With models you always need to look at assumptions. The paper there assumes 0 adaptation/replacement, simply correlates reductions in mortality with USAID funding trends without even attempting to assess causality, operates entirely at population (as opposed to individual) level analysis, and more. So you're looking at an extreme high end ballpark that is not intended to be used as a prediction or measurement of what will happen. In the study's own words:

---

".. the primary aim of our analysis was not to model in detail the full range of potential USAID defunding scenarios or their varied effects. Instead, it was to estimate the consequences of a worst-case scenario—defined by the most severe funding cuts—relative to a business-as-usual continuation of USAID operations. Importantly, our goal was not to forecast the exact all-age and child mortality by 2030, but to compare two parallel scenarios that differ only in the path of the main exposure variable.."

---

The vacuum USAID has left began being rapidly filled and replaced by other countries - most notably China but also including the EU nations, Mideast, and many others. You'll be able to look at the death rate of countries like Nigeria (where USAID spending was quite high) in the near future to see the impact. If it was anywhere near as significant as the extreme scenarios, it'll show up very visibly in the death rate data. I do not expect we will see anything like that.