You're asking the wrong questions: you can't know in advance that a breakthrough in energy won't bring about a disruptive impact on healthcare. For example, if

You're asking the wrong questions: you can't know in advance that a breakthrough in energy won't bring about a disruptive impact on healthcare. For example, if you were to able to stabilize a biobattery of mitrochondria, you might incidentally solve a problem related to diabetes (example made up). In other words, because you can't predict the impact of your effort, you can't reliably infer where to direct your effort. By analogy with machine learning, instead of throwing lots of effort at a non-convex problem, you might instead choose a convex problem instead: e.g. what daily process allows you to achieve high problem solving output? what expertise can you acquire and put to use regularly, which is sustainable (no burnout, pays rent, etc)?