Come to think of it, the general idea is "If you, as an insurance company, foot the bill for these treatments with long-term benefits for your clients, this will cost $X, but save you $Y and other nearby insurance companies $Z in the future. Therefore, we the association will pay you $z < $Z to do it, and by assumption $Y+$Z > $X, so there should exist values of $z that make this profitable for both parties (anything between $X-$Y and $Z)." The carrot is the payment of $z.
In terms of whether this association could become a price-fixing cartel... It's probably easier to verify compliance when compliance consists of performing a positive behavior (i.e. paying healthcare providers to do certain treatments; whatever the details of the payment, the important thing is whether the treatment was performed) instead of a negative behavior (i.e. not selling extra product at a lower price on the side). It doesn't seem obvious that this would make price-fixing much easier than it currently is.