Community-Based Accompaniment and Psychosocial Health Outcomes in HIV-Infected Adults in Rwanda: A Prospective Study

Abstract

We examined whether the addition of community-based accompaniment to Rwanda’s national model for antiretroviral treatment (ART) was associated with greater improvements in patients’ psychosocial health outcomes during the first year of therapy. We enrolled 610 HIV-infected adults with CD4 cell counts under 350 cells/μL initiating ART in one of two programs. Both programs provided ART and required patients to identify a treatment buddy per national protocols. Patients in one program additionally received nutritional and socioeconomic supplements, and daily home-visits by a community health worker (“accompagnateur”) who provided social support and directly-observed ingestion of medication. The addition of community-based accompaniment was associated with an additional 44.3 % reduction in prevalence of depression, more than twice the gains in perceived physical and mental health quality of life, and increased perceived social support in the first year of treatment. Community-based accompaniment may represent an important intervention in HIV-infected populations with prevalent mental health morbidity.

Resumen

Este estudio evaluó si el agregado del acompañamiento comunitario al modelo nacional de tratamiento antirretroviral (TARV) utilizado en Ruanda, se asocia a mejores resultados en la salud psicosocial de los pacientes durante el primer año de tratamiento. Se enrolaron 610 adultos infectados con VIH, con recuento de células CD4 inferior a 350 por microlitro y que iniciaron el TARV en uno de dos programas. Ambos programas proporcionaron tratamiento antirretroviral y, según protocolo nacional, se les solicitó a los pacientes la identificación de un compañero para el tratamiento. En uno de los programas, los pacientes recibieron además: ayuda nutricional y socioeconómica y visitas diarias de agentes sanitarios de la comunidad (acompañadores) que proporcionaron apoyo social y que vigilaron directamente la toma de los medicamentos. El acompañamiento comunitario se asoció a una reducción adicional de 44,3 % en la prevalencia de depresión, a un aumento de más del doble en la calidad de vida física y mental percibida y a un aumento de la percepción del apoyo social durante el primer año de tratamiento. El acompañamiento comunitario puede representar una importante intervención en las poblaciones infectadas por el VIH y con una predominante morbilidad en la salud mental.

Appendices

Appendix 1

Profile Analysis

Continuous outcome estimates (social support score, mental health quality of life score, and physical health quality of life score) for baseline, 6, and 12 months were modeled as:

$$ Y_{ij} = \, \alpha \, + \, {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + \, {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + \, B_{3} \left( {trt*time_{ij} } \right) + {\varvec{\beta}} \left( \varvec {X} \right) + \, \varepsilon_{ij} $$

(1)

where health outcomes followed a normal distribution and were a function of a constant term (α), treatment group (trt _i ), measurement occasion (time _ij ), interaction of treatment group across measurement occasions (trt * time _ij ), and a set of covariates (X) with associated vectors of coefficients (β), while ε _ij was an error term that varied among individuals (i) and across time (j). These models accounted for repeated measures of individuals with an unstructured covariance matrix.

We conducted a similar analysis to calculate the prevalence of depression, a common binary outcome, except we assumed the outcome followed a Poisson distribution and used a robust variance estimator:

$$ \log \left( {Y_{ij} } \right) = \, \alpha \, + \, {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + \, {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + \, B_{3} \left( {trt*time_{ij} } \right) + {\varvec{\beta}} \left(\varvec {X} \right) + \, \varepsilon_{ij.} $$

(2)

Using the estimated outcomes for each individual and time point from the above models, we estimated expected change of each individual from baseline for all outcomes including depression, and then averaged within treatment groups:

$$ \Updelta Y_{i} = Y_{ij = 6} - Y_{ij = 0} $$

(3)

$$ \Updelta Y_{i} = Y_{ij = 12} - Y_{ij = 0} $$

(4)

Adjust for Baseline Analysis

Continuous outcome estimates (social support score, mental health quality of life score, and physical quality of life health score) for 6 and 12 months were modeled as:

$$ Y_{ij} = \alpha + {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + B_{3} \left( {trt \, * \, time_{ij} } \right) + {\varvec{\beta}}_{4} \left( {baseline_{i} } \right) + {\varvec{\beta}} \left( \varvec {X} \right) + \varepsilon_{ij} $$

(5)

where health outcomes followed a normal distribution and were a function of a constant term (α), treatment group (trt _i ), measurement occasion (time _ij ), interaction of treatment group across measurement occasions (trt * time _ij ), baseline score (baseline _i ), and a set of covariates (X) with associated vectors of coefficients (β), while ε _ij was an error term that varied among individuals (i) and across time (j). These models accounted for repeated measures of individuals; because outcomes are modeled at just two time points, a compound symmetry covariance structure was specified.

We conducted a similar analysis to calculate the prevalence of depression, a common binary outcome, except we assumed the outcome followed a Poisson distribution and used a robust variance estimator:

$$ \log \left( {Y_{ij} } \right) = \alpha + {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + B_{3} \left( {trt*time_{ij} } \right) + {\varvec{\beta}}_{4} \left( {baseline_{i} } \right) + {\varvec{\beta}} \left( \varvec {X} \right) + \varepsilon_{ij}. $$

(6)

To model change since baseline, we used the following model for all outcomes including depression:

$$ Y_{ij = 6} -baseline_{i} = \alpha + {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + B_{3} \left( {trt \, * \, time_{ij} } \right) + \left( {{\varvec{\beta}}_{4} - \, 1} \right)\left( {baseline_{i} } \right) + {\varvec{\beta}} \left( \varvec {X} \right) + \varepsilon_{ij}; $$

(7)

$$ Y_{ij = 12} -baseline_{i} = \alpha + {\varvec{\beta}}_{1} \left( {trt_{i} } \right) + {\varvec{\beta}}_{2} \left( {time_{ij} } \right) + B_{3} \left( {trt \, * \, time_{ij} } \right) + \left( {{\varvec{\beta}}_{4} - \, 1} \right)\left( {baseline_{i} } \right) + {\varvec{\beta}} \left( \varvec {X} \right) + \varepsilon_{ij}. $$

(8)

Appendix 2

See Tables 5 and 6.

Table 5 Multivariable models of community-based accompaniment and psychosocial health outcomes at six months, profile analysis

Table 6 Multivariable models of community-based accompaniment and psychosocial health outcomes at 12 months, profile analysis