In a mailed physician survey, questionnaire length had a threshold effect on response rate

To examine the association between questionnaire length and response rate in a mailed survey of generalist physicians randomly selected from the American Medical Association master file.

In a pilot study, otherwise similar questionnaires of 30 different lengths (849 to 1,867 words) were mailed to 192 physicians in April 1999. In the main study, questionnaires of 16 different lengths (564 to 988 words) were mailed to 1,700 physicians between June 1999 and January 2000.

In the pilot study, response rate decreased from 60% for questionnaires 849 words in length to 16.7% for questionnaires over 1,800 words in length. Logistic regression revealed an odds ratio of 0.887 (95%CI 0.813, 0.968; p=0.006) for word count, expressed in units of 100 words. In the main study, response rate varied between 51.5% and 71.4%. Logistic regression showed no association between response and word count (OR 0.988; 95%CI 0.896, 1.090; p=0.81).

There appears to have been a threshold in these studies of approximately 1,000 words. Questionnaires above the threshold had lower response rates than those below it (38.0% vs. 59.4%).