Original Article
Predictors identified for losses to follow-up among HIV-seropositive patients

https://doi.org/10.1016/j.jclinepi.2005.11.024Get rights and content

Abstract

Background and Objective

This study aims to describe predictors of loss to follow-up (LFU) in the French Hospital Database on HIV infection (FHDH).

Methods

We studied the prevalence and predictors of LFU among 34,835 patients enrolled in FHDH in 1999. Impacts of demographic and clinical factors were studied by using multivariate logistic regression analysis.

Results

Among included patients, 1,478 (4.2%) died and 2,950 (8.5%) were lost to follow-up. LFU was more frequent among recently diagnosed patients (<1 year, 16.8%; > 1 year, 7.1). Among recently diagnosed patients, LFU was less frequent among men who have sex with men (MSM) [odds radio (OR) = 0.6, 95% confidence interval (CI) = (0.5;0.7)], and among patients with AIDS [OR = 0.5, 95%CI = (0.4;0.6)], and more frequent among immigrants [OR = 1.3, 95%CI = (1.0;1.5)]. Among less recently diagnosed patients, LFU was more frequent in French Departments of America than in the Paris area. The proportion of LFU fell with age, and LFU was more frequent among intravenous drug users (IVDU) than among MSM [OR = 1.2, 95%CI = (1.1;1.4)]. Patients with viral load >5,000 copies/mL or CD4 cell counts <200/mm3 were more likely to be lost to follow-up.

Conclusions

Recently diagnosed patients, IVDU, and immigrants are more often lost to follow-up, and should therefore receive special attention.

Introduction

Cohort studies are a major HIV/AIDS research tool. In the epidemiologic setting they are particularly useful for determining the incidence of rare diseases [1], [2]. In the era of highly active antiretroviral therapy (HAART), cohort studies have proven useful for measuring the impact of drugs and to evaluate prognosis factors in the HAART era [3]. However, the results of such studies may be skewed by losses to follow-up and missing outcome data.

To our knowledge, patients lost to follow-up (LFU) in HIV cohort studies have never been thoroughly characterized. Although LFU are systematically recorded in clinical trials, this is not generally the case in cohort databases. Better knowledge of such patients could help to improve the management of HIV disease. From a methodologic point of view, it would also be useful to know the frequency and predictors of LFU to evaluate the potential bias due to attrition [4], to build tools for tracing patients [4], [5], and to impute missing outcomes [6].

In the HAART era the epidemiology of HIV/AIDS has changed in the northern hemisphere [7]. Mortality rates have fallen markedly, but new complications have arisen, such as myocardial infarction and cancer [1], [8], [9], [10], [11]. These changes have led to an increase in life expectancy, more geographic mobility, and a diversification of physicians caring for HIV-infected patients. All these factors are likely to contribute to a higher rate of LFU in cohorts where data are mainly collected in infectious diseases and internal medicine wards.

The objectives of this study were to determinate the prevalence of LFU in the French Hospital Database on HIV infection (ANRS CO 04 FHDH), and its determinants.

Section snippets

Description of the FHDH

The characteristics of the French Hospital Database on HIV (FHDH) have been described in detail elsewhere [12]. Briefly, FHDH is a clinical epidemiologic network implemented, since 1992, in 62 French University Hospitals belonging to 29 HIV treatment and information centers (CISIH) located both in continental France and overseas territories. The only FHDH inclusion criteria are documented HIV-1 or HIV-2 infection and written informed consent. Data are recorded prospectively by trained research

Results

Among the 34,835 patients included in the analysis, 2,950 (8.5%) were lost to follow-up and 1,478 (4.2%) died after their last visit in 1999.

Among the 4,962 patients whose seropositivity had been diagnosed <1 year previously, 834 (16.8%) were lost to follow-up and 185 (3.7%) died (Table 1). Losses to follow-up were more frequent among migrants. Among patients diagnosed <1 year previously, 629 (12.7%) had stayed in sub-Saharan Africa for more than 6 months, 54 (1.1%) in European Union member

Discussion

In 1999, the overall prevalence of LFU in the French Hospital Database on HIV infection was 8.5%. This prevalence is lower than that observed in the American Foundation of AIDS Research database (16.5%) [16] and similar to those of the French southwestern cohort in 2003 (10.9%) [17] and a prospective study on retention among HIV-infected women [18]. In French cohorts, LFU cannot be explained by the cost of care, as all care is provided free of charge.

We found that patients with a recent

Conclusion

It is important to study factors linked to LFU and their impact on cohort findings. Factors influencing the risk of LFU might also have an impact on patients' health, and a better knowledge of these factors might help to improve patient management. The better the follow-up, the better the patients' clinical and biologic status and the lower the bias. Programs taking into account patients' particularities—recent diagnosis, IVDU, migrants—and geographic specificities should see improved follow-up.

Acknowledgments

This work was supported by the Institut National de la Santé et de la Recherche Médicale, the Agence Nationale de Recherches sur le SIDA, and the Fondation pour la Recherche Médicale.

We are indebted to all the participants in the cohort, the clinical centers, and especially the local research assistants, without whom this work would not have been possible.

References (27)

  • M. Mary-Krause et al.

    Increased risk of myocardial infarction with duration of protease inhibitor therapy in HIV-infected men

    AIDS

    (2003)
  • B. Ledergerber et al.

    Risk of HIV related Kaposi's sarcoma and non-Hodgkin's lymphoma with potent antiretroviral therapy: prospective cohort study. Swiss HIV Cohort Study

    BMJ

    (1999)
  • M. Herida et al.

    Incidence of non-AIDS-defining cancers before and during the highly active antiretroviral therapy era in a cohort of human immunodeficiency virus-infected patients

    J Clin Oncol

    (2003)
  • Cited by (77)

    • Mortality in migrants living with HIV in western Europe (1997-2013): A collaborative cohort study

      2015, The Lancet HIV
      Citation Excerpt :

      Not all cohorts were able to cross-match their data with the corresponding national death registries, and it is unclear how differential loss to follow-up might have affected our results. Loss to follow-up was more common for some groups of migrants than for native groups, which has also been reported in other cohort studies.7,17,26 Higher rates of loss to follow-up could be explained by legal status or deportations, individuals returning to their home countries, higher geographical mobility, or barriers to regularly accessing services.

    View all citing articles on Scopus

    A complete list of participants from the Clinical Epidemiology Group of the French Hospital Database is available on the journal's website at www.elsevier.com.

    View full text