Why diagnosis hierarchies matter

Why Diagnosis Hierarchies Matter

The Challenge of Monitoring Thousands of Diagnoses

Emergency department data contain thousands of individual ICD-10 diagnosis codes.

Monitoring each diagnosis individually would be impractical for surveillance because:

• many diagnoses occur very rarely
• some diagnoses represent closely related conditions
• emerging health events may appear across multiple related diagnoses

To address this, TreeScan organizes diagnoses into a hierarchical tree structure.

---

What the Diagnosis Tree Represents

In a diagnosis tree, related diagnoses are grouped into broader categories.

For example:

Respiratory disease
├── Upper respiratory infections \ │
├── Acute nasopharyngitis
│
└── Acute sinusitis \ └── Lower respiratory disease
├── Pneumonia
└── Bronchitis \

Each level of the tree represents a different level of diagnostic specificity.

---

How TreeScan Uses the Hierarchy

TreeScan evaluates clusters at multiple levels of the diagnosis tree simultaneously.

This means the method can detect clusters involving:

• very specific diagnoses (e.g., a particular ICD-10 code)
• broader diagnostic categories (e.g., respiratory illness)
• intermediate groups of related diagnoses

By scanning across the hierarchy, TreeScan can detect signals that might otherwise be missed if only individual diagnoses were monitored.

---

Why This Is Important for Surveillance

Emerging health events often affect groups of related diagnoses rather than a single diagnosis code.

For example:

• a respiratory outbreak may appear across multiple respiratory diagnoses
• gastrointestinal illness may appear across several related ICD-10 codes
• environmental exposures may generate symptoms coded in multiple ways

The hierarchical tree allows TreeScan to detect these broader patterns.

---

Balancing Specificity and Sensitivity

The diagnosis tree allows TreeScan to search across:

• specific nodes, which capture narrow diagnostic categories
• broader nodes, which capture related conditions

This balance helps the method detect both:

• small, specific clusters
• broader increases across related diagnoses

---

Why the Tree Must Be Defined Carefully

Because the diagnosis hierarchy determines how diagnoses are grouped, the structure of the tree influences the types of signals that can be detected.

For this project, the diagnosis tree is based on standard ICD-10 groupings used in syndromic surveillance systems.

Maintaining a consistent tree structure across jurisdictions helps ensure that signals detected in different locations are comparable.