OSHA have got a very prescriptive account of what is considered first aid treatment in the US (as whether treatment is medical or first aid is a criterion in their legal accident recording requirements), so if you want a comprehensive set of criteria to avoid bickering over what counts you could use that.

If your aim is to reduce your accident statistics, then all you need to do is to create obstacles to reporting and recording them, and to punish those who do.

If your aim is to reduce the actual number of accidents, then you should focus on investigating and analysing the accidents and near misses that you have had.  But I expect you know this.

To add to Kate's comment, 5 years after the BP Texas City disaster, OSHA said they were going to start coming down with a ton of bricks on companies that continued to play all the games that enable accident statistics to look "better".

However, those games continue. Saw the report by one multinational following the principles set out by the Global Reporting Inititiave that showed the nice downwards trend in OSHA recordables and hardly mentioned the disaster that had killed nearly 20 in a single incident. Due to the size of the business those fatalities didn't really impact the overall reportable incidence rate.