TABLE 3.

Adjusted odds ratios for switching treatment among subjects with virologic failure, sampled from a cohort of HIV^*-infected persons in San Francisco, California, 2000–2004

Covariate	Adjusted odds ratio†
Current diagnosis with an opportunistic disease	1.21
No. of protease inhibitors tried (per drug)	1.11
Most recent HIV RNA level undetectable	0.44
% average adherence (per 10%)	0.92
Most recent CD4 T-cell count (per 100 cells/μl)	0.92
Nadir of CD4 T-cell count (per 100 cells/μl)	1.06
Most recent HIV RNA level >1 month prior	0.90
Age (per 5 years)	0.80

^*HIV, human immunodeficiency virus.

^†Based on a multivariable logistic regression model of the treatment mechanism, fitted data-adaptively using the deletion/substitution/addition algorithm (15) and fivefold cross-validation. Variables for the treatment mechanism were selected among a larger sample of nonsuppressed subjects in the SCOPE cohort (Study on the Consequences of the Protease Inhibitor Era): 255 persons and 368 episodes, including people with unknown loss of suppression time and missing outcome data. The coefficients in the selected model (corresponding to the odds ratios reported here) were then refitted in the population with known loss of suppression time. Standard errors and p values are not shown, to emphasize the role of the treatment mechanism in the construction of weights.