In today's item response theory (IRT) the response to the test item is considered as a probability event depending on the student's ability and difficulty of items. It is noted that in the scientific literature there is very little agreement about how to determine factors affecting the item difficulty. It is suggested that the difficulty of the item increases with the number of key elements, which are the data elements used in solving the item. Based on the statistical analysis of solutions of specially designed test items it has been concluded that there is a linear dependence of the item difficulty in the Rasch model on the number of key elements of the solution. The result allows taking an unbiased look at the difficulty of test items at the stage of their development, as well as predicts the outcome of testing.