1. Loss to follow-up correction increased mortality estimates in HIV-positive people on antiretroviral therapy in Mozambique
- Author
-
Nanina Anderegg, Matthias Egger, Laura F. Jefferys, Marloes H. Maathuis, Jonas Hector, Jochen Ehmer, Juan Burgos-Soto, Michael A Hobbins, and Lukas Meier
- Subjects
Adult ,Male ,Epidemiology ,Anti-HIV Agents ,Maximum likelihood ,Human immunodeficiency virus (HIV) ,HIV Infections ,medicine.disease_cause ,Competing risks ,Sensitivity and Specificity ,Article ,03 medical and health sciences ,Young Adult ,0302 clinical medicine ,Acquired immunodeficiency syndrome (AIDS) ,Risk Factors ,Medicine ,Humans ,Cumulative incidence ,030212 general & internal medicine ,Cd4 cell count ,610 Medicine & health ,Mozambique ,business.industry ,Reproducibility of Results ,medicine.disease ,Antiretroviral therapy ,Confidence interval ,Regression ,HIV-positive people ,Home visits ,Female ,Lost to Follow-Up ,business ,360 Social problems & social services ,030217 neurology & neurosurgery ,Demography ,Follow-Up Studies - Abstract
Objectives People living with human immunodeficiency virus (HIV) on antiretroviral therapy (ART) may be lost to follow-up (LTFU), which hampers the assessment of outcomes. We estimated mortality for patients starting ART in a rural region in sub-Saharan Africa and examined risk factors for death, correcting for LTFU. Study Design and Setting We analyzed data from Ancuabe, Mozambique, where patients LTFU are traced by phone and home visits. We used cumulative incidence functions to estimate mortality and LTFU. To correct for LTFU, we revised outcomes based on tracing data using different inverse probability weights (maximum likelihood, Ridge regression, or Bayesian model averaging). We fitted competing risk models to identify risk factors for death and LTFU. Results The analyses included 4,492 patients; during 8,152 person-years of follow-up, 486 patients died, 2,375 were LTFU, 752 were traced, and 603 were found. At 4 years after starting ART, observed mortality was 11.9% (95% confidence interval [CI]: 10.9–13.0), but 23.5% (95% CI: 19.8–28.0), 21.6% (95% CI: 18.7–25.0), and 23.3% (95% CI: 19.7–27.6) after correction with maximum likelihood, Ridge regression, and Bayesian model averaging weights, respectively. The risk factors for death included male sex, lower CD4 cell counts, and more advanced clinical stage. Conclusion In ART programs with substantial LTFU, mortality estimates need to take LTFU into account.
- Published
- 2020