Test-enhanced web-based learning: optimizing the number of questions (a randomized crossover trial)

• Published in: Academic medicine Vol. 89; no. 1; p. 169
• Main Authors: Cook, David A, Thompson, Warren G, Thomas, Kris G
• Format: Journal Article
• Language: English
• Published: United States 01.01.2014
• Subjects: Adult; Computer-Assisted Instruction; Cross-Over Studies; Education, Medical, Graduate - methods; Educational Measurement - methods; Family Practice - education; Female; Humans; Internal Medicine - education; Internet; Internship and Residency; Male; Minnesota; Minnesota
• ISSN: 1938-808X
• DOI: 10.1097/ACM.0000000000000084

Questions enhance learning in Web-based courses, but preliminary evidence suggests that too many questions may interfere with learning. The authors sought to determine how varying the number of self-assessment questions affects knowledge outcomes in a Web-based course. The authors conducted a randomized crossover trial in one internal medicine and one family medicine residency program between January 2009 and July 2010. Eight Web-based modules on ambulatory medicine topics were developed, with varying numbers of self-assessment questions (0, 1, 5, 10, or 15). Participants completed modules in four different formats each year, with sequence randomly assigned. Participants completed a pretest for half their modules. Outcomes included knowledge, completion time, and module ratings. One hundred eighty residents provided data. The mean (standard error) percent correct knowledge score was 53.2 (0.8) for pretests and 73.7 (0.5) for posttests. In repeated-measures analysis pooling all data, mean posttest knowledge scores were highest for the 10- and 15-question formats (75.7 [1.1] and 74.4 [1.0], respectively) and lower for 0-, 1-, and 5-question formats (73.1 [1.3], 72.9 [1.0], and 72.8 [1.5], respectively); P = .04 for differences across all modules. Modules with more questions generally took longer to complete and were rated higher, although differences were small. Residents most often identified 10 questions as ideal. Posttest knowledge scores were higher for modules that included a pretest (75.4 [0.9] versus 72.2 [0.9]; P = .0002). Increasing the number of self-assessment questions improves learning until a plateau beyond which additional questions do not add value.