I want a patient's record to be right at the registration desk and the bedside, before any report or insurance claim depends on it.
The first data set I was trusted with was a household survey in 2000, in villages north of Kochi. I had completed a master's degree in statistics at Ponnezhuthu Institute of Applied Statistics that May, after a bachelor's degree at Mazhavilkara College. At Ithalmara Field Research, a small survey company, I checked questionnaires as they came back from field investigators. One investigator's forms were too tidy: ages ending in zero or five, and a working tap in every house. He had not invented anything deliberately. He had rounded what respondents were unsure of and filled gaps with what seemed likely. I learnt then that data quality is mostly decided by the person holding the pen, and that correcting a table later never recovers what was not asked at the time.
Most errors in hospital data begin as small, reasonable shortcuts: a name typed the way the patient happened to say it, a note written for the next shift rather than the next reader, a consent box ticked to keep a queue moving. I would rather correct those moments with the people who own them than build a cleaner copy further downstream and call the problem solved.