Read evidence and design programs for the population
Define the outcome and denominator
Prevalence describes existing disease in a defined population at a specified assessment time. Incidence concerns new cases arising in people at risk during follow-up. Neither measure is meaningful without a case definition, denominator, and time frame. A survey of regular maintenance patients may provide useful clinic information but can misrepresent community disease because nonattenders may face different risks and barriers. Before comparing studies, check whether they measured the same condition, population, and outcome.
Distinguish association from causation
A cross-sectional study may reveal an association without showing which factor came first. A prospective cohort can establish temporal ordering, but confounding and loss to follow-up remain concerns. A conventional case-control study samples by outcome status and commonly estimates an odds ratio rather than direct population risk. Recall bias can occur if cases and controls report past exposure differently. Random allocation in a trial reduces systematic baseline differences probabilistically; allocation concealment prevents manipulation of enrollment, while blinding addresses later measurement or behavior. These safeguards solve different problems.
Translate numbers into useful decisions
A risk ratio compares risks by division. Absolute risk reduction subtracts the intervention risk from the comparator risk, and its reciprocal gives a number needed to treat for the stated outcome and follow-up. Keep relative and absolute effects separate. Confidence-interval width reflects precision, not freedom from bias. Statistical significance does not establish that an effect matters to patients. Interpret the size of benefit with uncertainty, harms, burden, and applicability to the population under consideration.
Evaluate screening and equity
Sensitivity is the proportion of true cases detected, and specificity is the proportion of people without the condition correctly classified negative. Predictive values also depend on disease prevalence. A positive result in a low-prevalence population can have a substantial false-positive component even when sensitivity and specificity are unchanged. Screening is a program, not just a test: consider confirmation, access, treatment, and follow-up. Equity recognizes that identical service offers do not give everyone an equal practical opportunity to benefit. Address language, cost, disability, transport, and other barriers using evidence and community input rather than assuming attendance alone measures need.