The average enrollment scores of the dependent variable HRQOL [the physical and mental component summaries (PCS and MCS) of the SF-36] will be compared among women at different stages of menopause using ANOVA. If the HRQOL scores are not normally distributed on the original scale, a transformation is necessary. We will examine the association between both the baseline presence and severity of symptoms (0-4) and HRQOL using regression techniques. The association between the use of HT (yes/no) and HRQOL will be analyzed by ANOVA. The association between the use of CAM and HRQOL will be analyzed by ANOVA. The association between attitudes towards menopause and HRQOL will be analyzed by linear regression. Confounders, including age, comorbid medical conditions, and social support, will be included in the models. A final model examining the impact of all factors on HRQOL will be created using stepwise linear regression.

We will record all methods used by women in the cohort to manage menopause. Methods will be categorized as use of: no method, HT, and CAM. A frequency table will quantify use of each method as well as type of CAM. We will analyze the relationship between both menopausal symptoms and attitudes towards menopause and methods used by Chi-square or contingency table tests.

Using regression analysis techniques, we will examine the relationship between HSU and our independent variables, HRQOL and menopausal stage at assessment. The model will also adjust for confounders such as age and the number and type of comorbidities at the time of analysis. Because all women in the study cohort will have utilization data for at least one physician visit, we will use a linear regression model. If cost data are not normally distributed, we will apply appropriate transformations.

We will calculate the concordance correlation coefficient to investigate correlation between HSU data extracted from MARS and that obtained by direct patient interview to uncover under-reporting in the MARS database.