TABLE 5 ] .

Potential Data Sources for Value Metrics

Data Source	Benefits	Limitations	Typical Use (Example)
Administrative	Commonly available across institutions for large groups of patients No new data collection required	Delay in coding (not real time) Limited granularity Potential for coding errors	Structure (staffing ratio12) Process (RBC transfusion58) Outcome (length of stay) Cost (cost/d4)
Manual chart abstraction	Good detail on focused areas Can translate free text from chart into more-structured data	Resource intensive per chart reviewed Limited scalability Depends on clinician documentation	Process (antibiotic administration for sepsis52) Outcome (length of stay)
EMR extraction	Larger population More discrete clinical data Potential to automate measurement, reducing ongoing resource utilization (more efficient in the long run) Potential real-time assessment	Depends on clinician documentation Initial development cost Advance system planning required (eg, variables, data repositories) Balance between discrete data and narrative	Process (antibiotic administration for sepsis52) Outcome (length of stay)
Survey	Provides data commonly not present in the medical record May be scaled depending on scope of project	Respondent burden Limited ability to integrate into other data reporting	Structure (staffing ratio12) Outcome (patient satisfaction71)
Ancillary system (eg, infection surveillance and adverse event reporting systems)	Provide information commonly not present in medical records	Challenging to integrate with other systems for data gathering and reporting purposes Initial cost to develop system	Outcome (CLABSI36) Outcome (unplanned extubation35)

EMR = electronic medical record. See Table 1 legend for expansion of other abbreviation.