TreeScan Overview

TreeScan Overview

What TreeScan Does

TreeScan is a statistical surveillance method used to detect unusual patterns in healthcare data.

In this project, TreeScan analyzes emergency department (ED) visit data and searches for clusters of diagnoses that occur more frequently than expected within a recent time window.

Rather than monitoring a small set of predefined syndromes, TreeScan evaluates all diagnosis groups simultaneously using a hierarchical tree of ICD-10 codes. This allows the method to detect unexpected patterns that may not fit into traditional surveillance categories.

TreeScan therefore supports syndrome-agnostic surveillance, meaning it can detect signals even when the underlying syndrome has not been predefined.

---

What TreeScan Is Looking For

TreeScan searches for combinations of:

• diagnosis groups
• time periods

where the number of observed cases is unusually high compared with the expected baseline.

For example, TreeScan might identify:

• a sudden increase in respiratory diagnoses over several days
• an unusual cluster of gastrointestinal diagnoses at a specific point in time
• a spike in a rare diagnosis category that is not normally monitored

These unusual patterns are called signals.

---

What a Signal Means

A TreeScan signal indicates that the observed number of cases in a particular diagnosis group and time window is unlikely under the recent baseline distribution of diagnoses.

Importantly, a signal does not mean that an outbreak or causal event has been confirmed.

Signals can arise for many reasons, including:

• real public health events
• changes in clinical coding practices
• data reporting artifacts
• random statistical fluctuation

TreeScan therefore functions as a screening tool that identifies patterns requiring further review.

---

Why TreeScan Is Useful for Syndromic Surveillance

Traditional syndromic surveillance systems typically monitor a fixed set of syndromes, such as respiratory illness or gastrointestinal illness.

While useful, this approach can miss unusual patterns that fall outside predefined categories.

TreeScan addresses this limitation by:

• scanning across all diagnosis codes
• evaluating many potential clusters simultaneously
• identifying patterns that may not have been anticipated

This makes TreeScan particularly useful for detecting novel or emerging health events.