Diagnostic Test¶

Diagnostic measurement procedure — instantiates Hidden-Type Screening

Applies a standardized measurement with known error rates to reveal a hidden state — disease, defect, readiness — and reads the result against the population's base rate.

A Diagnostic Test screens by taking a direct physical or informational measurement whose relationship to a hidden state is characterized in advance. Unlike a self-report or a credential, it produces a fresh observation — a marker, a reading, a pass/fail — and unlike a subjective assessment, its errors are quantified: it has a known rate of missing true positives (sensitivity) and a known rate of flagging true negatives (specificity). That quantification is what makes this mechanism distinctive. Because its error profile is known, the whole art of using it is choosing where to set the positive threshold given which error is worse, and — since no single test is perfect — routing ambiguous or high-stakes results to a confirmatory fallback rather than acting on one reading.

Example¶

A clinic deploys a rapid screening test for a condition that is present in only ≈1% of the population it serves. The test is good: it correctly flags ≈95% of true cases and correctly clears ≈95% of healthy people. Yet because the condition is rare, most positive results are still false alarms — among the flagged, healthy people vastly outnumber the few true cases, so a positive result raises the probability of disease far less than the ≈95% figures suggest. The clinic anticipates this. It treats a positive not as a diagnosis but as a trigger for a slower, more specific confirmatory test, and it counsels patients that a screen-positive most often resolves to disease-free.

The screen's value is not that it is decisive — it isn't — but that it is cheap, fast, and sensitive enough to catch nearly all true cases while safely deferring the expensive confirmatory measurement to the small flagged subset.

How it works¶

Measure directly. Produce a fresh, standardized observation of the hidden state rather than relying on claims or records.
Set the threshold by error cost. Choose the positive cut-point to trade sensitivity against specificity according to which mistake — a miss or a false alarm — is worse in this use.
Read the result against the base rate. Interpret a positive using how common the condition actually is, not the test's headline accuracy, because rarity inflates false positives.
Confirm high-stakes positives. Route flagged cases to a more specific fallback measurement before any consequential action, rather than acting on a single screen.

Tuning parameters¶

Positive threshold — where the cut between positive and negative sits. Lower it to catch more true cases (fewer misses) at the cost of more false alarms; the right height follows the error cost.
Sensitivity/specificity target — which error the test is optimized against. A rule-out screen maximizes sensitivity; a rule-in confirmatory test maximizes specificity.
Confirmatory sequencing — whether a positive triggers a second, independent test. Adds cost and delay but sharply cuts acting on false positives.
Retest interval — how often the state is re-measured. States change; a one-shot reading goes stale, while frequent testing costs more and multiplies false positives over time.

When it helps, and when it misleads¶

A diagnostic test is the right screen when the hidden attribute is a measurable state and a validated instrument exists — it delivers a fresh, objective reading with an honest error budget, which few other screens can claim. Its signature failure is base-rate neglect: applied to a low-prevalence pool, even an accurate test produces mostly false positives, so a positive result read as near-certainty misleads badly.^[1] It can also be misused as a blunt gate — deploying a rule-out screen where a rule-in decision is needed, or treating a screen-positive as a diagnosis and acting irreversibly on it. The disciplines that keep it honest are interpreting every result through the population's actual base rate, matching the threshold to which error the domain can least afford, and confirming consequential positives with an independent measurement before commitment.

How it implements the components¶

screening_signal_or_test — it produces the observed signal directly: a fresh, standardized measurement of the hidden state.
false_positive_false_negative_policy — its known sensitivity/specificity make the miss-versus-false-alarm trade explicit and tunable via the threshold.
fallback_measurement_path — confirmatory testing of positives is a built-in second channel for when one reading is not enough to act on.

It does not set the downstream routing consequences of a positive across a whole pool (that is Risk Scoring Model and Underwriting Assessment), nor review the fairness of who gets tested (that belongs with Structured Application and Structured Interview).

Instantiates: Hidden-Type Screening — the direct-measurement-with-known-error variant of a screen.
Sibling mechanisms: Risk Scoring Model · Work Sample or Audition · Underwriting Assessment · Credential Verification · Self-Selection Menu · Background Check · Reference Check · Structured Application · Structured Interview · Pilot Project · Probationary Period · Challenge or Proof-of-Work

Notes¶

The most consequential decision in deploying a diagnostic screen is whom you test — the base rate of the tested pool determines what a positive means. Screening a broad, low-prevalence population with even an excellent test can generate more harm from false positives than benefit from true catches, which is why the pool boundary and the confirmatory fallback are inseparable from the test itself.

References¶

[1] The base-rate fallacy — reading a test's positive result by its sensitivity/specificity while ignoring how rare the condition is. When prevalence is low, the positive predictive value (the chance a positive is a true case) can be small even for a highly accurate test, which is why a screen-positive should trigger confirmation rather than action. ↩