Content Validity Index (CVI)

The Content Validity Index (CVI) is a quantitative measure of the extent to which the items on a measurement instrument are relevant to and representative of the construct they are intended to measure. It is calculated based on expert ratings of each item's relevance, typically on a four point scale (1 = not relevant, 2 = somewhat relevant, 3 = quite relevant, 4 = highly relevant). The CVI can be calculated at the item level (I CVI), representing the proportion of experts who rate an item as relevant (3 or 4), and at the scale level (S CVI), representing the average of the I CVIs across all items or the proportion of items rated as relevant by all experts.^[3]

Content validity is the foundation of any measurement instrument: if an instrument's items do not adequately represent the construct of interest, no amount of statistical validation can compensate. The CVI summarizes experts' judgments of relevance in a transparent, numerical form; it quantifies those judgments rather than replacing them.^[4] The CVI is widely reported in studies that develop or adapt questionnaires, scales, or assessment tools, especially in nursing and health research.^[3][4] A high CVI strengthens the argument that the instrument measures what it claims to measure and that subsequent validity testing (construct, criterion) is worth pursuing.

Researchers assemble a panel of experts (typically five to ten individuals with relevant content expertise) and ask them to rate each item on the instrument for its relevance to the construct being measured. Using the four point rating scale, items rated 3 or 4 are considered relevant. The I CVI for each item equals the number of experts rating the item as relevant divided by the total number of experts. Items with an I CVI below 0.78 (for panels of six or more experts) are typically revised or removed. The S CVI can be calculated as the average of all I CVIs (S CVI/Ave) or as the proportion of items rated relevant (3 or 4) by all experts (S CVI/UA, or universal agreement).^[3] An S CVI/Ave of 0.90 or higher is generally considered acceptable.

Calculate the CVI during the development of a new survey, questionnaire, or assessment instrument that is meant to measure a defined construct. It should also be computed when adapting an existing instrument for a new population, language, or cultural context, because items that were relevant in the original setting may not be relevant in the new one. The CVI is a common step in instrument development that precedes pilot testing and larger scale psychometric analyses such as exploratory or confirmatory factor analysis.

Selecting experts who lack genuine expertise in the content area produces unreliable CVI values. Another common error is using too few experts (fewer than three), which inflates chance agreement and makes the CVI unstable. Researchers sometimes report only the S CVI without examining I CVIs for individual items, potentially retaining problematic items that are masked by overall high averages. Failing to account for chance agreement is another issue; the modified kappa statistic can adjust I CVI values for chance.^[4] Finally, some researchers conflate content validity with face validity; face validity is a superficial assessment of whether an item "looks right," whereas content validity requires systematic expert evaluation against a defined construct.

A team of nursing researchers develops a 30 item instrument to measure patient engagement in self management of diabetes. They recruit eight endocrinologists, diabetes educators, and behavioral scientists to serve as the expert panel. Each expert rates every item on the four point relevance scale. After calculating I CVIs, four items fall below the 0.78 threshold and are removed. Two additional items are revised based on expert comments and re evaluated in a second round, where both achieve I CVIs above 0.85. The final 26 item instrument has an S CVI/Ave of 0.93, supporting its content validity before proceeding to pilot testing with patients.