The search for internal validity in improvement Frank Davidoff Learning Lab – 2013 IHI Forum.
-
Upload
tyrone-cross -
Category
Documents
-
view
213 -
download
0
Transcript of The search for internal validity in improvement Frank Davidoff Learning Lab – 2013 IHI Forum.
The search for internal validity in improvement
Frank DavidoffLearning Lab – 2013 IHI Forum
Improvement has a two-part mantra
• Part 1:
–All improvement involves change• Part 2:
–Not all change is improvement
Local project, part 1: make change
• “Here’s how we made (system-level) change happen…”– Identified a dysfunction in the system– Came up with an innovation (better process; change
strategy for getting there)– Implemented our strategy in local context– Used small tests of change to refine innovation– Spread and maintained changes
• (Way different from giving a pill…)• So what’s next?
Local project, part 2: find out whether change is improvement
• “Here’s how we learned whether our change was improvement…”– Chose outcomes: processes, patients’ clinical
condition– Developed outcome measures– Created informal systems for collecting,
displaying, using outcomes data (quantitative, qualitative)
– Used these data locally to study the impact of changes, modify change strategy
Yes, Virginia, there is “study” in local improvement projects
• Informal study is an inherent part of all meaningful improvement– Used to check on impact of change (“Did it
work?”)– Especially visible in “Plan-Do-Study-Act” cycles
(originally called “Plan-Do-Check-Act” cycles)– Not related to whether project is meant “for
publication”
Where have we gotten in our local project?
• In part 1: we made change happen • In part 2: we produced informal outcomes data to
demonstrate improvement– Good enough: allows project staff to modify, spread, maintain
change• BUT data quality (completeness, accuracy) is uncertain;
no control for confounders, biases – “lite” study data• Result: weak internal validity!– i.e., unlikely to convince skeptics elsewhere about
improvement
How can we strengthen the evidence?• Shift “up” to formal planning and study
– Identify plausible theory of performance change– Adopt specific study design– Select/define relevant outcomes– Develop reliable data collection process, robust data
quality control– Analyze results (e.g., grounded theory; statistics; time
series, esp. statistical process control)• Creates “research level” data• Result: stronger internal validity!
– i.e., more likely to convince editors, peer reviewers, rest of the world – that our change was really an improvement
Summing up
• Making change locally includes informal study of outcomes– Useful: makes project possible – but data
somewhat “fuzzy”– Result: Internal validity is weak
• Formal study of change process and outcome requires “research level” methods– Scholarly: contributes to general knowledge– Not every formal study feature is required, but the
more features the better– Result: internal validity is stronger
Internal validity in improvement studies:key reference
• Solberg L, et al. The three faces of performance measurement: improvement, accountability, and research. Joint Comm J Qual Improvement 1997;23:135-47.