1 / 17

Why Local Investigations?

Design of Local Investigations in Community Practice Settings: An Effectiveness / Cost-Effectiveness Framework with Finite Patient Horizon Naihua Duan, Ph.D. Columbia University Joint work with Ken Cheung (Columbia University) and Jeff Cully (Houston VA HSRD Center). Why Local Investigations?.

ravi
Download Presentation

Why Local Investigations?

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Design of Local Investigations in Community Practice Settings: An Effectiveness / Cost-Effectiveness Framework with Finite Patient HorizonNaihua Duan, Ph.D.Columbia UniversityJoint work with Ken Cheung (Columbia University) and Jeff Cully (Houston VA HSRD Center) Finite Patient Horizon NYSPI

  2. Why Local Investigations? • Think globally, act locally • Quality improvement programs in community practice settings often need to adapt and/or improvise in order to accommodate local conditions • Impact of adaption is usually unknown • Decision-making often based on expert judgment • Empirical evaluation might be warranted, but not widely used • “Small n problem” (Small N, not n) Finite Patient Horizon NYSPI

  3. External Program Evaluation vs. Internal Quality Improvement Evaluation • Program evaluation usually takes an external perspective: does the implementation program work? • Does the engine work overall? • Internal evaluation for nuts and bolts might be warranted for quality improvement • How do we do intake calls, case referrals, follow-up reminders, care management (maybe IT enhanced), etc., most effectively for a specific clinical setting? • One size might not fit all (HTE; Kravitz-Duan 2004) Finite Patient Horizon NYSPI

  4. Statistical Literature • Statistical literature focused mainly on global investigations, not on local investigations • Zelen (1969) an unusual exception • “New” statistical perspective warranted for local investigations in implementation studies • Global knowledge for external consumption and local knowledge for internal consumption call for distinct statistical frameworks Finite Patient Horizon NYSPI

  5. Patient Horizon (Population Size) • Essentially infinite for usual clinical studies aimed to produce generalizable knowledge for external consumption • Large Phase III trial, sample size << N • Strong accuracy (small , high power) warranted to protect welfare of future patients • Can be very limited (hundreds, even fewer) for local investigations aimed to produce local knowledge for internal consumption Finite Patient Horizon NYSPI

  6. Hypothetical Example • N = 500 patients • Randomize 2n patients to novel procedure vs. standard procedure (50/50) • Apply local knowledge gained to remaining patients (N – 2 n) • d = 0.25,  = 5%, 80% power ==> n = 250 • No patients left to consume knowledge gained! Finite Patient Horizon NYSPI

  7. Cost Effectiveness Framework • Criterion: incremental net benefit per capita • INB = b  – c •  = effect size, say, symptom reduction resulting from new procedure (unknown) • b = value for each unit of treatment effect (known) • c = incremental cost for novel procedure vs. standard procedure (known) • Effectiveness framework if c = 0 Finite Patient Horizon NYSPI

  8. Cost Effectiveness Framework (II) • Bayesian framework to incorporate existing knowledge among local experts •  ~ N(0, 2) •  measures local experts’ uncertainty • Cost-effectiveness equipoise: • b 0 = c • Effectiveness equipoise: 0 = 0 (c = 0) Finite Patient Horizon NYSPI

  9. Non-Empirical Adoption Strategies • Based adoption decision on expert opinion, without empirical local investigation • Never adopt strategy (bench mark) • Always adopt strategy • Expected gain compared to never adopt: • G1 = N × (b 0 – c) Finite Patient Horizon NYSPI

  10. Evidence-Based Adoption Strategy • Randomize 2n patients to trial (evaluation phase) • One-sided test H0: INB  0 vs. HA: INB > 0 with significance threshold  • Apply finding to remaining N – 2n patients (consumption phase) • Each combination of design parameters (n, ) is a possible design for the local investigation • Find the design that maximizes expected net gain across N patients • Compare with non-empirical strategies Finite Patient Horizon NYSPI

  11. Evidence-Based Adoption Strategy • Expected gain: G2 = {n + (N – 2n) A} (b 0 – c) + (N – 2n) B A  (0,1); B > 0 • Compared to non-empirical strategies: G0 = 0 G1 = N × (b 0 – c) Finite Patient Horizon NYSPI

  12. Optimal Design for Local Investigation Finite Patient Horizon NYSPI

  13. Finite Patient Horizon NYSPI

  14. Noteworthy Results • Under equipoise: • Any local investigation is better than non-empirical strategies • Optimal  = 50% • Optimal n = N / {3 + sqrt(9 + 4 R)}  N / 6 • R = N 2 / (2 2) • Under mild optimism: optimal  > 50% • Under strong optimism: forthright adoption without local investigation Finite Patient Horizon NYSPI

  15. Finite Patient Horizon NYSPI

  16. Discussions • More emphasis on local investigations might enhance quality improvement programs, leading to improved patient safety, quality of care, and health outcomes • More emphasis on local investigations might help transform current top-down organization for health care knowledge production, and facilitate the development of learning communities, and motivate more comprehensive data acquisition Finite Patient Horizon NYSPI

  17. Discussions (II) • Appropriate statistical methodologies (not the 5%-80% ritual) can facilitate wider use of local investigations in quality improvement studies • Human subjects and publication issues Finite Patient Horizon NYSPI

More Related