Process-Driven Metrics and Process Evaluation of Bundled Interventions: The Agriculture to Nutrition (ATONU) Trial

• Activity Number: 208 - Survey Estimation
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Evidence Matangi* and George McCabe and Tshilidzi Madzivhandila and Farai Gwelo and Bertha Munthali and Simbarashe Sibanda and Wafaie Fawzi and Nilupa Gunaratna
• Companies: and Purdue University and FANRPAN and FANRPAN and FANRPAN and FANRPAN and Harvard and Purdue University
• Keywords: compliance; cluster randomized controlled trials; metrics; multilevel model; process evaluation; variance decomposition

Abstract

Public health interventions increasingly comprise multiple bundled components, intended to additively or synergistically improve health outcomes. These interventions are often evaluated in cluster randomized controlled trials (cRCTs) in which implementation may be heterogeneous. Intervention impact is determined by how participants interact with intervention components. We propose process-driven metrics to capture participation in interventions with multiple components and heterogeneous implementation. Using data from the Agriculture to Nutrition (ATONU) cRCT in Ethiopia, we compare the proposed metrics with traditional process metrics in describing implementation dynamics. Variation in the metrics at differing geographical scales, assessed using beta-logistic and logistic multilevel models, indicated specific needs to improve implementation quality. Generalized mixed models were applied to link the intervention and study outcome through the proposed metrics. One compliance metric was a significant determinant for the intervention’s impact on the primary outcome, women’s dietary diversity. Improved process metrics can strengthen the evidence for impact of complex interventions.