Lmst

How can we reliably spread evidence-based practices at the speed and scale modern health challenges demand?

At a symposium of the American Society for Tropical Medicine and Hygiene (ASTMH) Annual Meeting, I explored how peer learning could help us tackle five critical challenges that limit effectiveness in global health.

Performance: How do we move beyond knowledge gains to measurable improvements in health outcomes?
Scale and access: How do we reach and include tens of thousands of health workers, not just dozens?
Applicability: How do we ensure learning translates into changed practice?
Diversity: How do we leverage different perspectives and contexts rather than enforce standardization?
Complexity: How do we support locally-led leadership for change to tackle complex challenges that have no standard solutions?

https://www.youtube.com/watch?v=Q08dbbzUzzc

For epidemiologists working on implementation science, peer learning provides a new path for solving one of global health’s most persistent challenges: how to reliably spread evidence-based practices at the speed and scale modern health challenges demand.

The evidence suggests we should view peer learning not just as a training approach, but as a mechanism for viral spread of effective practices through health systems.

How do we get to attribution?

Of course, an epidemiologist will want to know if and how improved health outcomes can be attributed to peer learning interventions.

The Geneva Learning Foundation (TGLF) addresses this fundamental challenge in implementation science – proving attribution – through a three-stage process that combines quantitative indicators with qualitative validation.

The process begins with baseline health indicators relevant to each context (such as vaccination coverage rates, if it is immunization), which are then tracked through regular “acceleration reports” that capture both metrics and implementation progress.

Rather than assuming causation from correlation, participants must explicitly rate the extent to which they attribute observed improvements to their intervention.

The critical innovation comes in the third stage: those claiming attribution must “prove it” to the community of peers, by providing specific evidence of how their actions led to the observed changes – a requirement that both controls for self-reporting limitations and generates rich qualitative data about implementation mechanisms.

This methodology has proven particularly valuable in complex interventions where randomized controlled trials may be impractical or insufficient.

What are examples of peer learning in action?

Here are three examples from The Geneva Learning Foundation’s work that demonstrate scale, reach, and sustainability.

Within four weeks, a single Teach to Reach cohort of 17,662 health workers across over 80 countries generated 1,800 context-specific experiences describing the “how” of implementation, especially at the district and community levels.

In Côte d’Ivoire, working with Gavi and The Geneva Learning Foundation, the national immunization team used TGLF’s model to support community engagement. Within two weeks, over 500 health workers representing 85% of the country’s districts had begun implementing locally-led innovations. 82% of participants said they would use TGLF’s model for their own needs, without requiring any further assistance or support.

In TGLF’s COVID-19 Peer Hub, 30% of participants successfully implemented recovery plans within three months – a rate seven times higher than a control group that did not use TGLF’s model.

Participants who actively engaged with peers were not only more likely to report successful implementation, but could demonstrate concrete evidence of how peer interactions contributed to their success, creating a robust framework for understanding not just whether interventions work, but how and why they succeed or fail across different contexts.

Quantifying learning

Using a simple methodology that measures learning efficacy across five key variables – scalability, information fidelity, cost effectiveness, feedback quality, and uniformity – we calculated that properly structured peer learning networks achieve an efficacy score of 3.2 out of 4, significantly outperforming both traditional cascade training (1.4) and expert coaching (2.2).

But the real breakthrough came when considering scale. When calculating the Efficacy-Scale Score (ESS) – which multiplies learning efficacy by the number of learners reached – the differences became stark:

Peer Learning: 3,200 (reaching 1,000 learners)
Cascade Training: 700 (reaching 500 learners)
Expert Coaching: 132 (reaching 60 learners)

Learn more: Calculating the relative effectiveness of expert coaching, peer learning, and cascade training

The mathematics of scale

For epidemiologists, the mechanics of this scaling effect may feel familiar.

In traditional expert-led training, if N is the total number of learners and M is the number of available experts who can each effectively coach K learners, we quickly hit a ceiling where N far exceeds M×K.

TGLF’s model transforms this equation by structuring interactions so each learner gives and receives feedback from exactly three peers, guided by expert-designed rubrics.

This creates a linear scaling pattern where total learning interactions = 3N, allowing for theoretically unlimited scale while maintaining quality through structured feedback loops.

Information loss and network resilience

One of the most interesting findings concerns information fidelity. In cascade training, knowledge degradation follows a predictable pattern:

where Kn is the knowledge at the nth level of the cascade and α is the loss rate at each step. This explains why cascade training, despite its theoretical appeal, consistently underperforms.

In contrast, TGLF’s peer learning-to-action networks showed remarkable resilience. By creating multiple pathways for knowledge transmission and building in structured feedback loops, the system maintains high information fidelity even at scale.

Learn more: Why does cascade training fail?

References

Arling, P.A., Doebbeling, B.N., Fox, R.L., 2011. Improving the Implementation of Evidence-Based Practice and Information Systems in Healthcare: A Social Network Approach. International Journal of Healthcare Information Systems and Informatics 6, 37–59. https://doi.org/10.4018/jhisi.2011040104

Hogan, M.J., Barton, A., Twiner, A., James, C., Ahmed, F., Casebourne, I., Steed, I., Hamilton, P., Shi, S., Zhao, Y., Harney, O.M., Wegerif, R., 2023. Education for collective intelligence. Irish Educational Studies 1–30. https://doi.org/10.1080/03323315.2023.2250309

Watkins, K.E., Sandmann, L.R., Dailey, C.A., Li, B., Yang, S.-E., Galen, R.S., Sadki, R., 2022. Accelerating problem-solving capacities of sub-national public health professionals: an evaluation of a digital immunization training intervention. BMC Health Serv Res 22, 736. https://doi.org/10.1186/s12913-022-08138-4

Share this:

#AmericanSocietyForTropicalMedicineAndHygiene #ASTMH #attribution #cascadeTraining #globalHealth #implementationScience #peerLearning #TheGenevaLearningFoundation #TropMed24

ASTMH 2024 How can we reliably spread evidence-based practices at the speed and scale modern health challenges demand

Calculating the relative effectiveness of expert coaching, peer learning, and cascade training

A formula for calculating learning efficacy, (E), considering the importance of each criterion and the specific ratings for peer learning, is:

This abstract formula provides a way to quantify learning efficacy, considering various educational criteria and their relative importance (weights) for effective learning.

Variable DefinitionDescription SScalabilityAbility to accommodate a large number of learners IInformation fidelityQuality and reliability of information CCost effectivenessFinancial efficiency of the learning method FFeedback qualityQuality of feedback received UUniformityConsistency of learning experience Summary of five variables that contribute to learning efficacy

Weights for each variables are derived from empirical data and expert consensus.

All values are on a scale of 0-4, with a “4” representing the highest level.

ScalabilityInformation fidelityCost-benefitFeedback qualityUniformity4.003.004.003.001.00Assigned weights

Here is a summary table including all values for each criterion, learning efficacy calculated with weights, and Efficacy-Scale Score (ESS) for peer learning, cascade training, and expert coaching.

The Efficacy-Scale Score (ESS) can be calculated by multiplying the efficacy (E) of a learning method by the number of learners (N).

This table provides a detailed comparison of the values for each criterion across the different learning methods, the calculated learning efficacy values considering the specified weights, and the Efficacy-Scale Score (ESS) for each method.

Type of learningScalabilityInformation fidelityCost effectivenessFeedback qualityUniformityLearning efficacy# of learnersEfficacy-Scale ScorePeer learning4.002.504.002.501.003.2010003200Cascade training2.001.002.000.500.501.40500700Expert coaching0.504.001.004.003.002.2060132

Of course, there are many nuances in individual programmes that could affect the real-world effectiveness of this simple model. The model, grounded in empirical data and simplified to highlight core determinants of learning efficacy, leverages statistical weighting to prioritize key educational factors, acknowledging its abstraction from the multifaceted nature of educational effectiveness and assumptions may not capture all nuances of individual learning scenarios.

Peer learning

The calculated learning efficacy for peer learning, , is 3.20. This value reflects the weighted assessment of peer learning’s strengths and characteristics according to the provided criteria and their importance.

By virtue of scalability, ESS for peer learning is 24 times higher than expert coaching.

Cascade training

For Cascade Training, the calculated learning efficacy, , is approximately 1.40. This reflects the weighted assessment based on the provided criteria and their importance, indicating lower efficacy compared to peer learning.

Cascade training has a higher ESS than expert coaching, due to its ability to achieve scale.

Learn more: Why does cascade training fail?

Expert coaching

For Expert Coaching, the calculated learning efficacy, , is approximately 2.20. This value indicates higher efficacy than cascade training but lower than peer learning.

However, the ESS is the lowest of the three methods, primarily due to its inability to scale. Read this article for a scalability comparison between expert coaching and peer learning.

Share this:

#cascadeTraining #expertCoaching #fellowship #mathematicalModeling #peerLearning

Calculating the relative effectiveness of expert coaching, peer learning, and cascade training

Why does cascade training fail?

Cascade training remains widely used in global health.

Cascade training can look great on paper: an expert trains a small group who, in turn, train others, thereby theoretically scaling the knowledge across an organization.

It attempts to combine the advantages of expert coaching and peer learning by passing knowledge down a hierarchy.

However, despite its promise and persistent use, cascade training is plagued by several factors that often lead to its failure.

This is well-documented in the field of learning, but largely unknown (or ignored) in global health.

What are the mechanics of this known inefficacy?

Here are four factors that contribute to the failure of cascade training

1. Information loss

Consider a model where an expert holds a knowledge set K. In each subsequent layer of the cascade, α percentage of the knowledge is lost:

Where is the knowledge at the nth level of the cascade. As n grows, exponentially decreases, leading to severe information loss.
Each layer in the cascade introduces a potential for misunderstanding the original information, leading to the training equivalent of the ‘telephone game’.

2. Lack of feedback

In a cascade model, only the first layer receives feedback from an actual expert.

Subsequent layers have to rely on their immediate ‘trainers,’ who might not have the expertise to correct nuanced mistakes.
The hierarchical relationship between trainer and trainee is different from peer learning, in which it is assumed that everyone has something to learn from others, and expertise is produced through collaborative learning.

3. Skill variation

Not everyone is equipped to teach others.
The people who receive the training first are not necessarily the best at conveying it to the next layer, leading to unequal training quality.

4. Dilution of responsibility

As the cascade flows down, the sense of responsibility for the quality and fidelity of the training dilutes.
The absence of feedback to drive a quality development process exacerbates this.

Share this:

#cascadeTraining #learningScience #pedagogy

#cascadeTraining

Client Info