Engagement with an LMS-based tobacco harm reduction and smoking cessation program among frontline healthcare workers in Malawi: threshold adoption dynamics and determinants

Background Digital learning management systems (LMS) are increasingly used to strengthen frontline healthcare worker capacity in tobacco harm reduction (THR) and smoking cessation, particularly in low- and middle-income countries (LMICs). However, implementation research has focused largely on training effectiveness rather than behavioral patterns of digital platform adoption. We characterized LMS engagement and examined determinants of training uptake among frontline healthcare workers in Malawi, testing a ‘threshold adoption’ pattern, in which participants either do not engage or complete all content, with limited intermediate participation. Methods We analyzed LMS data from 267 frontline healthcare workers in Lilongwe and Mzimba North districts participated in a blended THR and smoking cessation program comprising a one-day workshop and five asynchronous LMS modules. Outcomes included full completion (all five modules), total modules completed (0-5), and total quizzes completed (0-5). Engagement distributions were described, and coherence between module and quiz completion was assessed using Pearson correlation. Predictors of full completion were estimated using multivariable logistic regression. To test threshold adoption, we applied a two-part hurdle model: logistic regression for initiation (≥1 module vs 0) and zero-truncated Poisson regression among initiators (1-5 modules), reporting adjusted incidence rate ratios (aIRRs). Follow-up telephone feedback from non-completers (n = 26) was summarized using rapid content analysis and mapped to the Consolidated Framework for Implementation Research (CFIR). Results Engagement was polarized: 36.7% completed no modules and 38.2% completed all five. Modules and quizzes completed were strongly correlated (r = 0.777, p < 0.001). Compared with participants aged <30 years, older groups had lower odds of full completion (30-39: aOR = 0.48, 95% CI 0.24-0.98; 40-49: aOR = 0.22, 95% CI 0.08-0.60; 50+: aOR = 0.06, 95% CI 0.01-0.28). No measured variables predicted initiation. Among initiators (n = 169), participants in Mzimba North completed fewer modules than those in Lilongwe (aIRR = 0.85, 95% CI: 0.75–0.97). Reported barriers included workload constraints, connectivity limitations, and login or device challenges. Conclusion LMS engagement exhibited threshold adoption dynamics, with distinct determinants for initiation and persistence. Implementation strategies should pair early activation supports with context-sensitive mechanisms to sustain completion.