Why TreeScan does not adjust for risk factors

Why TreeScan Does Not Adjust for Risk Factors

TreeScan Is a Surveillance Method, Not a Causal Model

TreeScan is designed to detect unexpected patterns in healthcare data, not to estimate causal relationships between risk factors and disease outcomes.

The method therefore does not adjust for variables such as:

• vaccination status
• age
• comorbidities
• demographic risk factors
• behavioral exposures

These factors are important for epidemiologic analysis, but they are not part of the statistical signal detection process.

Instead, TreeScan focuses on identifying unusual changes in diagnosis patterns over time.

---

What TreeScan Is Actually Testing

TreeScan tests whether the number of observed diagnoses in a specific diagnosis group and time window is unusually high compared with recent patterns in the data.

The null hypothesis can be thought of as:

Diagnoses occur over time in the same proportions that have been observed historically.

If the number of cases in a particular diagnosis group and time window deviates strongly from this baseline expectation, the method identifies a signal.

This comparison is based entirely on patterns within the surveillance data itself, rather than external risk factors.

---

Why Risk Factors Are Not Included

Including risk factors such as vaccination status would change the nature of the analysis.

TreeScan is designed to be:

• rapid
• automated
• agnostic to specific diseases

Introducing additional covariates would require detailed modeling assumptions about how those factors influence disease risk.

This would make the system:

• slower
• harder to automate
• dependent on disease-specific models

For real-time surveillance across many diagnoses, this approach would not be practical.

---

Where Risk Factors Become Important

Risk factors such as vaccination status can become important after a signal has been detected.

Once TreeScan identifies an unusual cluster, investigators may examine additional information such as:

• vaccination status
• demographic characteristics
• geographic patterns
• travel history
• environmental exposures

These analyses help determine whether the signal represents a true public health event and what factors may explain it.

---

Surveillance First, Explanation Later

The design philosophy of TreeScan can be summarized as:

• Detect unusual patterns quickly
• Investigate signals using additional information

By separating signal detection from causal investigation, TreeScan allows surveillance systems to monitor large volumes of healthcare data efficiently.