The following is from Ray Falk, from OptInference LLC, who participated in the opening workshop for SAMSI’s Data-Driven Decisions in Healthcare program.
I found the kick-off workshop, Data-Driven Decisions in Healthcare, tightly focused on methodologically tractable projects with ‘shovel-ready’ data sources, primarily in the areas of operational patient, provider, and (material) resource allocation and scheduling and patient-centered (personalized) effectiveness assessment.
Inference, analysis and control of service systems were developed via extension of existing results and observations from call center management (Avi Mandelbaum), including stabilization of performance under time-varying conditions (Yunan Liu). Prodigiously comprehensive tracking of patient flows through the ICU (Carri Chan) and the remaining non-intensive departments of a hospital (Jim Dai) were related to specific insights regarding options for intervention (including proportional and absolute threshold effects), both for providers and patients. An interesting discussion addressed dependence of service times on (new vs. readmitted) patient status (Guodong Pang).
Personalized medicine was addressed from several perspectives, including heterogeneous treatment effects (with a heuristic re-sampling based approaches from Bob Obenchain and Sheldon Jacobson), appropriately targeted decision analysis based on high-dimensional modeling from large populations (centered on ROC curves from validation data sequestered from model development), and availability of comprehensive data, as well as methodologies for data acquisition and alignment, inference, and modeling(OMOP) (David Madigan, Patrick Ryan) . Assessment and prediction from retrospective observational studies and tracking data was addressed via meta-analysis, propensity matching or weighting, and spontaneous event reporting (with no identifiable population at risk) (Alan Menius)
Guidance and policy for the national initiative in patient-centered evidence-based assessment of medical effectiveness was addressed via standards elaborated via PCORI (Constantine Gatsonis, Sally Morton). Issues and perspectives included:
Value of Information for prioritization, Efficiency of adaptive experimental design, individually heterogeneous treatment effects, observational vs. prospective randomized trials, data infrastructure (storage of electronic records and ontologies for integrated access), and technology assessment and dissemination, hyper-inclusive access to results for review (beyond standard publications in English). An auxiliary resource is the Society for Medical Decision Making (Turgay Ayer).
– Ray Falk