ANÁLISIS DEL SUELO A BIODEGRADAR

Descriptive statistics were used to illustrate sample characteristics and differences between those with depression and those without depression. These statistics were also used to characterize antidepressant users and non- users in the depressed population. Chi-square statistics were used to compare categorical variables across groups and the two-sample t-test was used to compare continuous variables across groups. For all regression models, an a priori selection of variables based off the conceptual framework were used. To test the hypothesis that those with antidepressant use will have improved adherence to hormone therapy in the depressed population, a generalized linear regression with logit-link and binomial distribution for repeated measures was used. This analysis is similar to logistic regression and does yield odds ratios from the parameter estimates. A generalized linear regression with log-link and gamma distribution for repeated measures of total cost was used to test the hypothesis that those on antidepressants will incur lower cost in the depressed population and to determine the incremental cost of depression. Kaplan Meier estimates were used to determine the initial association of depression and

antidepressant use in the depressed population with persistence and survival. Kaplan-Meier estimates and a time interaction variable in the cox proportional hazards model were used to test the proportional hazards assumption. If proportional hazards were indicated (curves did not cross

and the interaction variable was not statistically significant), the association of depression and antidepressant use in the depressed population was reported from the adjusted Cox proportional hazards model. If proportional hazards were not indicated, (curves did cross and the time interaction variable was statistically significant), the distribution was determined using a linear survival model that fits different distributions. The distribution with the best fit (lowest AIC value) was used to estimate the association of depression and antidepressant use in the depressed population, adjusting for clinical and demographic characteristics.

The following model is the general form used for the analysis for the association of depression with adherence to hormone therapy. Y indicates the probability of adhering to hormone therapy so e^β is the adjusted odds ratio for the parameter110, 111_{. For this study, e^ β1 is} the estimate of interest as it indicates the odds of adhering to hormone therapy if a person has depression adjusting for demographic and clinical characteristics110_.

Log (Y/1-Y) = β1 *depression + β2 * age + β3 *race + β4 *co-morbidity + β5 *SEER site + β6 *urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression + β12 *count indicating repeated measures

The following model is the general form used for the analysis for the association of depression with survival. Hi represents an individual’s hazard of death and t indicates length of time a person survives. H0 is the baseline hazard function. No assumption is needed for the baseline hazard function when there are proportional hazards between groups and H0 is not in the model. If proportional hazards are not indicated, then H0 is in the model as a baseline hazard function for the sample. For this study, e^(β1) is the estimate of interest as it indicates the risk of death if a person has depression adjusting for demographic and clinical characteristics112_.

*urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression) )

The following model is the general form used for the analysis for the association of depression with cost. Y represents total cost per patient per year and Log (E(Y|variable)) is the log cost based off the log link used for this association. For this study, β1 is the estimate of interest as it indicates the increase in cost a person has depression adjusting for demographic and clinical characteristics113_{. The intercept in this case is baseline cost for the sample and estimates} are added to the intercept for the total cost94_.

Log (E(Y|variable)) = intercept+ β1 *depression + β2 * age + β3 *race + β4 *co-morbidity + β5 *SEER site + β6 *urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression + β12 *count indicating repeated measures

The following model is the general form used for the analysis for the association of antidepressants with adherence to hormone therapy in the depressed population. Y indicates the probability of adhering to hormone therapy so e^β is the adjusted odds ratio for the parameter110, 111_{. For this study, e^ (β1 is the estimate of interest as it indicates the odds of adhering to hormone}

therapy if a person has depression adjusting for demographic and clinical characteristics.

Log(Y/1-Y) = β1 *antidepressants + β2 * age + β3 *race + β4 *co-morbidity + β5 *SEER site + β6 *urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression + β12 * number of 30 day antidepressant supplies + β13 * count indicating repeated measures

The following model is the general form used for the analysis to determine the association of antidepressants with survival in the depressed population. Hi represents an individual’s hazard of death and t indicates length of time a person survives. H0 is the baseline hazard function. No assumption is needed for the baseline hazard function when there are proportional hazards between groups and H0 is not in the model. If proportional hazards are not indicated, then H0 is in the model as a baseline hazard function for the sample. For this study, e^(β1) is the estimate of interest as it indicates the risk of death if a person with depression uses antidepressant adjusting for demographic and clinical characteristics112_.

Hi = H0 * e^( β1 *antidepressants + β2 * age + β3 *race + β4 *co-morbidity + β5 *SEER site + β6 *urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression + β12 * number of 30 day antidepressant supplies)

The following model is the general form used for the analysis to determine the

association of antidepressants with cost in the depressed population. Y represents cost and Log (E(Y|variable)) is the log cost based off the log link used for this association. For this study, β1 is the estimate of interest as it indicates the increase in cost a person with depression uses

antidepressants adjusting for demographic and clinical characteristics113_{. The intercept in this} case is baseline cost for the sample and estimates are added to the intercept for the total cost94_.

Log (E(Y|variable)) = intercept+ β1 *antidepressants + β2 * age + β3 *race + β4 *co-morbidity + β5 *SEER site + β6 *urban + β7 *cancer stage + β8 *chemotherapy + β9 *radiation+ β10 *cancer grade + β11 *history of depression + β12 * number of 30 day antidepressant supplies + + β13 * count indicating repeated measures

CHAPTER 4 RESULTS