• usual practice;
• another programme;
• lower exposure;
• an earlier period;
• another area;
• a threshold or target.

A weak comparator can make a precise outcome answer the wrong policy question.

#Outcome

Choose an outcome that is measurable, interpretable, and connected to the problem. Distinguish:

• process from outcome;
• proxy from patient or population consequence;
• first dose from completed schedule;
• notification from confirmed case;
• statistical significance from practical relevance.

#Time

Specify when exposure or intervention occurs, how long follow-up lasts, and when the outcome is measured. Time determines whether the proposed relation is temporally plausible and whether the study captures immediate, delayed, or sustained effects.

#Step 4 — Define the estimand in plain language

Before choosing a model, complete this sentence:

We want to estimate the difference or association in [outcome] between [comparison groups or conditions] in [population] over [time], under the stated definitions and assumptions.

This does not replace a formal estimand, but it reveals whether the question is about prevalence, incidence, risk, rate, mean difference, time to event, coverage, inequality, diagnostic performance, or another quantity.

#Step 5 — Test whether the question is worth and able to be answered

A structured question can still fail before data collection. Assess:

• Feasibility: adequate data, participants, expertise, time, and resources;
• Ethics: acceptable risk, privacy, consent or lawful data use, and appropriate oversight;
• Relevance: a real uncertainty for stakeholders or decisions;
• Novelty or confirmatory value: whether the work adds, tests, or transfers knowledge rather than repeating a settled result without purpose;
• Actionability: how plausible answers would alter policy, practice, research, or resource allocation;
• Equity: whether important groups are excluded by the question or data;
• Measurement validity: whether the proposed variables represent the concepts named in the question.

FINER is a commonly used mnemonic for feasibility, interest, novelty, ethics, and relevance, but it should support judgment rather than become a scoring ritual.^[3]

#Worked example: investigating slower notification response

#Felt problem

"Our region's outbreak-notification response has become slower."

This statement combines several possible processes: detection, notification, triage, case-definition assessment, investigation, and communication.

#First structured question

Among notifications received by the regional public-health unit in 2026, was the median time from receipt to case-definition assessment longer than for notifications received in 2025?

Element	Specification
Population	Notifications received by the same regional unit
Exposure	Calendar year 2026
Comparator	Calendar year 2025
Outcome	Time from receipt to completed case-definition assessment
Time	Two defined calendar years

#What the structure reveals

The question is answerable from routine operational data, but it does not prove that the whole response became slower. It examines one interval. Before analysis, the team should still ask:

• Did the case mix or notification volume change?
• Did definitions or workflow states change between years?
• Are timestamps generated consistently?
• Were urgent notifications prioritized differently?
• Is median time sufficient, or do extreme delays matter?
• Does a year-to-year comparison confound trend, seasonality, and system change?

The framework makes the scope visible; it does not remove the need for design and causal reasoning.

#A second example: vaccination coverage

#Concern

"Coverage is low in rural children."

Possible questions include:

1. Descriptive: What proportion of children aged 2–5 in rural districts had completed the specified schedule by 31 December 2025?
2. Comparative: Was documented completion lower in rural than urban districts after applying the same denominator and record definition?
3. Intervention: Did extending clinic hours increase completion within six months compared with similar districts retaining usual hours?
4. Equity: Did the intervention narrow or widen differences by distance, deprivation, or migration status?

These are not interchangeable. They require different data, assumptions, and designs even though they originate from the same concern.

#Evidence and method

Richardson and colleagues introduced the well-built PICO question as a method for making clinical questions explicit enough to search and act upon.^[1] Cochrane continues to use PICO for intervention-review scope while emphasizing stakeholder input and the additional complexity of context, implementation, and equity.^[4] Morgan and colleagues provide a direct methodological basis for PECO in exposure–outcome questions.^[2] Willis provides a recent practical account of combining PICO with FINER when formulating a research question and hypothesis.^[3]

The public-health examples in this guide are synthetic. They illustrate how to make hidden choices visible; they are not templates that determine a correct design without local expertise and stakeholder input.

#Practical implications

Before selecting a design or dataset, require a short question record containing:

• the problem and intended decision;
• the structured question;
• the target population and data population;
• intervention or exposure definition;
• comparator;
• primary outcome and measurement rule;
• time frame;
• feasibility and ethics constraints;
• key assumptions;
• important subgroup or equity dimensions;
• what a positive, negative, or inconclusive result would mean.

This record makes later changes inspectable. If the available data force a new population, proxy outcome, or comparator, the question has changed and should be versioned rather than silently rewritten in the analysis.

#Limitations and uncertainty

PICO, PECO, PICOT, PECOT, and FINER are aids to structured judgment, not validity certificates. A perfectly formatted question can still be trivial, unethical, infeasible, based on an invalid measurement, or disconnected from the decision that motivated it. Complex interventions, systems questions, qualitative inquiry, implementation research, diagnostic accuracy, prediction, and evidence synthesis may require additional or different frameworks. The methodological quality of the eventual study depends on the design and conduct that follow, not on the acronym used at the beginning.

#Connection to Weavidence

Weavidence Journey is designed to preserve the relationship between a question, the alternatives considered, the selected design, and the protocol generated from those decisions. That trace can make scope changes and assumptions easier to review; it does not guarantee that the original question was important or that the selected design is valid.