Differentiated Care for Improved Health Systems Efficiency and Health Outcomes in Zambia

This is a joint endeavor of the Zambian Ministry of Health (MoH), Ministry of Community Development, Mother \& Child Health (MCDMCH) and the Centre for Infectious Disease Research in Zambia (CIDRZ) with funding from the Bill \& Melinda Gates Foundation.

The study evaluates four models of streamlined HIV care in reducing patients' time and cost: (1) community adherence groups (CAG), (2) urban adherence groups (UAG), (3) Fast-Track clinic visits and (4) streamlined ART initiation (START). CAG and UAG are assessed using a cluster-randomized design; Fast-Track and START using an observational approach to compare changes in the outcome through a difference in difference approach.

1. Specific Objectives

Objective 1:

To determine acceptability, appropriateness, and feasibility of a differentiated care system in Zambia, by assessing perspectives of (1) patients and family members (2) health care workers and (3) government and local leaders.

Objective 2:

To evaluate effectiveness (retention), efficiency (cost-effectiveness), and health care quality (quantitative and qualitative measures) of a differentiated and targeted care system.

Objective 3:

To develop a "methodologic" toolkit for assessment of local needs, preferences and scale-up of differentiated care models in varied contexts, using data from Objectives 1 and 2.

2.0 Methodology 2.1 Objective 1 Mixed methods approach will be used to evaluate perceptions on challenges in accessing and adhering to life-long HIV treatment and to assess needs/preferences for differentiated care models collected from patients and family members (demand side), health facilities -professional and community health workers (supply side), and government and local leaders using in-depth interviews (IDIs), focus group discussions (FGDs), and discrete choice surveys (DCS) to sequentially analyze supply and demand side factors that may influence uptake.

Population Key government and other stakeholders in HIV or other relevant community based service delivery will be interviewed. Professional and lay health workers in ART program, Pre-ART and ART patients and, a sample of household members of ART or pre-ART patient registered at selected health facility will be interviewed.

Measurements/Procedures For Objective 1: IDIs, FGDs, medical chart reviews, DCS/general survey Semi-structured guides by participant type will be used to conduct IDIs and FGDs in English or a local language (Nyanja, Bemba, or Tonga). Where permitted, IDIs/FGDs will be audio-recorded for transcription and analysis. Participants will be reimbursed for travel. IDIs will last 1 and FGDs 1-2 hours.

A cross-sectional survey will be used to measure patient access and psychosocial needs. Access needs will be assessed by questions on appropriateness of current clinic operating hours; distance, cost, and time to reach clinic; household income, assets, and expenditures; and opportunity costs of a clinic visit. Patient costs related to health care utilization, including transport, communication, and food costs will be elicited. To assess psychosocial needs, validated scales and indices to measure stigma (Revised Berger), depression (PHQ-9), alcohol consumption (AUDIT-C), and domestic violence will be applied. The survey administered in the language of participant's choice by trained survey enumerators using tablets will take approximately 30-45 minutes and uploaded to the central database.

The electronic clinical information system (SmartCare) will be utilized to evaluate patient clinical needs. Distribution of clinical needs will be assessed for degree of immunosuppression at entry into care, frequency of opportunistic infections after treatment initiation, and frequency of "ill" versus "healthy" visits.

DCSs will be used to assess patient preference for different, discrete characteristics of various models of community-based ART delivery. All possible combinations of key attributes/aspects of community-based ART models and number of levels/options for them will be put in subsets divided in blocks. Each participant is asked only one possible block of about 10 questions.

Analytic Approach Qualitative data will be imported into software for managing qualitative analysis (e.g. N-Vivo or Atlas TI), iteratively coded and cross-referenced with survey data. Scale scores may be analyzed using descriptive statistics and linear regression modeling. For DCS, main effects plus two-way interactions between different parameters will be estimated using a nested logit framework.

Sampling Approach/Sample Size Considerations The convenience sample consists of 20 IDIs with government and community leaders; two FGDs each with professional HCW, Lay HCW and family members from1 urban and 1 rural clinic in each province; 4 FGDs with ART patients (2 urban and 2 rural clinic) in each province; and 2 FGDs with pre-ART patients (1 urban and 1 rural clinic) in Lusaka province only.

General survey: 1,600 patients (of which 800 are ART and 800 are pre-ART). A precision approach was used to calculate number of ART and pre-ART participants respectively necessary to estimate a binary indicator with 5% margin of error and 95% confidence interval, assuming a prevalence of p=0.5 and a design effect of 2 [ n=((1.96^2) * 0.5* 0.5 *2)/(0.05^2)=768].

2.2 Objective 2: To evaluate the effectiveness, efficiency, and health care quality of a differentiated system of care that includes targeted models of care.

Effectiveness of a differentiated system of HIV care delivery in which frequency, type, or intensity of contact with the health system is tailored to patient needs will be evaluated. CAGs, UAGs, fast-track, and START will be simultaneously implemented and evaluated using a common evaluation framework that includes effectiveness (retention), efficiency, and quality.

Methodology A matched cluster randomized design approach is proposed for CAG and UAG models to account for the clinic-level intervention and possible selection bias. With little to no observational/ implementation data for Fast-Track and START models globally, a difference-in-difference design will be used to calculate effect of the models on outcomes.

Population See inclusion and exclusion criteria Procedures Description of Interventions: Entered elsewhere

Patient recruitment:

At all intervention sites: Facility staff will identify eligible patients during routine clinic visits and notify the study research assistant about potential study participants. The research study assistant will assess patient willingness to participate in the intervention and will obtain informed consent for study participation.

START model: Staff will use national guidelines to determine ART eligibility and, with participants' informed consent, expedite provision of on-site CD4 testing using point of care platform for all newly diagnosed and ART-naive patients due for routine CD4 monitoring. Participants will not be actively followed in the 12-months after ART initiation.

At control sites: Eligible patients at control sites, willing to join CAG and UAG studies, will be enrolled. Consent will be obtained to extract their SmartCare data.

Baseline viral load testing:

For participants enrolled into CAG, UAG and Fast-Track intervention, dry blood spot specimens will be obtained via finger prick. Specimens will be transmitted to CIDRZ central laboratory for baseline viral load testing. Results will be communicated to providers and patients and, documented in SmartCare.

Costing A subset of participants in each model will be interviewed on direct and indirect health costs of the intervention using a semi-structured questionnaire. Administrative data may also be used.

Exit viral load testing and patient exit-survey All participants will have an exit viral load test in the 12th month. In CAG, UAG, and Fast-Track models, a patient exit-survey will assess patient satisfaction and patient centeredness.

Measurements Standard de-identified socio-demographic, laboratory and clinical data will be extracted from SmartCare.

Information in model-specific registers (attendance, receipt of medications, symptom checklist, patient preference between intervention and standard care, up-referral due to presence of symptoms or a patient's desire to return to standard care, or down-referral to model when patients with acute illness stabilize) will be entered into the study database using a tablet.

Patient satisfaction and centeredness will be captured by the semi-structured patient-exit survey conducted after the 12-month intervention period.

Baseline and exit viral load testing will be conducted. A semi-structured costing questionnaire used with a subset of intervention participants will assess related direct and indirect health costs. To calculate resource use of each activity, time-in-motion studies will be conducted among patients and health care workers prior to and during the intervention.

Process data will be collected facility-level assessments of pharmacy stock outs and through HCW interviews at end of follow-up period.

Analytic Approach For START, date of ART initiation is the primary outcome. For CAG and UAG interventions, Kaplan-Meier estimates of time to the primary outcome, stratified by intervention condition, using log rank tests will be calculated. For Fast-track and START, primary analysis will take a difference-in-difference approach to estimate difference in the change in the rate of the outcome in intervention settings. In secondary analyses of the primary outcome, longer intervals of lateness that comprise a missed pharmacy visit (14 and 30 days) as well as adjustment for baseline patient characteristics will be explored along with analysis using medication possession ratio (MPR) as the outcome and other metrics of retention that incorporate both frequency of missed visit and interval of time before return.

Sample Size Considerations CAG/UAG: Program data suggest that 65% of patients are > 7 days late for a pharmacy refill visit in their first year of ART in 95% (+15%) of clinics. Assuming a conservative matched co-efficient of variation of 0.10 and a 50% reduction in missed pharmacy visits, selection of five clinics and 90 patients per site will yield power of 96% for CAG intervention. Under more conservative assumptions about between clinic variability and the effect size, this sample size yields >80% power. For UAG intervention, target sample size is 4 UAGs of 30 people each in five intervention sites (120 participants per site; 600 across sites). Under the same assumptions as the CAG model, this sample size yields power of > 90% and is robust to varying assumptions about correlation.

Fast-track: Targeting 200 patients at each of four sites (2 intervention and 2 control sites), a fall to 40% of the assumed 80% patients that miss a pharmacy visit in the first year, gives 80% power.

START: Assuming ART initiation at 2 weeks rises from 60% to 85%, 100 patients at each of eight sites (4 intervention and 4 control sites; n=800), rho of 0.15 yields an anticipated power of 78%.

2.3 Objective 3: Methodological Toolkit After data synthesis, a methodological toolkit will be developed to guide assessment of feasibility, acceptability, needs, preferences and implementation for differentiated care models. Data will be collected using work process tracking and costing; and interviews with study staff and key decision makers. The toolkit will be easily digestible and lead to actionable plans to move from current to differentiated care models in differing circumstances and contexts. With input from reviewers, the toolkit will be finalized and digital and hard copies will be distributed at dissemination events with the MoH, MCDMCH and other national and sub-national stakeholders.