In the project, I used health belief model (HBM) to investigate how its construct are associated with the utilization rate of a particular screening test.

HBM has 6 constructs, that are perceived barrier/ perceived benefit/ perceived severity/ perceived susceptibility/ self efficacy and cue to action.

in my questionnaire I have some questions under each construct, ranged from 4 to 16 questions.

For every question, I used a 5 point likert scale for choices, and they are Strongly disagree/disagree/don't know/agree/strongly agree

---------------------------------------

one of my hypothesis is that the 6 constructs are respectively directly associated with the outcome variable----ever have a particular test (yes/no for answer choices)

------------------------------------------

so I would like to ask, what method should I use to analyze the 5-point likert scale data?

For example, in construct A, there are 6 questions.

If I code the items "Strongly disagree=1/disagree=2/don't know=3/agree=4/strongly agree=5", I know how to calculate the mean and SD of 1 of the question, but are there any methods to calculate the mean and SD of the who construct A, that is, for all 6 questions?

Another question is, should I code "don't know" as score =3? It is different from "neutral" and I am wondering whether coding it as 3 will affect the meaning of the result.

Thanks!