The Screening Data That Doesn't Add Up

Treat 38% as a hypothesis, not a fact. Three sources of error are visible in the setup:

1. Administration variance - substitute, log-in failures, incomplete records. Those 9 students aren't data; they're missing data, and excluding them (or coding them as failures) biases the rate in opposite directions.
2. Instrument change - a longer window on the same measure is not the same measure. A 17-point jump in one semester is more likely a change in the ruler than a change in the children unless something else explains it (new curriculum, new population, a real disruption).
3. Seasonal effect - fall screening after summer break typically dips; compare to last fall, not last spring.

Before Monday:

• Re-screen or complete the 9 missing students within the week; do not let them sit in the denominator.
• Pull last fall's third-grade rate and this year's second-grade spring rate to see whether the jump survives a like-to-like comparison.
• Cross-check with a second source: running records, unit assessments, teacher nomination. If screening says 38% and three other sources say 22%, trust the convergence.
• Then sort the confirmed intensive group by skill (phonemic awareness vs. fluency vs. comprehension) - a group that is 38% intensive for three different reasons cannot be served by one purchased program.

Agree with the urgency, disagree with the sequence - and make the cost concrete.

• Capacity arithmetic: if 38% of third grade is roughly 45 students and Tier 2 groups run 4–5 students per 30-minute block, that is 9–11 groups. With one reading specialist, a 60%-fidelity history, and no interventionists hired, the plan produces groups on paper and instruction that doesn't happen. Over-placement also pulls the strongest Tier 1 teacher time out of the core block, which is how a school creates the next year's 38%.
• Fidelity is the binding constraint, not placement speed. Last year's groups ran 60% of scheduled minutes. Before adding students, find out why - scheduling conflicts, pull-out during core instruction, absenteeism, staff coverage. Doubling caseload onto a broken schedule will drop fidelity further.
• Propose a defensible two-track rollout: place the students with confirmed data and clear skill profiles now; hold the ambiguous cases for two weeks of diagnostic follow-up in Tier 1 small groups while re-screening is finished. Give the principal a written timeline with dates so "later" isn't indefinite.
• Bring the disproportionality check: before finalizing groups, look at who is being placed by subgroup, language status, and classroom. A jump concentrated in one classroom is an instructional signal, not a student deficit signal.

Work from the cheapest explanation outward - do not start by blaming the children.

1. Is the intervention actually happening? Check logs and do two unannounced walkthroughs. Flat data with 60%-type fidelity is a delivery problem, not an intervention problem.
2. Is the intervention matched to the skill? A fluency intervention given to a decoding problem produces flat data reliably. Sample the students' errors: are they guessing at words, or reading accurately but slowly?
3. Is the dose enough? Three 20-minute sessions a week is below most Tier 2 recommendations; check minutes per week against the program's design, not against what fits the schedule.
4. Is the measure sensitive to change? Weekly probes should show a trend in 4–6 weeks. If the probe is at the wrong difficulty level, the data will look flat even when the student is learning.
5. Only then consider changing the intervention - and change one variable at a time, so the next four weeks of data tell you something.

Document the sequence. If the student is later referred for special education evaluation, the team must be able to show that the intervention was delivered with fidelity and was appropriately matched - otherwise the evaluation is testing the school's implementation, not the child.