Data Validation During Entry

Validation during entry is one of the strongest advantages of electronic data capture. It allows errors to be detected near the point of origin, when study staff are more likely to remember the participant encounter and source documents are easier to consult. Validation rules can identify missing required values, invalid dates, out-of-range measurements, inconsistent responses, and incorrect formats. Validation should be designed before data collection begins. Numeric fields should have plausible ranges. Date fields should use standardized formats. Categorical fields should use coded options. Critical fields should be required where appropriate. Logic checks should compare related values, such as follow-up date occurring after enrollment date. Validation should be strong enough to prevent common errors but flexible enough to allow unusual true values with appropriate confirmation. Clinical plausibility is an important concept. A value may be unusual but real. For example, a hemoglobin value of 4 g/dL is alarming but possible in a severely ill participant. A validation warning should prompt verification, not necessarily prevent entry. Hard stops should be reserved for impossible values or values that violate essential structure, such as a negative age or a date entered in the wrong format. Validation also reduces unit errors. If a temperature field expects Celsius and uses a plausible range of 30 to 45, a Fahrenheit value such as 100 will trigger a warning. If weight is expected in kilograms, a value entered in pounds may appear implausible depending on the participant population. Field labels, units, completion guidelines, and validation rules should work together. Validation rules must be tested. A rule that appears correct during design may behave unexpectedly when fields are blank, when branching logic hides fields, or when data are collected across events. Testing should include valid values, invalid values, boundary values, missing values, and unusual but plausible values. Users should also be trained on how to respond to validation warnings. They should verify values against source documents rather than simply overriding warnings to move forward. **Table 4.4: Examples of Entry Validation Controls**