In the simplest terms, you can just use PixelMath with an expression like this:
Ha / 2 + OIII / 2
That will give you a monochrome image with data from both filters. The issue that you'll run into is that the relative contributions of the two bands are probably going to vary wildly, based on the target object. In some cases, the Ha will dominate and the OIII will be almost invisible. And in some cases, it will be reversed. There are two ways that you can address this issue. First (and simplest), you could just do a LinearFit on the dimmer of the two images, using the brighter of the two as the reference. That will ensure that both images contribute more or less equally.
The issue is that you may not want equal contribution. In that case, there are various ways to use PixelMath to adjust the contribution, ranging from really simple to pretty complicated. The simplest way to just to play around with the "divided by 2" parts of the expression above.
Personally, I generally combine the channels into an RGB image with PixelMath. I have a few expressions that equalize the background sky level and adjust the slope of the dimmer image to get the level of contribution that I want (I get the values for my expressions using previews of interesting areas of the uncombined images). Once I have the color image balanced to taste, then I extract the luminance. This results in a black and white image that has the specific contributions for each of the channels that I want.