Causal vs Predictive Analysis: How DEPTh Typing Shapes Clinical Research Design

The analysis design stage forces a fundamental decision:

🧠 Are we explaining a cause, or predicting an event?

This logic determines how we build occurrence equations—and whether our statistical model needs confounder adjustment or predictive calibration.

---

🧠 Two Core Logics in Analysis Design

DEPTh Category	Occurrence Equation Template	Logic Type	Research Intent
Etiognostic	Y = f(X | confounders)	Causal	Explain: “Is X a cause of Y?”
Therapeutic	Y = f(Treatment | confounders)	Causal	Explain: “Does treatment lead to outcome?”
Diagnostic	Y = f(Signs, Symptoms, History, Labs)	Predictive	Predict disease status
Prognostic	Y = f(Biomarkers, Vitals, Clinical Data)	Predictive	Forecast future events

---

⚖️ Causal Occurrence Relations

Common in etiognostic and therapeutic studies. The goal is to test whether and how strongly X affects Y—while controlling for confounding and ensuring proper temporal order.

Examples:

• Etiognostic“Does heavy caffeine intake increase risk of first-trimester miscarriage?”

  
  

• 👉 เพื่อหาว่าการบริโภคคาเฟอีนมาก เป็น สาเหตุ ของการแท้งบุตรหรือไม่
  

• Therapeutic“Does early mobilization reduce length of hospital stay in elderly hip fracture patients?”

  
  

• 👉 เพื่อพิสูจน์ว่า intervention (early mobilization) มีผลต่อ outcome (length of stay)
  

---

🔮 Predictive Occurrence Relations

Found in diagnostic and prognostic research. Focused not on why, but on how well we can anticipate an outcome based on measurable variables.

Examples:

• Diagnostic: “Which symptoms and signs best predict acute appendicitis in children?”

  
  

• 👉 เพื่อประเมินว่า ข้อมูลใดช่วยในการวินิจฉัย โรคได้

• Prognostic“Which clinical variables predict ICU admission in COVID-19 patients?”

  
  

• 👉 เพื่อหาว่าตัวแปรใด พยากรณ์ การเข้า ICU ได้ดีที่สุด
  

---

🧩 Case Revisit: Benzodiazepine–Delirium

Let’s plug this back into a classic clinical puzzle:

• DEPTh Type: Etiognostic

• Logic: Causal

• Equation:

  
  

• 👉 จุดประสงค์: ต้องการทราบว่า ยาเบนโซ เป็น สาเหตุให้เกิด ภาวะเพ้อหรือไม่ (ไม่ใช่แค่คาดเดา)
  

---

🧠 Key Takeaways

• DEPTh typing clarifies whether your model should explain causality or predict outcomes.

• Causal designs (Etiognostic, Therapeutic) must address confounding and temporality.

• Predictive designs (Diagnostic, Prognostic) prioritize discrimination and calibration.

• Let your research question guide your equation logic, not the other way around.