Statistics are almost useless for analysis. There are many subcategories, so most people trying to parse out meaningful data will adjust or compensate trying for greater relevance. Unfortunately, that means that assumptions determine the outcome.

For example, if we look at all the head injuries due to simple falls, we find a large percentage occur among the elderly. Not because the elderly are more prone to head injury, but simply because the elderly tend to fall more often for a variety of reasons. So, we might adjust the numbers to remove all TBIs occurring in nursing homes and hospitals and find that the picture has changed significantly.

We might likewise look at TBI data among bicyclists, and see if there are age group disparities, ie. children or adult over 65. Also, it's hard to get realistic data on actual usage, so even if we accept the raw numbers, we can't compute a rate.

So, while we do know that some bicycle crashes lead to TBIs, we don't have a decent sense of the likelihood, during anyone's lifetime cycling career, or what kinds of riding, ie. urban commuting, sport road riding, are more or less likely to lead to a TBI.

So, IMO, all this debate about the data is pointless, since the numbers can be tweaked any way someone wants.

On the flip side, even though the large percentage of TBIs from simple falls, isn't on point for discussing TBI risk when bicycling, it is relevant from a personal risk assessment. There is a risk, and we can approach it from the point of view of reducing every risk, or frame the relative risk within our other daily risks. Do you accept living near a nuclear reactor, diet/heart disease or cancer, other sports activities, drive a sports car, home with steep stairs, etc.

This is how I made my personal decision about helmets, approaching the decision based on what level of risk I consider acceptable, combined with my expectations/belief about the degree to which a helmet might make a difference.