Screening, Assessment, Evaluation, and Diagnosis: Tools, Scores, and Convergence
Domain 2 tests whether you can choose the right measure, read scores correctly, and separate difference from disorder. The exam favors psychometric reasoning and clinical decision-making over memorized test names.
Screening is not diagnosis
Screening flags risk quickly and cheaply; it never establishes a diagnosis or eligibility. Sensitivity is the proportion of true cases correctly flagged; specificity is the proportion of unaffected people correctly cleared. High sensitivity suits screening (few false negatives). Positive predictive value collapses when prevalence is low, so many "fails" in a general population are false alarms. A failed screen triggers assessment, nothing more.
Reading scores
- Standard scores: mean 100, SD 15. Scaled subtest scores: mean 10, SD 3. Percentile ranks are not percent correct.
- The standard error of measurement generates a confidence interval. A standard score of 85 with a 95% CI of 81–89 does not differ reliably from average, so one low score cannot carry a diagnosis.
- Use the standard error of difference before claiming a subtest gap is real.
- Basal and ceiling rules change the raw score; translated items or extra cues are nonstandard administration and invalidate the norms, so report them as nonstandard.
Diagnosis versus eligibility
Diagnosis requires convergence: case history, standardized and criterion-referenced measures, a language sample, dynamic assessment, and evidence of functional impact. Under IDEA, eligibility needs a disability plus adverse educational impact; a clinical label alone does not qualify a student, and criteria vary by state.
Worked example
A 7-year-old sequential bilingual child scores 78 (7th percentile) on an English-only vocabulary test. A bilingual language sample shows typical grammar in L1, conceptual scoring across both languages places him near the mean, and dynamic assessment shows he learns new words after two exposures. Reasoning: the standard score indexes English exposure, not capacity; convergent data support language difference. Recommendation: no language disorder label; monitor and support English vocabulary growth.
Traps that cost marks
- Treating a failed screening as a diagnosis.
- Reading a percentile as percent correct.
- Calling a small subtest gap significant without the standard error of difference.
- Ignoring the confidence interval when a score sits near a cutoff.
- Using a translated test and reporting standard scores as if norms applied.