Assessment: needs, capacity, and data you can defend
Assessment answers one question before anything else is designed: what does this population actually need, and what assets exist to meet it? Items here test how you choose data sources, judge data quality, and defend a prioritisation decision.
Core rules
- A needs assessment moves from purpose → data collection → analysis → gap identification → prioritisation. Skipping the purpose statement is the most common design flaw; purpose determines method, not the reverse.
- Primary data is collected by you (surveys, interviews, focus groups, observation, windshield tours). Secondary data already exists (vital statistics, BRFSS, census, hospital discharge data, county health rankings). Secondary data is cheaper and faster but may not fit your question or your subgroup.
- Quantitative methods answer how many/how much; qualitative methods answer why/how. Key informant interviews and focus groups never generalise to the whole population.
- Probability sampling (simple random, systematic, stratified, cluster) supports generalisation; nonprobability methods (convenience, purposive, quota, snowball) do not.
- Reliability is consistency of a measure; validity is whether it measures what it claims. A scale can be reliable and invalid.
- PRECEDE-PROCEED opens with social and epidemiological diagnosis: quality of life, the health problem, then behavioural and environmental determinants.
- Prioritisation criteria typically include magnitude, severity, feasibility, community concern, cost, and existing capacity.
Worked example
A county wants to know why prenatal visit attendance is low. Staff survey 120 women at one WIC clinic and find 55% missed at least one visit; someone proposes declaring this the county rate. Reasoning: the sample is convenience and venue-based, so women with no WIC contact - often the highest-risk group - are structurally excluded. The finding is a useful signal, not a prevalence estimate. Fixes: add address-based random sampling, triangulate with birth certificate and Medicaid claims data, and run focus groups with women who dropped out of care.
Traps that cost marks
- Treating a high response rate as proof of representativeness; it only reduces nonresponse error within the sampled frame.
- Confusing assessment with evaluation (assessment precedes the programme; evaluation judges it).
- Presenting a needs assessment with no community input - professional judgement alone is not a needs assessment.
- Calling a windshield tour quantitative observation data.