Skip to main content

Table 3 Variables, Measures and Methods of Analysis

From: The HAART cell phone adherence trial (WelTel Kenya1): a randomized controlled trial protocol

Variable/Outcome Hypothesis Outcome Measure Methods of Analysis
1) Primary Intervention improved outcome from baseline to 6 months   
a) Adherence at 12 months   Percent adherence in previous 30 days >95% [binary] Chi-squared test
b) Suppression of HIV viral load at 12 months   Viral load ≤400 copies/ml [binary] Chi-squared test
2) Secondary
Adherence percentage at 12 months
improvement occurred Adherence % (>95%) [binary] Chi-squared test
HIV viral load at 12 months improvement occurred Viral load (copies) T-test
Immune reconstitution (change in CD4 T cell count from baseline) improvement occurred Cd4 T-cells/mm3 (continuous) T-test
Time to virological failure Improvement occurred Virological failure after successful suppression Kaplan-Meier survival analysis
Weight gain [lbs] and BMI improvement occurred Change in weight (lbs) and BMI [continuous] T-test
Occurrence of opportunistic infections (OIs) improvement occurred Presence of AIDS defining opportunistic infection [binary] Chi-squared test
Time to reporting of adverse drug events (ADEs) improvement occurred Presence of drug-related adverse event [time to event] Kaplan-Meier survival analysis
Deaths (all cause) improvement occurred All-cause mortality [binary] Chi-squared test and Kaplan-Meier survival analysis
SF-12 improvement occurred Quality pf life questionnaire [continuous] T-test
Satisfaction with care provided improvement occurred Questionnaire [continuous] T-test
Level of disclosure of HIV status improvement occurred Disclosed to a family member [binary] Chi-squared test
Impression of stigma improvement occurred Questionnaire [continuous] T-test
Family dyamics improvement occurred Questionnaire [continuous] T-test
Employment attendance improvement occurred Questionnaire [continuous] T-test
Household member school attendance improvement occurred Questionnaire [continuous] T-test
Cell phones lost/stolen improvement occurred Presence of cellphone [binary] Poisson regression
Stopped taking HAART improvement occurred Self-report [binary] Chi-squared test
Required active tracing for 12 month follow-up improvement occurred Field officers [binary] Chi-squared test
3) Subgroup Analyses:    Regression methods with appropriate interaction term
   Urban vs. rural Distance affects adherence   
Female vs. male Sex affects adherence   
Phone ownership (owned vs. shared) Ownership affects adherence   
Level of education Low education affects adherence   
4) Sensitivity Analyses: improvement occurred All outcomes  
   a) Per protocol analysis       a) Chi-squared/T-test test
   b) Adjusting for baseline covariates       b) Multivariable regression
   c) clustering among individuals within a clinic    c) GEE
  2. • The GEE[11] is a technique that allows to specify the correlation structure between patients within a hospital and this approach produces unbiased estimates under the assumption that missing observations will be missing at random. An amended approach of weighted GEE will be employed if missingness is found not to be at random.[12]
  3. • In all analyses results will be expressed as coefficient, standard errors, corresponding 95% and associated p-values.
  4. • Goodness-of-fit will be assessed by examining the residuals for model assumptions and chi-squared test of goodness-of-fit.
  5. Bonferroni method will be used to adjust the overall level of significance for multiple secondary outcomes.