Epidemiologic measures and inference in psychiatry (MEQ)

(i) Frequency measures. Point prevalence: existing cases at one time / population. Period prevalence: people who were cases during an interval (for example 12 months) / population. Cumulative incidence: new cases over a fixed period / population initially at risk. Incidence rate: new cases / person-time at risk.[1] Current community caseload for service planning is best informed by period (especially 12-month) prevalence plus severity/need, not lifetime prevalence or crude hospital counts alone.[1]

(ii) RR, OR, PAF. RR compares risks (risk_exposed / risk_unexposed). OR compares odds; in case-control work it is the natural association measure and approximates RR when the outcome is uncommon, but overstates relative risk if misread as RR when outcomes are common.[1] PAF estimates the proportion of population cases attributable to an exposure if the association is causal and estimates are unbiased; it depends on exposure prevalence and association strength and is easily misused when assumptions fail or correlated factors are naively summed.[3]

(iii) Bias families (examples). Selection: non-response or clinic-only sampling enriching trauma and depression co-occurrence.[2][8] Information: recall bias — depressed adults over-reporting childhood adversity relative to controls.[2][8] Confounding: childhood socioeconomic disadvantage as a common cause of adversity exposure and later depression risk, distorting the crude OR if uncontrolled.[2]

(iv) Screening. With Sn and Sp both 90% but prevalence about 2%, false positives greatly outnumber true positives among screen-positives, so PPV is low — most “positives” are not cases.[4][5] Before rollout, require Wilson–Jungner-style programme conditions: important condition, acceptable accurate test, agreed pathway and capacity for assessment/treatment, acceptable false-positive burden, cost balance, and quality-assured continuous programme — not test marketing alone.[6]

(v) Rewrite claim. Defensible language: “In this cross-sectional survey, childhood adversity was associated with higher odds of adult depression (OR 2.8). A model-based PAF near 40% would require strong causal assumptions, absence of major bias/confounding, and correct exposure prevalence; cross-sectional data cannot establish temporality. Longitudinal and intervention evidence are needed before claiming that adversity causes 40% of depression cases.” Use Hill viewpoints as weight-of-evidence, not proof.[2][3][7]