Perhaps there is something more professional seeming about an authorize.net style merchant account, but it's easier for us if we don't have to handle any credit card information. How much does using Google Checkout hurt us?
So if you want people to visit and order from your site, don't put any obstacles in their way. Whatever you do, don't force visitors to register."-Paul Graham, "The 10 Secrets of Selling Online: #4. Make Your Site Easy," http://store.yahoo.com/easy.html
By accepting only Google Checkout, you are adding an obstacle to everyone who does not already have a Google Checkout account, and thus discouraging casual browsers from executing a purchase. I'd personally encounter a lot less "mental friction" logging into an existing Paypal account or the like and pressing a few buttons than going to another site, starting creating an account, finding and typing in a credit-card number, enter an address, glance through the Terms of Service, etc...
Of course, depending on your product, there may not be such thing as a "casual browser" who may leave and look at a similar product merely based on the inconvenience of making an account, but it is nevertheless a large obstacle; it's just that the seeker and prize on either side have suddenly increased in strength.
Beyond that, it is hard to say exactly how much you will be hurt. A few years ago a site I was working on moved from exclusively PayPal to an on-site form, and sales increased instantly about 20% and then over 100% after tweaking of checkout process. That was back when PayPal forced you to create a PayPal account, which I believe Google does now.
In the past I have used CyberSource and VeriSign, and have recommended CyberSource in the news.yc discussions. (The PayFlow division of Verisign has been acquired by PayPal.) CyberSource has a small business option that has worked well for my businesses. They have an easy to use API, etc.
However, for this project I am trying out PayPal Pro. Not that I love PayPal or anything, but I thought I would give it a shot. (Last time I evaluated PayPal, they didn't have a merchant option.) Two things I like about them so far: 1) they don't seem to require an explicit personal guarantee, although I am not live yet and it might be implicit in the terms; and 2) they seem to have American Express integration from the get-go, as opposed to all the others where you have to sign up with them directly. I could be wrong on both of those, though.
Another thing. In your case, the difference from off-site/on-site checkout might be substantial because I imagine trust matters a lot in the medical arena. It is one thing if you are buying a T-shirt and the merchant has you leave the site to pay, but on a professional site for a serious product, you expect the site to just have that checkout process as part of their site. That is not to say you can't additionally offer Google Checkout and/or PayPal for those customers who prefer it.
Ebay's Yahoo Stores competitor ProStores seems like a reasonable alternative to Yahoo Stores, it has really cheap pricing for those with a small catalog of items. And while I don't have a problem integrating a payment processing API into my own shopping cart, it seems like it is outright cheaper to use a prepackaged off-site solution so I'd rather not do it myself.
But that it is a different thing from off-site vs on-site. I can pretty much guarantee the optimal checkout process (among the universe of checkout processes) would be an integrated solution minimizing the number of steps and abrupt transitions in the overall process. Such an optimal process might include additional options for Google Checkout, PayPal, and others, i.e. third parties for those people who want that.
What it comes down to, is that for every step, there will be a drop off %. Putting checkout on your site allows for less steps. And it allows for more possibilities in look and feel, etc. That is, it allows you to move more freely around the universe of checkout processes toward the optimal process for your particular situation.
From experience, this universe of checkout processes can have vastly different outcomes in terms of sales. By the way, I worked on another site that saw an immediate increase in sales from moving off Yahoo Store to an integrated checkout, and then another big increase when the checkout process was further optimized with, among other things, upselling and coupons right in the process.
I just don't understand the want to move this off-site when you have the capability to do it on-site. I assume the main goal of this site is to sell stuff. So I would imagine that you would at least want to experiment with one of the main things involved with selling stuff, i.e. the checkout process. You could even run experiments testing off-site vs various on-site configurations. I suppose I could see a compelling argument if it took months to create something functional. But it is probably an order of magnitude shorter.