At this point, SNP genotyping is pretty much obsolete for health-related uses because you can now get a full genome sequence for about $1,000 from just a few drop of saliva - well the raw data, at least - a custom interpretation for a suspected genetic condition might easily run you $20K. SNP genotyping can still be useful for detecting losses or duplications of large parts of a chromosome ("structural" variations) - but mainly because the analysis software is more mature.
I am aware of the difference between the cost of data and the cost of healthcare resulting from said data, but the poster was explicitly asking about the relevance of "open source" raw data so that he could do the interpretation himself. I am also explicitly mentioning arrays, because this is the cheapest technology and the one used by 23andme. Nevertheless, this kind of technology (and yes, the genome too!) has not made a very meaningful impact in clinical practice, with the obvious exception of clinical genetics. Now, you probably already know that today it's much easier to get a diagnosis of some rare muscular dystrophy or some weird anomaly of metabolism thanks to genome sequencing. However, with the exception of some genetic conditions (BRCA1/2, Lynch syndrome, Cystic fibrosis), clinical genetics are only useful for a small percentage of the population. That is exactly what I am saying: there is great progress for a small percentage of the population with rare, high-penetrance, devastating diseases, but there is not meaningful genetic test for type 2 diabetes, hypertension, cardiovascular disease, COPD and other frequent conditions. I already explained the reason above but you can also read more about it elsewhere (for example: http://www.nature.com/nrg/jour..., frequent diseas)
Based on the presentations at the American Society for Human Genetics (ASHG) in Baltimore last month, many of the major healthcare organization in the USA are making plans to implement genetic testing for adverse reactions to these drugs. It may be a few more years before it all actually gets rolled out universally (there are currently a number of large pilot studies ongoing) - and it may just be full genome sequencing rather than SNP genotyping - but this is coming. It's a question of "when" rather than "if".
I agree, which is why I said that this kind of application is more mature and that's why I called it an "exception". I already occasionally ask for DPYD or UGT1A1 genotyping for my patients, so certainly this is not a very remote scenario. Nevertheless, some kind of HARD data will have to justify the expense: reduction of hospitalisations, chemotherapy mortality or something similar. Just showing that universal genotyping predicts
There were quite a few people who found out through 23andMe that they were carriers for dangerous BRCA variants. 23andMe almost certainly saved some people from dying of breast cancer. Personally, I'd say that not dying of breast cancer is useful.
Well, that is an interesting byproduct of chance and certainly these people were lucky. However, you are certainly aware that SNP genotyping is NOT a valid test for BRCA variants for many different reasons. So, I would counter argue that some people were also falsely reassured by a negative result of a suboptimal test.
If people need to know their BRCA status, they need a proper test (=sequencing with sufficient depth at an accredited lab), a proper interpretation of variants (especially class 2-3!) and discussion of the family tree by a medical geneticist. I understand that cheap DIY healthcare is appreciated in the US, but you can't seriously consider 23andme results as sufficient in this context.