It's easy to jump to conclusions from this one single meta analysis however based on my experience I would want to see a more deep dive review. I would guess that underreporting, especially in the pre-2010 era after which electronic medical records became much more prevalent, as well as the changeover to ICD10 from ICD9, are likely culprits. Data quality in healthcare today is a D- at best, 20 years ago it was absolute and total garbage. Though this persists today, the excel 65k row limitation and excel date bugs are reponsible for an unimaginable amount of healthcare data destruction.<p>Up until 2012 even, most CDC reporting was done as manual chart reviews typed into an excel sheet. That was a very error prone process.