Table 1 Measurements

Nutritional Pathway

Nutritional status

Nutritional status was assessed through body mass index (BMI) and mid-upper arm circumference (MUAC), commonly used to assess nutritional status (Physical Status 1995; Collins et al. 2000). The BMI reflects protein and fat reserves (James et al. 1988) and was assessed using an established grading system (Ferro-Luzzi et al. 1992). For MUAC, we used WHO sex-specific cut-offs of 22.0 cm for females and 23.0 cm for males with chronic energy deficiency (Doocy et al. 2005).

Semi-annually

Food frequency

Food frequency, the number of different foods or food groups and the frequency consumed over a given reference period (Hoddinott & Yohannes 2002), as adapted from the World Food Programme Food Consumption Score was collected.

Quarterly

Behavioral Pathway

Pill count ART adherence

Participants received an unannounced visit to inventory medications and count pills (Bangsberg et al. 2001a; Bangsberg et al. 2001b; Bangsberg et al. 2000), a technique closely correlated with electronically monitored adherence, HIV viral load (Bangsberg et al. 2000) and progression to AIDS (Bangsberg et al. 2001c). The count of existing pills was reconciled with the participant’s pharmacy refill history to determine the percentage of pills not yet consumed.

Quarterly

Competing demands

Questions were modified from Gelberg and Anderson’s Behavioral Model for Vulnerable Populations (Gelberg et al. 2000; Gelberg et al. 1997) to assess how often lack of food interferes with ability to procure drugs or visit the clinic.

Semi-annually

Healthcare access

Utilization of health care services including hospitalizations and clinic visits over the preceding 6 months were collected.

Semi-annually

Mental Health Pathway

Mental health & depression

Mental health status was measured using the Medical Outcomes Study HIV Health Survey (MOS-HIV), a tool for assessing health-related quality of life (Wu et al. 1991) that has been validated in resource-limited settings (Chatterton et al. 1999; Mast et al. 2004). Depression was screened using the Hopkins Symptom Check-list for depression, a 15-item scale (Derogatis et al. 1974) which has been validated in sub-Saharan Africa (Bolton et al. 2004).

Semi-annually

HIV-related stigma

We used the Internalized AIDS-Related Stigma Scale (Kalichman et al. 2009).

Semi-annually

Disclosure of HIV status

We asked about disclosure of HIV status to partners, family members, friends, colleagues, and public. These questions were adapted from our previous studies in Uganda, Botswana and Swaziland (Weiser et al. 2007; Wolfe et al. 2008; Tsai et al. 2013).

Semi-annually

Alcohol use

To measure alcohol use, we adapted the Alcohol Use Disorders Identification Test (AUDIT-C) indicators. The AUDIT-C is a 3-item alcohol screen that can help identify persons who are hazardous drinkers or have active alcohol use disorders.

Semi-annually

Empowerment

Gender empowerment

Empowerment indicators were adapted from a large cluster-randomized trial of an intervention including: greater challenges to established gender roles, communication with relationship partner about sexual matters in the prior 3 months, measures of financial decision-making, measures of attitudes towards gender roles and gender-based violence, and experience of controlling behavior by relationship partner in prior 3 months (Pronyk et al. 2006). In addition, we used the Sexual Relationship Power Scale (SRPS) (Pulerwitz et al. 2000), which conceptualizes sexual relationship power as a multi-dimensional construct consisting of relationship control and decision making dominance. The SRPS has been used successfully in observational research conducted in South Africa (Dunkle et al. 2004; Jewkes et al. 2010) and Uganda.(Weiser et al. 2010a, b) We also collected data on sexual victimization and perpetration in the prior 3 months.

Semi-annually

Proximal Mediators

Food insecurity

The Household Food Insecurity Access Scale (HFIAS) has been validated in eight countries (Coates et al. 2006a; Swindale & Bilinsky 2006; Frongillo & Nanama 2006; Coates et al. 2006b) and used successfully by our team in rural Uganda. (Weiser et al. 2010a; Tsai et al. 2011; Weiser et al. 2012; Tsai et al. 2012; Weiser et al. 2010a, b; Miller et al. 2011; Tsai et al. 2011)

Semi-annually

Agricultural measures

Agricultural measures were adapted from outcome evaluations developed by Kickstart in Kenya and supplemented by outcome indicators found in an earlier pilot study and a rural assessment. These measures were designed to evaluate uptake and adoption, and to measure changes in agricultural practices including crop diversity and agricultural practices and production. In addition, we evaluated the effectiveness of the training, and specific topics within, so as to refine the training for the subsequent larger cluster-randomized trial.

Quarterly

Household economic indicators

A modification of the World Bank Living Standards Measurement Study (LSMS) questionnaire (Grosh & Glewwe 1998) was used to measure: a) expenditures (food, health, education and productive investments); b) consumption (food and non-food); c) income (from agriculture and all sources); and d) inter-household commodity and cash transfers.

Quarterly

Behavioral Outcome

Risky sexual behaviors

The primary transmission risk outcome was unprotected sex. Other outcomes included: number of non-spousal/non-cohabiting sexual partners, sex-exchange (exchanging sex for money, food, or other resources) Dupas & Robinson 2010); Robinson & Yeh 2011).

Semi-annually

Health Outcomes

HIV-related mortality

Burial permits and information from family members were used to determine cause of death.

As needed

Viral load suppression

Viral load testing was performed on venous blood on the COBAS TaqMan HIV viral load platform (Roche Molecular Diagnostics, Pleasanton, CA) with a lower limit of detection of <20 copies/mL.

Semi-annually

CD4 Count

We abstracted data for CD4 counts from participant’s medical records. Absolute CD4 count testing was performed on whole blood using the BD FACSCount (BD Bioscience, San Jose, CA).

Semi-annually

HIV morbidity

HIV morbidity was measured through key outcomes from the medical record. We abstracted data every 3 months for ART treatment interruptions and episodes of opportunistic infections. We also gathered self-report data on opportunistic infections and symptoms during structured interviews.

Quarterly/Semi-annually

Physical health

Health status was measured using the MOS-HIV, a tool for assessing health-related quality of life (Wu et al. 1991) that has been validated in resource-limited settings (Chatterton et al. 1999; Mast et al. 2004).

Semi-annually

Covariates

Demographics

Age, religion, education, marital/partnership status, number of children and household census.

Baseline

Social support

To measure social support, we adapted the Functional Social Support Scale (Antelman et al. 2001), a modified version of the Duke University-University of North Carolina Functional Support Questionnaire (Broadhead et al. 1988) consisting of questions related to perceived emotional and instrumental support. Higher scores reflect higher levels of social support.

Semi-annually

ART history and Self-reported ART adherence

Detailed ART history, ART self-reported adherence and barriers to ART adherence were collected. For self-report adherence, we used the visual analog scale (Oyugi et al. 2004) and the three day recall.

Semi-annually