Incidence rate monitoring in DCE fieldwork

Incidence rate should be stable throughout fieldwork. A rising IR is the clearest early warning sign of organised panel fraud.

This article explains incidence rate monitoring as a fraud detection tool in DCE fieldwork, how to track IR in SurveyEngine, and what to do when IR shows an anomalous pattern.

Ben White, 07.07.2026

What is incidence rate and why does it matter?

Incidence rate (IR) is the proportion of recruited respondents who pass the screener and qualify to complete the main survey. If your study targets patients with a specific diagnosis and 30% of panel respondents claim that diagnosis and pass the screener, your IR is 30%.

In a legitimate study with genuine panel respondents, IR should be approximately constant throughout fieldwork. Each new respondent's eligibility is an independent event - whether they qualify should not depend on who completed before them. A monotonically increasing IR over time therefore indicates that early respondents are passing information about the screener answers to later respondents - a direct sign of organised panel fraud.

Why rising IR is a fraud signal

In organised panel fraud, small groups of fraudsters coordinate to pass screening questions in online surveys. Early fraudsters may fail but provide information to later fraudsters about what answers pass the screener. As this knowledge spreads, the proportion passing increases monotonically.

The fraud detection toolkit (Iskiwitch, White et al., ISPOR 2024) identifies monotonically increasing IR as one of the most reliable indicators of coordinated fraud. The method requires no additional survey questions - it uses only the timing data that is automatically collected in SurveyEngine's fieldwork dashboard.

IR monitoring is also valuable for legitimate reasons - to identify screener questions that are not discriminating as expected, or to detect changes in panel composition over the course of fieldwork that may affect sample representativeness.

TLDR Quick links

Fraud detection toolkit

Forum: data quality

Open a free account

Monitoring incidence rates in SurveyEngine

Step 1: Set up a rolling IR calculation. In SurveyEngine's fieldwork dashboard, the IR is calculated automatically from the ratio of completions to contacts. Monitor this metric at least twice daily during active fieldwork.

Step 2: Calculate a rolling average over a meaningful window. Point-in-time IR is noisy. Calculate a rolling average over the previous 50–100 contacts to smooth out random variation while detecting genuine trends.

Step 3: Set an alert threshold. Define an alert threshold - for example, IR increasing by more than 10 percentage points from the initial baseline over a 3-day rolling window. Set up an email or dashboard alert when this threshold is exceeded.

Step 4: Investigate anomalies promptly. When an IR anomaly is detected, pause fieldwork and investigate. Check whether the anomaly is confined to a specific panel source, time period, or demographic sub-group. Contact the panel provider for an explanation.

Step 5: Document the monitoring in your study report. Report the IR trajectory over the course of fieldwork in your methods section. If an anomaly was detected and investigated, describe the actions taken and any data excluded as a result.

Worked example - IR monitoring in a health preference study

A patient preference study targeting patients with Type 2 diabetes sets an expected IR of 35% (prevalence-based estimate). During the first two days of fieldwork, IR is 33%. By day 4, the rolling IR has increased to 51%, and by day 6, to 68%.

SurveyEngine's dashboard alert fires at the 45% threshold. Fieldwork is paused and the panel source is investigated. The panel provider confirms that a fraudulent sub-group has been sharing screener answers. 94 responses from the affected period are excluded. Fieldwork restarts with an enhanced screener - adding a disease-specific question that genuine patients can answer but fraudsters cannot. IR stabilises at 38% for the remainder of fieldwork.