Case Study 1. The purpose of Case study 1 was to evaluate the impact of adherence on lipid panel changes (25% reduction from baseline for LDL, TC, and non-HDL). Adherence to statin medication was based on the medication possession ratio (MPR), the main exposure variable in the multivariate regression model. Lipid panel changes were measured at 12 months after initiation of a statin prescription and included low-density lipoproteins (LDL), non-high-density lipoprotein (non- HDL), and triglycerides (TG). Adherence was the main exposure variable of interest and was categorized into adherent or non- adherent based on an arbitrary MPR threshold level of 0.80. Patients who were at or above the threshold were considered adherent; patients who were below the threshold were considered non- adherent. MPR is a surrogate marker for adherence and does not indicate that the patient ingested the medication; however, several studies have reported that it provides an accurate depiction of patient adherence using pharmacy claims data.63,64 MPR was calculated as the number of days supplied of prescription medication divided by days of observation.63,64 MPR was right-skewed; however it was assumed that this would not violate the assumptions of the multiple logistic regression. The dependent variable was change in lipid panel levels which included LDL, non-HDL, and TG. Although these outcomes were continuous, the study categorized them into achieving a 25% reduction from baseline which is considered a clinically significant change.65,66 This level of reduction has been associated with improved clinical outcomes.
Case Study 1. In Case study 1, there was a moderate amount of missing data that ranged from 16% to 19% for individual variables. In the multivariate models, the initial sample size of 7,739 subjects was reduced to 6,074 (~22% missing) due to missing data. Crude analysis was performed to provide a reference for comparison with the complete-case analysis and multiple imputation method. The conclusions from the complete-case analysis were similar to the crude analysis. Further comparisons between the conclusions of the multiple imputation method and complete-case analysis were also similar controlling for age, BMI, gender, baseline lipid values, comorbid conditions (diabetes, hypertension, congestive heart failure, history of myocardial infarction, angina, vascular disease), statin use, ethnicity, and starting medication count. In all scenarios, there were significant associations between adherence and achieving 25% reduction in lipid panel levels.
Case Study 1. In Case study 1, adherent patients were reported to have a higher odd of achieving 25% or more reduction in lipid panel levels relative to non-adherent patients controlling for potential confounding factors. Xxxxxxxxx, et al.36 reported similar findings in that patients who were adherence (MPR ≥ 0.80) had a significant reduction in LDL and non-HDL level relative to non- adherent patients in a veteran population. The use of multiple imputation did not change the conclusion of the complete-case analysis and crude analysis. However, it increased the sample size of the study and improved precision on the confidence limits when compared to the complete-case analysis. It also provided supporting evidence that the conclusions were appropriate and unaffected by missing data.
Case Study 1. Xxxxx, a 45 year old female, works as a primary school teacher. She has been divorced from her husband for six months, and lives with her three children, aged 5, 7, and 9. She has been dating a man in her neighborhood, and reports him as her only new sex partner. She presents to the health facility today because her husband was recently diagnosed with TB, and she was told to come for HIV testing. Her last HIV test was during ANC, when she was pregnant with her last child. It was negative. She has a history of high blood pressure that is well managed with medication. Her HIV test is positive.
