### Task

You are performing high-quality theme extraction and per-theme sentiment analysis for the {domain} domain to support reflective supervision, practice improvement, and targeted worker support.

Your task involves seven thinking steps.  
Steps 1 to 4 should be written out and shown. Step 5 is for internal reasoning only.  
Step 6 is the final output and should be shown.

---

### Step 1: Initial Themes with Sentiment 
Carefully read the COMMENT.  
List all distinct issues, conditions, supports, risks, constraints, or practices mentioned, in the order they appear.  
Next to each item, indicate its sentiment based on the practitioner’s tone and context, choosing one of:  
  – strongly negative  
  – negative  
  – neutral  
  – positive  
  – strongly positive  

If an item describes extra effort, conflict, or a workaround, mark it as negative or strongly negative.  

Format as a numbered list like:  
1. Topic phrase — sentiment  
2. Topic phrase — sentiment  
…  

---

### Step 2: Add Domain-Relevant Insights
Re-read the COMMENT using the QUESTION and its domain as a guide.  
Ask yourself:  
• What is the reflection focus for practitioners?   
• What would a policymaker care about here?  
• What barriers, supports, or missed opportunities matter most in this domain?  

Output:
1. A brief summary of your domain-guided insights.  
2. A bullet list of any additional issues not captured in Step 1:
   - New topic phrase  
   - Another new topic phrase  
   - …  

---

### Step 3: Filter, Merge, and Label Impact 
Review your combined list from Steps 1 to 2.  
• Remove trivial, event-based, or duplicate topics.  
• Merge overlapping items under one abstract label, when merging opposing dynamics (barrier + response), set impact to mixed.  
• Preserve impact labels from Step 1 when deciding what to merge or drop.  

Output the revised list as a numbered list of entries with their impact in parentheses:

1. Refined topic A (positive)  
2. Refined topic B (negative)  
3. Refined topic C (mixed)  
…

---

### Step 4: Standardise Theme Names
Rewrite each topic from Step 3 as a concise noun phrase (2 to 5 words) in Title Case:  
• Start with the core subject  
• End with a category word  
• Use modifier–head order  
• Avoid prepositions, personal or event references, or vague process words  

Output the fully standardised list as a numbered list:

1. Standardised Topic A  
2. Standardised Topic B  
…

---

### Step 5: Add Evidence, Impact, and Root Cause (internal)  
For each standardised theme:  
• Attach a verbatim quote from the COMMENT as `"evidence"`  
• Assign `"impact"` by reasoning how the theme helps or hinders real-world practice outcomes, choosing one of:  
  – positive  
  – neutral  
  – negative  
  – mixed  
• If `"impact"` is `"negative"`, infer a short `"root_cause"` phrase revealing a deeper structural, procedural, cultural, or relational factor (do not restate the evidence)  

Do not output this step.

---

### Step 6: Output Final JSON  
Convert all themes into this JSON:

{{  
  "themes": [  
    {{  
      "topic": "…",  
      "evidence": "…",  
      "impact": "positive|neutral|negative|mixed",  
      "root_cause": "…",    // or null  
      "sentiment": "strongly negative|negative|neutral|positive|strongly positive"  
    }}  
    // add more as needed  
  ]  
}}

Return the outputs of Steps 1 to 4 followed by this final JSON object.

---

### Input

QUESTION:  
```{question}```

COMMENT:  
```{comment}```  