You could also maybe structure this as a set of recurring conditional prediction markets on the probability that emergency services will find someone in need of help if they go to any particular location at a particular time. That would involve a lot of markets though, and it doesn't allow for the same depth of information that a generic 911 call does.
The idea here is that if you know someone is hurt or something's on fire, you can get rewarded for reporting it (and if you're wrong, you just lose the fee you paid, which covers the cost of dispatching emergency services unnecessarily). I would expect this to mostly be used by companies and dedicated trading firms rather than individuals, collecting data and running statistical models to figure out that someone needs help before (or maybe instead of, if they're really hurt) they can call for it themselves. For instance, a prediction company might offer to pay anyone for their live location data and other phone data (like acceleration). If, say, someone in a car suddenly decelerates way too fast and everyone around them stops moving, that company could tell emergency services know before anyone on the scene even knows what just happened.
I'm not totally sure how this would work when it comes to reporting crimes. It should be possible to use this report active crimes where someone is in danger, like a mugging or a kidnapping, but what happens if someone uses this to send the police to someone's home where there's evidence of a past crime but no active crime? That brings up a bunch of poison pill issues, since we don't know the evidence the reporter used to make their determination. It could have been illegally obtained, or just not kosher for police investigations. Possibly you can just say that the reward only pays out if there's an active emergency, and if it turns out there's no emergency, police cannot use any information they found for anything. This still means that the police have an incentive to manufacture an emergency somewhere they expect to find evidence of a crime, but they already have that incentive under the current system.
You could accomplish a similar thing by auctioning off a financial instrument that allows the holder of the instrument to tell you where to go for dinner, and pays out depending on the rating you give for the restaurant they send you to.
However, this is worse than a proper market. If Alice thinks they know a really good restaurant for you, they'll outbid everyone else and send you there. But if Bob thinks Alice is overestimating that restaurant, there's nothing Bob can do to make money on that information. So a mistaken trader can cause a bunch of people to go to suboptimal restaurants for relatively little money (the difference in the reward for the rating you would have given the best restaurant and the rating you gave for the one they sent you to). Maybe Alice owns a restaurant, and she's buying the instrument so she can send you there, even if she doesn't think you'll like it! With a prediction market, Bob can continue to trade on the conditional market for you going to Alice's recommendation for as long as he wants to make money (forever), bringing down your predicted rating and causing you to go somewhere else. Alice would need to accept arbitrarily high losses to actually get you to go to the wrong restaurant.