Prompt Chain: Evaluation Report Generator from Structured Data

For Speech-Language Pathologists

Tools: Claude Pro | Time to build: 2-3 hours | Difficulty: Intermediate-Advanced Prerequisites: Comfortable using Claude for documentation. See Level 3 guide: "Use Claude Pro to Draft Evaluation Reports"

What This Builds

A multi-step prompt chain that takes your raw evaluation data (case history bullets, standardized test scores, clinical observations) and produces a complete, structured evaluation report section by section, with each section's output feeding into a more coherent final document. Instead of a one-shot "write my eval report" prompt (which produces generic output), this chain produces reports that feel like they were written by a clinician who actually knows the case. SLPs using this system report reducing evaluation report writing time from 6-10 hours to 1.5-3 hours.

Prerequisites

• Claude Pro account ($20/month), since long reports exceed free tier limits
• A completed evaluation with test scores scored and recorded
• Your standard eval report template (if you have one, upload it to the Claude Project)
• Optional: a Claude Project set up (see Level 4 guide: "Your Personal SLP Documentation Assistant")

The Concept

A prompt chain is a series of connected prompts where you complete them in order so that each step's output informs the next step. Writing an eval report is naturally a chain: background → testing → observations → diagnosis → recommendations. Running these as separate, connected prompts produces better output than asking for the whole report at once, because each step focuses Claude's attention on one task at a time, and later sections can reference what was already established in earlier ones.

Build It Step by Step

Part 1: Set Up Your Data Sheet

Before opening Claude, fill out this data sheet for the evaluation. You'll paste from it throughout the chain. Takes 10-15 minutes, and most of this information already exists in your notes.

=== EVALUATION DATA SHEET ===

PATIENT PROFILE
- Age: [X years, X months]
- Sex: [M/F]
- Referral source: [who referred and why]
- Primary concern: [presenting complaint in 1-2 sentences]

BACKGROUND HISTORY (bullet points)
- Developmental history: [milestones, any delays noted]
- Medical history: [relevant diagnoses, medications, hearing/vision]
- Educational history: [grade, any services, teacher concerns]
- Family history: [family history of communication disorders if relevant]
- Prior speech-language therapy: [yes/no; if yes, duration, focus, progress]

BEHAVIORAL OBSERVATIONS
- Cooperation/engagement: [description]
- Attention/focus: [description]
- Communication attempts: [did they initiate? respond? gesture?]
- Rapport with examiner: [description]
- Any notable behaviors: [fatigue, frustration, anxiety, etc.]

STANDARDIZED TEST RESULTS
- [Test name 1]: Standard Score [X], Percentile [X]th, Confidence Interval [X-X]
- [Test name 2]: Standard Score [X], Percentile [X]th
- [Test name 3 or subtest]: [score] with a brief description of what it measures
- [Continue for all tests administered]

INFORMAL ASSESSMENT FINDINGS
- [Connected speech sample / language sample / informal observation]: [description of findings]

CLINICAL IMPRESSIONS
- Diagnosis/diagnoses: [your clinical conclusion]
- Severity: [mild / moderate / severe]
- Prognosis: [good / fair / guarded, with rationale]

RECOMMENDATIONS
- Speech-language therapy: [yes/no; frequency/duration if yes]
- Other referrals: [audiology, OT, psychology, pediatrician, etc.]
- Home program: [yes/no; focus area]
- School/team communication: [what you want communicated]

Part 2: Run the Prompt Chain in Order

Open Claude (ideally your Documentation Assistant Project). Run these prompts one at a time in the same conversation, waiting for each response before continuing.

Prompt 1. Background Section:

I'm writing a speech-language evaluation report. I'll share my evaluation data section by section. Please draft each section as a formal clinical report narrative when I provide the data. Start with the Background/Case History section.

Background data:
[PASTE your Background History bullets from the Data Sheet]

Patient: [age]-year-old [sex] referred by [source] for [reason].

Prompt 2. Behavioral Observations:

Now draft the Behavioral Observations section from this data:
[PASTE your Behavioral Observations bullets]
Write 1-2 paragraphs. Note implications for test validity where relevant.

Prompt 3. Assessment Results:

Now draft the Assessment Results section. For each test, write 1-2 sentences interpreting the score clinically (not just restating the number). Include an overall profile statement at the end.

Test results:
[PASTE your Standardized Test Results]

Informal findings:
[PASTE your Informal Assessment Findings]

Prompt 4. Diagnostic Impressions:

Based on the background, observations, and test results you've written about so far, draft the Diagnostic Impressions section. Use this clinical conclusion:
Diagnosis: [diagnosis]
Severity: [severity]
Prognosis: [prognosis with rationale]

Include appropriate hedging language ("consistent with," "findings support," "presents with"). Reference the key evidence from prior sections.

Prompt 5. Recommendations:

Draft the Recommendations section. Recommendations:
[PASTE your Recommendations bullets]

Format as a numbered list with brief rationale for each recommendation. Include functional impact language connecting the recommendations to daily life/school/work.

Prompt 6. Executive Summary (optional):

Finally, write a 2-3 sentence summary paragraph suitable for the top of the report as an executive overview. It should capture: the presenting concern, key findings in plain language, and the primary recommendation.

Part 3: Assemble and Review

Copy each section from Claude into your report template in order. Then do a final review pass:

• Do the sections flow together and reference each other coherently?
• Are all test score numbers accurate? (Compare to your scoring forms)
• Does the Diagnostic Impressions section logically follow from the Results?
• Does the Recommendations section address what the referral question was asking?
• Remove any [REVIEW THIS] flags before finalizing

Real Example: Pediatric Language Evaluation

Patient setup: 7-year-old male referred by first-grade teacher for concerns about reading readiness and following classroom directions.

Data Sheet excerpt:

• CELF-5 Core Language Score: 79, 8th percentile
• CELF-5 Language Memory Index: 72, 3rd percentile
• Phonological Awareness (DIBELS): Below benchmark
• Informal language sample: MLU 5.2, age-appropriate; reduced use of complex syntax; frequent revisions and mazes

Prompt 3 output example (Assessment Results excerpt): "Standardized assessment revealed performance in the Low Average to Borderline range across language domains. On the CELF-5, the patient earned a Core Language Score of 79 (8th percentile), indicating language skills below age expectations. The Language Memory Index (72, 3rd percentile) was particularly notable, suggesting difficulty retaining and recalling linguistic information, consistent with classroom reports of difficulty following multi-step directions. Phonological awareness assessed via DIBELS fell below benchmark, placing him at increased risk for reading difficulties. A spontaneous language sample demonstrated age-appropriate mean length of utterance (MLU 5.2); however, reduced use of complex syntactic structures and increased disfluencies (revisions, mazes) were noted, suggesting effortful language formulation."

Time used: The full 6-prompt chain took 35 minutes to run. Final report assembly and review took 45 minutes. Total: 80 minutes vs. a historical average of 6 hours.

What to Do When It Breaks

• Section 4 (Impressions) doesn't connect to earlier sections → Add "Refer to the test results and observations from earlier in this conversation when writing this section" to the prompt
• Output is too long → Add "Keep this section under [X] words" to any prompt
• Test score interpretation is generic → Add the test's full name and what it measures: "CELF-5 measures language processing across receptive and expressive domains". Claude interprets scores more accurately with this context
• Hit a message limit → Upgrade to Claude Pro, or start a new conversation with a summary of what's been written so far

Variations

• Simpler version: Run all 6 prompts in a single long prompt if your report is short and simple; works for brief evaluations with 1-2 tests
• Extended version: Create a Claude Project with your standard eval report template uploaded, so the AI can see your exact format and reproduce your headers, formatting, and style precisely

What to Do Next

• This week: Run this chain on a recently completed evaluation you've already written. Compare the output to your report and calibrate what additional input produces better output
• This month: Build a version of this chain for each of your 3-4 most common evaluation types (speech sound, language, voice, dysphagia)
• Advanced: Upload de-identified example reports to your Claude Project so the AI can match your exact writing style

Advanced guide for speech-language pathologist professionals. These techniques use sophisticated AI features and require a paid Claude subscription. Always review AI-generated clinical documentation before signing.