Comparing the use of binary to ordinal logistic regression models in analyzing the twelve items general health questionnaire (GHQ-12) for complex survey data