• The development of interventions/programs should not be based on people's assumptions, but rather on an analysis of gaps in care.
• Random sampling of charts (and not convenience sampling) is the best way to go; otherwise you might select only the best patients or the best results (for example, those patients who are going through your program).
• Data collection has to be done by someone who is qualified (medical records) and with standardized, blind, data collection forms, with no physician or patient identification (remember, you are not trying to evaluate individual practices here).
• There is always a risk of missing relevant information: the data you are collecting is often coming from administrative databases (that could have been coded incorrectly) or medical charts (where some information may have not been recorded).

 

Do not assume that the best practices are in place in your institution, or that you know all about them; you need to do a formal assessment "Gaps in care".