Realistically, a Constitution score below 14 is nearly suicidal on its own in Pathfinder. Characters with large HD tend to be on the front lines, and so can’t afford to miss out on any hp they can get, and characters with small HD desperately need to get hp somewhere, and as a result everyone is basically forced to sink a 14 (or better) in Constitution if they want to stay alive.
This, of course, wreaks havoc with the rest of the game because many, many characters really need that 14 in another score. Monks are the most notorious for this (needing Strength, Dexterity, Constitution, and Wisdom), but really anyone who needs more than one non-Constitution abilty score is in trouble. So a barbarian who only needs Strength and Constitution does OK, but a fighter who needs those and a 13 in Intelligence for Combat Expertise is in trouble, and a paladin who needs those and Charisma for all their supernatural features (and really wants better than a 13) is in a lot of trouble.
Even when rolling scores and getting very lucky, at higher levels you still have a problem because of the exorbitant costs of magically enhancing all of those ability scores. Here, even the barbarian is pretty badly off because of Paizo’s inane insistence that all physical-ability-enhancing magic items must all be belts, and thus incur a 50% surcharge when you want to enhance more than one of them, starts to really shaft them.
So realistically the only classes that are really OK as far as abilities are concerned are the pure spellcasters, who need one mental stat and Constitution, and that’s it. That’s a big part of why magic dominates the game.
All of which is to say that, quite simply, Pathfinder is not an even remotely balanced game. It full of tons and tons of options that are written as though they were equally strong, equally valid, but are in fact wildly variable in their effectiveness. Choosing to have “average” Constitution, i.e. 10, is treated as just as reasonable as having “average” Charisma—except that the former choice is suicide and the latter choice is nearly meaningless unless you’re a Charisma-based spellcaster. Likewise, all the classes are presented as equal, when they are massively not.
So what can you do about it?
Really, and I say this as someone who plays a lot of Pathfinder (and D&D 3.5e, which it’s based on and from which it inherits many—though not all—of these problems), the best answer is probably to find a system that isn’t so ridiculously imbalanced. This is not a system where you can just pick up and play, making decisions based on what sounds cool or appropriate and expecting the system to make that work.
Which means that, if you stick with Pathfinder, the only workable solution I have found is for everybody to simply get to know the system really well, and learn how to avoid all the traps. I can point out a few of the biggest ones for you, but ultimately what you really need is expertise, which I can’t just transfer to you.
Everyone needs Con 14 or better
This is just something your players have to learn, and plan around. Someone with Constitution 10 isn’t average, they’re suicidally low. Someone with Constitution 14 isn’t moderately strong in this stat; they’re pretty much average.
Consider houserules to make this easier, if you like. If using point buy, having Constitution start at 14 rather than 10 might be a start—going above 14 still costs more, and you can’t get above 18 at all, but you ensure that everyone has Constitution 14 and still has points for the other things their classes rely on.
Alternatively, you could consider maximizing HD each level, instead of rolling or using the average. Con is still pretty important for Fortitude, but lower Con values are less inherently suicidal. It also eliminates one of the ways in which the order of levels matters, which is always a good thing.
Avoid 1st level altogether—probably 2nd too
Pathfinder at 1st level is utterly unlike Pathfinder at any other level of the game. Commonly referred to as “rusty dagger shanktown,” this is a problem found, actually, across many editions of D&D and throughout many D&D spin-offs like Pathfinder. The issue is that a whole lot of values in the game are linear, with little or no initial offset. That means those values are literally doubled by reaching 2nd level, and by 3rd level they’ll have been tripled (relative to 1st), and so on. Effectively, the math works out such that 1st level characters simply haven’t accumulated enough (hp, save bonuses, AC, gold, etc. etc.) to survive.
The solution chosen by a huge number of tables is to just skip 1st level. Skipping 2nd is pretty popular too. An enormous number of games start at 3rd level for exactly this reason. This avoids a lot of the worst problems, at least for a while, though really it starts getting bad again as early as 7th level—there’s a reason why E6 is popular.
Class choice needs to be careful
Monk is simply a bad class, for example. Any given “monk” themed character can be far more effectively brought to life using another class’s rules. For instance, there’s the unchained version, which is pretty solid, though there are still better options. This theme continues for tons of other classes as well.
Meanwhile, some classes are just fairly unreasonable from the outset. Pretty much every spellcaster that can get 9th-level spells becomes a problem sooner rather than later—spells just get so absurd. They are, themselves, a big part of the reason why E6 stops at 6th, because at 7th level they get 4th-level spells and those are insane.
Challenging a party that mixes classes of these two wildly differing power levels becomes extremely difficult for a GM—anything the former can reasonably handle could be a cakewalk for the latter, and anything actually challenging for the latter can be impossible for the former. (Playing a spellcaster well can be challenging too, so this isn’t entirely automatic, but playing a warrior well has its own challenges and ultimately they become much harder as levels increase.)
This is why the tier system exists. This method of roughly categorizing the relative power and versatility of Pathfinder classes allows players and GMs to find a way to make their characters more equitable, to avoid some of those headaches. Maybe cleric and wizard should be avoided, or at least the people playing them should intentionally choose some very weak options. Maybe you should skip monk or rogue, and go with something more powerful that can have similar flavor—or go with the unchained versions, which have exactly the same flavor.
Maybe those who are dead-set on a certain weaker class should do some research, and find ways of making that class as powerful as it can be. The kineticist is borderline unplayable if played the way it’s described, but with certain very specific choices, it can be OK. Not every class can be “salvaged” like that, but you can definitely improve matters in this fashion.
Again, E6 helps matters here.
The other option is to just have everyone stay in, say, Tier 3 or so. I have had very successful games where everyone picks one of the “two-thirds” casters, like magus or warpriest. That’s very effective! Those classes hit a very nice sweet spot in the system, where their spells help ensure they’ll have options for handling things that come their way, but the delays on them make sure you aren’t dealing with the real game-breakers, or at least not until later.
The alternative, ish
The other option is the one used by Paizo in modules: cater to the lowest common denominator. Most modules are filled with challenges well below the capabilities of fairly typical adventurers of that level—to make sure that the monk or whatever actually can contribute. For very casual players—that don’t happen to stumble upon the strongest options—this can work out alright. But many players, and GMs, get pretty bored with this.