An occupational health advisory board is reviewing the implementation of a new serum biomarker assay used to screen for industrial heavy metal toxicity. The board is evaluating how different policy decisions regarding diagnostic thresholds and target populations will impact the statistical performance of the test. Match each proposed policy change to its most direct statistical consequence.
- Lowering the minimum serum biomarker concentration required to diagnose toxicity.Increases sensitivity and negative predictive value while decreasing specificity.
- Requiring a significantly higher serum biomarker concentration for a positive diagnosis.Increases specificity and positive predictive value while decreasing sensitivity.
- Moving the assay from a general workforce screen to a specialized clinic for workers with respiratory symptoms.Increases positive predictive value and decreases negative predictive value without changing inherent test characteristics.
- Expanding the screening program to include administrative office staff with no environmental exposure.Decreases positive predictive value and increases negative predictive value without changing inherent test characteristics.
Answer
Lowering the cutoff increases sensitivity/NPV; raising the cutoff increases specificity/PPV; testing a higher prevalence population increases PPV; testing a lower prevalence population decreases PPV. Prevalence shifts do not alter sensitivity or specificity.
Changes in a diagnostic test's performance must be separated into cutoff shifts versus prevalence shifts. Lowering a diagnostic cutoff increases the test's sensitivity and NPV, while decreasing its specificity and PPV. Raising the cutoff has the exact opposite effect, increasing specificity and PPV while decreasing sensitivity and NPV. Alternatively, altering the screened population only changes the pretest probability (prevalence) without affecting inherent test characteristics. Screening a higher-risk symptomatic population increases the PPV and decreases the NPV. Screening a lower-risk unexposed population decreases the PPV and increases the NPV. In both of these prevalence shift scenarios, the intrinsic sensitivity and specificity of the assay remain totally unaffected.
Step-by-Step Solution
Key Concept
Impact of disease prevalence and diagnostic thresholds on test characteristics and predictive values.
Estimated Time:1m 30s