Keep on the Shadowfell is balanced for parties of 5 players

The default party size for 4e is 5 players, and all the official modules are designed to be an appropriate challenge for a party of 5. This is not to say that the game won't work well with 3 players (my experience has been that it starts having trouble when you have 2 or less players or 8+ players), just that the amount of monsters you'll find in each fight are intended to be a typical challenge for a group of 5 players, rather than 3.

Keep on the Shadowfell is admittedly pretty poorly balanced, since it's the very first module ever produced for 4e, back before everything was really thoroughly playtested, so how close to ideal it is for a party of 5 is debatable; nonetheless, it was designed for 5. I highly recommend checking out the rewrite of KotS that Brian Ballsun-Stanton linked.

Regardless of whether you stick with the original module or switch to the rewrite, you're still looking at something built for 5 players, so the real question is how do you deal with that? I've got 3 possible approaches.

1. You can rebalance things for a party of 3 by removing approximately 40% of the XP worth of monsters for each fight; not 40% of the monsters, but 40% of the fight's total XP value worth of monsters.

2. If your players are happy with the current difficulty level, you can always keep going as things are. Later on your players may need to seek advice (the official 4e Character Optimization forum would be a good place) on how to keep their characters sufficiently capable.

3. You can compromise between approaches 1 & 2 by reducing the XP total for fights by 20%; this will balance things for 4 players, which should still be fairly challenging for a party of 3 but will also reduce the "almost dying every time" factor.

You Should Be Dead, But...

Save-or-die mechanics are pretty awful for straight-up challenges. I mean, you wouldn't exactly get a lot of tactical thrills from a game that boils down to "Flip a coin to see if you lose," would you? But that's not the only way they've been used.

Practices and opinions vary pretty widely in the OD&D/OSR community, but one way of looking at "save-or-die" is that the saving throw is the "second chance" mechanic.

Proponents of this approach say that falling into lava or getting stabbed through the chest ought to be lethal, so really the game is about avoiding those things altogether, not surviving them if they do come to pass. And the saving throw is there to decrease character death without removing the actual threat.

What this means in play and adventure design, though, is that you can't make save-or-die situations a thing that just happens to the PCs. Rather, you have to telegraph the threat and the PCs' main goal is avoiding it altogether.

For example, if the PCs are entering the lair of a gorgon ("medusa" in D&D terms), they'll know it from the crazy-looking 'statues' all around. The challenge is sneaking past the gorgon, or fighting her without looking at her (probably using a trick of some sort), or even negotiating with the clever and cunning monster for safe passage. If they're having to roll saves to avoid petrification at all, it means they're screwed up the actual plan badly.

Unfortunately, this approach doesn't really work if you want an adventure to involve a series of challenges the PCs are mostly expected to face head-on — because that's your idea of heroism in the story, because you want to play some thrilling tactical battles, &c. What happens then is the players will be rolling those saves not as a failure consequence but as a result of engaging the scenario at all, and of course some will fail and die kinda out of nowhere. That's one of the reasons 4th Edition D&D in particular removed the "save-or-die" angle from the game.

Adding More Second Chances

So, you don't want characters to die all the time, but you want them to feel like their lives are constantly at risk (which is a bit of a contradiction, yes, and it's good to recognize that).

Well, if you want to maintain the "threat" of death, I recommend trying to lessen its occurrence, not its impact. Dependable resurrection mechanics essentially redefine "death" to "XP/loot penalty" or "sidequest" (it's worth noting that some older D&D editions had "system shock" rolls to keep resurrection from being a sure thing). If you want death to be scary, I think it should have some finality to it. Focus on giving players a way to narrowly cheat death rather than a way to straight-up undo it.

One way to do it is to bolt on an explicit death-cheating mechanic. Some established patterns include:

• One approach that feels rather "old-school" is the "death and dismemberment" table. Make a random chart of nasty things that happen to you. Death is on there. So is other stuff, though. The idea is to replace death with "a chance of death or maybe you just get screwed up some other way." Here's an example with a variety of brutal but non-lethal outcomes.

• Many games try to balance gritty combat with survivable heroes using limited metagame currency for avoiding failure consequences, in the style of WFRP's Fate Points. If you can straight-up rewrite the outcome with a point, then they're kinda like 'extra lives.' If you want to make it less of a sure thing, have the points give you a reroll instead.

  I recommend using a pure metagame resource instead of something in the fiction (like resurrection scrolls or whatever) because I think creating fictional elements that allow you to defy death necessarily draws a lot of attention to those elements, and invite the PCs to go messing about trying to figure out how to 'game' the system (e.g. score more resurrection scrolls so they can't run out).

Save-Or-Die and Converting Between Editions

Another thing to note is that the different D&Ds have different save mechanics.

• OD&D and AD&D (and many OSR games, likely) have a chart with fixed saving throw numbers. Most effects just trigger a save on the chart. This means that, as you level, you'll consistently get better at actually making those saves.
• D&D3 and D&D4 (and most other D20 games) have rising modifiers, but you're rolling against DCs that scale with the level of the challenge. Thus, characters can end up falling behind in their saves (especially any "weak" ones) as they level.

Be mindful of this when converting: mid-level characters in OD&D or an OSR game might actually be way, way better at making their saves than equivalent characters in a D20 game.