Custom risk thresholds per CHW

[garethbowen]

Currently, each individual's risk is compared to a single threshold to determine whether they are high risk or not. We know that high risk individuals will often be located near each other, (for example, an area of a community that has a higher percentage of very poor families). This means that high risk individuals are not distributed evenly across health workers, and we have seen some health workers have a much larger number of new high risk tasks than others.

Risk thresholds should be customizable by the health worker to ensure each health worker gets a reasonable number of high-risk tasks.

One approach we have discussed is a Hybrid Risk Threshold: We maintain an absolute threshold which individuals must be above to receive any tasks, but also include relative thresholds for users who have more high-risk tasks than would likely be possible. This would allow us to manage health worker workloads while also preserving a concept of absolute risk.

Original issue

[yembrick]

It turns out that we might not be great at choosing the right thresholds. We'll know more after next week's design visit (week of June 10). To clarify one of the motivations here, we want to prevent health workers from getting overloaded with tasks, if so many of the individuals they serve are high risk (for example, if half of their people are high risk)-- we might also think about setting a maximum number of high risk patients per health worker, or maximum number of high risk tasks in a month, where the highest risk scores are prioritized. theres more to discuss here

[yembrick]

After the site visit and more discussion, were thinking we should consider normalizing risk scores by CU. But we want to collect more data and learn more before doing this, so its likely to be a feature required for the next round of scaleup, not the current pilot extension.