1. Sampling-Based Approach to Determining Outcomes of
Patients Lost to Follow-Up in Antiretroviral Therapy
Scale-Up Programs in Africa
Online article and related content
current as of August 3, 2008. Elvin H. Geng; Nneka Emenyonu; Mwebesa Bosco Bwana; et al.
JAMA. 2008;300(5):506-507 (doi:10.1001/jama.300.5.506)
http://jama.ama-assn.org/cgi/content/full/300/5/506
Correction Contact me if this article is corrected.
Citations Contact me when this article is cited.
Topic collections Drug Therapy, Other; World Health; Infectious Diseases; HIV/AIDS; Prognosis/
Outcomes; Drug Therapy; Public Health
Contact me when new articles are published in these topic areas.
Subscribe Email Alerts
http://jama.com/subscribe http://jamaarchives.com/alerts
Permissions Reprints/E-prints
permissions@ama-assn.org reprints@ama-assn.org
http://pubs.ama-assn.org/misc/permissions.dtl
Downloaded from www.jama.com by guest on August 3, 2008

2. LETTERS
1. Harbarth S, Fankhauser C, Schrenzel J, et al. Universal screening for methicillin- Mbarara University of Science and Technology. Patients gave
resistant Staphylococcus aureus at hospital admission and nosocomial infection in
surgical patients. JAMA. 2008;299(10):1149-1157. oral consent.
2. Robicsek A, Beaumont JL, Paule SM, et al. Universal surveillance for methicillin- Results. A total of 3628 HIV-infected adults newly starting
resistant Staphylococcus aureus in 3 affiliated hospitals. Ann Intern Med. 2008;
148(6):409-418.
ART were evaluated. The median age was 35 years, 61% were
3. Cooper BS, Stone SP, Kibbler CC, et al. Isolation measures in the hospital man- women, and median pretherapy CD4 T-cell count was 95/mm3.
agement of methicillin resistant Staphylococcus aureus (MRSA): systematic re- Over a maximum of 3.75 years (median, 1.16 years; interquar-
view of the literature. BMJ. 2004;329(7465):533.
4. McGinigle KL, Gourlay ML, Buchanan IB. The use of active surveillance cul- tile range, 0.42-2.11 years), 829 patients became lost to follow-
tures in adult intensive care units to reduce methicillin-resistant Staphylococcus up, of whom a sample of 128 was sought. Of these 128, we could
aureus-related morbidity, mortality, and costs: a systematic review. Clin Infect Dis.
2008;46(11):1717-1725. ascertain the vital status of 111 patients (87%); 32 of 111 (29%)
5. Muto CA, Jernigan JA, Ostrowsky BE, et al. SHEA guideline for preventing noso- had died. Patients found by tracking (n=111) were similar in
comial transmission of multidrug-resistant strains of Staphylococcus aureus and
Enterococcus. Infect Control Hosp Epidemiol. 2003;24(5):362-386. age(P=.66),sex(P=.84),andCD4T-cellcount(P=.44)tothose
who were lost and did not have vital status ascertained (n=718).
Using the naive estimate, the cumulative incidence of death at
1, 2, and 3 years following ART was 1.7%, 2.1%, and 2.3%, re-
RESEARCH LETTER spectively. After incorporating deaths from the tracked sample
of lost patients (probability weight, 829/111), the corrected in-
Sampling-Based Approach to Determining
cidence of death at 1, 2, and 3 years was 7.5%, 10.3%, and 12.2%,
Outcomes of Patients Lost to Follow-Up
respectively (FIGURE 2).
in Antiretroviral Therapy Scale-Up
Comment. To address a critical barrier in evaluating ART
Programs in Africa
scale-up, we used a sampling-based strategy to account for
To the Editor: Evaluating outcomes among the millions of losses to follow-up. Searching for a sample of lost patients and
HIV-infected patients starting antiretroviral therapy (ART) determining vital status on a high percentage was feasible.
in settings with limited resources is key to understanding Using just a single tracker to locate lost patients and simple
the effect of current treatment programs and guiding fu- calculations to derive corrected survival estimates, the pro-
ture strategies. Accurately assessing survival outcomes has
been precluded by substantial numbers of patients not re-
turning for care. One year after starting ART, 15% to 30% Figure 1. Derivation of Probability Weights
of patients are lost to follow-up.1,2 Only by determining the (A)
outcomes of those lost can true survival and program im- Entire clinic population
pact be understood. We present a sampling-based strategy
to address this.
Methods. We evaluated all HIV-infected adults initiating
ART in a rural clinic in Mbarara, Uganda, between January
1, 2004, and September 30, 2007. Each month, a tracker
(B) (C) (D)
sought an unselected and consecutive sample of patients in Patients Patients Patients with vital
the community who were lost to follow-up (a 6-month ab- lost to sought status ascertained
sence from clinic) to obtain their vital status. Naive and cor- follow-up by tracking by tracking
rected estimates of cumulative incidence of mortality were
determined with Kaplan-Meier techniques. In the naive es-
timate, only deaths passively recorded by the clinic through
routine processes were included. In the corrected estimate, Continued in care or
the updated vital status among the tracked sample of lost died while in care
patients was used to represent outcomes among all those Patients lost
lost to follow-up by generating a probability weight (the ra- Pw =
to follow-up (B)
tio of all patients lost to follow-up to those lost and sampled Patients with vital
status ascertained
with subsequent updated vital status) (FIGURE 1). Lost pa- by tracking (D)
tients with sampling-updated vital status were assigned this
weight, and all other lost patients were dropped from the
analysis. This approach is equivalent to previously de-
The probability weight (Pw) allows a sample of patients who were lost to follow-up
scribed methods.3 Comparisons of lost patients with and to represent all lost patients in subsequent survival analysis. Population A repre-
without vital status ascertainment were tested with 2 and sents the entire clinic population, B represents all patients lost to follow-up, C is the
t tests. Significance was set at P .05. Analyses were per- sample of lost patients sought in the community, and D represents the sought-after
lost patients who had their vital status ascertained. Vital status outcomes among
formed using Stata 10 (StataCorp, College Station, Texas). those successfully sought (group D) are taken to represent outcomes among the
The study was approved by the institutional review boards remaining lost patients (group B–group D). Hence, the ratio of B/D is used to weight
patients in group D to allow them to represent all lost patients.
of the University of California, San Francisco, and the
506 JAMA, August 6, 2008—Vol 300, No. 5 (Reprinted) ©2008 American Medical Association. All rights reserved.
Downloaded from www.jama.com by guest on August 3, 2008

3. LETTERS
Nneka Emenyonu, MPH
Figure 2. Naive and Corrected Mortality Estimates Epidemiology Prevention and Intervention Center
0.20
Department of Medicine
Corrected estimate University of California, San Francisco
Naive estimate
Mwebesa Bosco Bwana, MBChB, MMed
Mbarara University of Science and Technology
0.15
Mbarara, Uganda
Proportion Deceased
David V. Glidden, PhD
Jeffrey N. Martin, MD, MPH
0.10
Department of Epidemiology and Biostatistics
University of California, San Francisco
0.05
Author Contributions: Dr Geng had full access to all of the data in the study
and takes responsibility for the integrity of the data and the accuracy of the
data analysis.
Study concept and design: Geng, Emenyonu, Glidden, Martin.
Acquisition of data: Geng, Emenyonu, Bwana, Glidden, Martin.
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Analysis and interpretation of data: Geng, Glidden, Martin.
Time Since Initiation of Antiretroviral Therapy, y Drafting of the manuscript: Geng, Glidden, Martin.
Critical revision of the manuscript for important intellectual content: Geng,
No. at risk 3628 2489 1897 1399 920 485 197 44
Emenyonu, Bwana, Glidden, Martin.
Statistical analysis: Geng, Glidden, Martin.
The naive estimate is determined using those deaths passively reported during the Obtained funding: Martin.
course of routine clinical care. The corrected estimate uses vital status outcomes in Administrative, technical, or material support: Emenyonu, Martin.
the tracked sample of lost patients to represent vital status outcomes in all lost patients. Study supervision: Emenyonu, Bwana, Martin.
Financial Disclosures: None reported.
Funding/Support: This study was funded by grants U01 AI069911, T32
cess is accessible and affordable. Incorporating updated out- AI06530, and R01 MH054907 from the National Institutes of Health, the East
Africa International Epidemiologic Databases to Evaluate AIDS (IeDEA) Consor-
comes from the tracked sample increased 5.3-fold the esti- tium, the United States President’s Emergency Plan for AIDS Relief (PEPFAR),
mate of mortality at 3 years, an absolute difference of 9.9%. and the Antiretroviral Treatment in Lower Income Countries (ARTLINC) Col-
laboration.
Although our sample was unselected and consecutive, it Role of the Sponsors: The sponsors had no role in the design and conduct of the
was not random; hence, the corrected mortality estimates study; in the collection, management, analysis, and interpretation of the data; or
in the preparation, review, or approval of the manuscript.
may potentially underestimate or overestimate true mortal- Previous Presentation: Presented in part at the 15th Conference on Opportunis-
ity. However, the lack of substantial differences in charac- tic Infections and Retroviruses; February 3-6, 2008; Boston, Massachusetts.
teristics between the lost patients with and without tracking- Additional Contributions: David Bangsberg, MD, University of California, San
Francisco (UCSF), assisted with the study concept; Hassan Baryahikwa, Uganda
updated vital status ascertainment suggests the sample was Research Institute (URI), tracked patients; John Bennett, MPH, UCSF, provided
representative and unbiased. Formally sampling and track- data management; Steven Deeks, MD, UCSF, assisted with the study concept;
Este Hudes, PhD, UCSF, provided statistical consultation; Nicolas Musinguzi,
ing a random subset of lost patients should strengthen this BA, URI, provided data management; Torsten Neilands, PhD, UCSF, provided
strategy. Sampling is well suited to the resource limitations statistical consultation; Adam Pepper, URI, provided data management; Larry
Pepper, MD, Mbarara University of Science and Technology, assisted with the
of global scale-up because the fraction of lost patients sought study concept; and Constantin Yiannoutsos, PhD, Indiana University, assisted
can be balanced against cost constraints. This approach to with the study concept. All persons except Dr Hudes and Dr Pepper received
compensation for their roles in this study.
losses to follow-up may provide value as a routine aspect of
understanding survival in program evaluation. 1. Rosen S, Fox MP, Gill CJ. Patient retention in antiretroviral therapy programs
in sub-Saharan Africa: a systematic review. PLoS Med. 2007;4(10):e298.
Elvin H. Geng, MD, MPH 2. Braitstein P, Brinkhof MW, Dabis F, et al. Mortality of HIV-1-infected patients
genge@php.ucsf.edu in the first year of antiretroviral therapy: comparison between low-income and
high-income countries. Lancet. 2006;367(9513):817-824.
Positive Health Program of the Department of Medicine 3. Frangakis CE, Rubin DB. Addressing an idiosyncrasy in estimating survival curves
San Francisco General Hospital using double sampling in the presence of self-selected right censoring. Biometrics.
San Francisco, California 2001;57(2):333-342.
©2008 American Medical Association. All rights reserved. (Reprinted) JAMA, August 6, 2008—Vol 300, No. 5 507
Downloaded from www.jama.com by guest on August 3, 2008