Data analysis and interpretation

Statistical analysis was undertaken using SPSS 23.0. Analyzing the data, responses were reverse coded as appropriate. Simple descriptive statistics such as frequencies, means, and standard deviations (SD) were employed to report the socio-demographic characteristics, clinical characteristics, EORTC QLQ-C30, EORTC QLQ-BR23, EQ-5D-5 L, and EQ VAS scores. Multivariable logistic regression was carried out to identify possible predicting factors for GQoL. GQoL, symptom and functional scales have been dichotomized into “Affected at any degree” and “Not affected at all”. A score below 75 (above 75 mean no problem at all) for functional and GQoL scales were defined as “Affected at any degree”. Scores above a 25 mean (below 25 indicates no symptom at all) were defined as “Affected at any degree” and binary logistic regression was conducted between the GQoL and independent variables to obtain candidate variables for multi-variable logistic regression analysis. Variables with p-value < 0.25 were candidate for multiple regression analysis. Due to many independent variables, forward stepwise method was used for the multivariable analysis and significance of association was determined at p-value < 0.05. Patient’s utility score is obtained using possible (3125) health states of patients with breast cancer defined by the 5 dimensions and disutility coefficient of general population. Thus, it was calculated using the Ethiopian general population utility value set [17]. One caveat in order is the limitation within the analysis where causality of the associations was not confirmed.