Multivariate logistic regression analysis was performed on the measures of adherence using selected explanatory variables including age, gender, occurrence of side-effects, education, occupation, cost of transport, other treatment-related costs, loss of income, months on treatment, missed appointments, and knowledge of HIV and ART (Annex 2). The variable of missed appointments was a significant predictor of adherence (Wald statistic χ²=4.851; p=0.028) and was associated with levels of adherence to ARV treatment (χ²=5.86;P=0.016). This finding was also reported by Nemes et al. (2004).

Independent t-tests between the sub-optimal and optimal groups for age, distance to facility, cost of transport and months on treatment indicated no significant differences (P>0.05) in both one-month recall methods (Table 4.5).

Table 4.5: Independent t-tests for comparisons between optimal and sub-optimal adherence groups