Analysis 1. Given that you have information on the health care expenses for two cohorts of patients (one cohort having a disease under study and the other cohort free from disease). a. Ignoring cohort differences in baseline patient characteristics, how would you assess whether the average cost across patients in the diseased cohort is significantly greater than that across disease-free patients? Describe tests or methods that might be used. b. How would you change the analysis if you wanted to account for the differences in characteristics between m cohorts at baseline? 2. Consider the following regression output to predict a person's annual health care costs (t-statistics are given under parameter estimates): cosr = 5000 + 200*AGE - 300*SEX + 700*HEARTCDX 100*EXRCISE (2.7) 44.2) (1.2) (3.2) (1-7) R2=0.45 AGE is measured in years SEX is 1 for male, 0 for female HEARTCDX is 1 if person has a heart condition, 0 if not EXRCISE is 1 if person regularly exercises, 0 if not 2. Consider the following regression output to predict a person's annual health care costs (t-statistics are given under parameter estimates): COST = 5000 + 200*AGE - 300*SEX + 700*HEARTCDX - 100*EXCISE (2.7) (4.2) (1.2) (3.2) (1.7) R2=0.45 AGE is measured in years SEX is 1 for male, 0 for female HEARTCDX is 1 if person has a heart condition, O if not EXRCISE is 1 if person regularly exercises, 0 if not a. Give a general interpretation of the regression equation, describing what the parameter estimates and t-statistics mean. Which variables can be considered predictors of health care costs? b. What are the expected annual health care costs for a 40-year-old female with a heart condition who exercises regularly? 3. Given the equation TC = HDL + LDL + TG*.20, solve for LDL given HDL=40 and TC=240.4. You are taking over a project started by another analyst on your team using medical claims data. The project studies a cohort of 300 schizophrenia patients newly taking oral antipsychotics. The primary outcome of interest is switch to a long acting injectable agent. You define a switch as the presence of a long acting injectable agent while at risk in the follow up (after first taking an oral antipsychotic). One of the limitations of using claims data is that you will not observe reliable information of medications received while patients are not enrolled with pharmacy benefits. The diagram below describes the study cohort inclusion/exclusion criteria implemented by the original analyst and shows observed events occurring in the follow up. It also shows a sample of 4 of the 300 patients. Legend Oral Antipsychotlc Medication (0A) 0 Long Acting Injectable Antlpsychotlc Medication D End of Heath Plan Pharmacy Benets Enrollment x Study Diagram lnclusiun' Baseline Enrollment in Health Plan with Medical and Pharmacy Benefits f% 350 Day Baseline Study Period Start Index: First DA Study Period End lOl Study Day: 1 60 120 130 240 300 360 mm 000 El X Purim: o O D x ...... o o o x El Patient 0 x a. Describe why the original analyst may have required a. Describe why the original analyst may have required enrollment in the 360 day baseline and excluded patients with Major Depressive Disorder or Bipolar Disorder. b. Calculate the observed outcome rate per 100 days at risk for the 4 patients in the diagram and describe the logic behind your calculation. c. Describe a method that can be used to calculate the probability a randomly selected patient in your cohort will have switched by 240 days