Health economics and causal modelling in Health Services Research
Download as PPTX, PDF3 likes567 views
The document discusses the challenges in economic evaluation of health services research, highlighting issues with measuring clinical outcomes and the need for causal modeling. It outlines a three-step process for causal modeling, which includes building a qualitative model, populating it with quantitative data and expert belief, and estimating intervention effectiveness using a Bayesian approach. Examples illustrate the impact of consultant presence in healthcare on various outcomes, such as length of stay and in-hospital mortality.
Health economics and causal modelling in Health Services Research
- 1. Health economics and causal modelling in health services research 23/04/2015 Yen-Fu Chen & Sam Watson Warwick Centre for Applied Health Research & Delivery (W-CAHRD) CLAHRC WM Programme Steering Committee Meeting, 15 April 2015
- 2. Challenges in economic evaluation of health services research • Clinical outcomes: too rare to measure reliably, e.g. transfusion of incompatible blood e.g. a £10 million computer system for a hospital with 50,000 admissions/year only needs to save 2 lives per 1000 patients • Process outcomes: too diffuse to measure • Experimental evidence may be scarce • Cost effectiveness = Costs / Effects • Difficulties in measuring intervention effects Lilford et al. BMJ 2010; 341:c4413
- 3. Use of causal modelling • Overcomes difficulties in measuring specific processes or outcomes • Allows the integration of all available evidence • Three key steps: 1.Build qualitative causal model 2.Populate the model o Systematic review of quantitative data o Elicitation of expert belief 3.Estimate intervention effectiveness using Bayesian approach
- 4. Service delivery causal chain Process Generic service intervention Targeted service intervention Policy intervention Clinical intervention Structure Generic process Outcome Targeted process Clinical process Context (moderating variables) HTA Explanatory (independent) variables Dependent variables Intervening variables
- 5. Consultant presence at weekendsStructure Intervening variables / mechanisms Higher level of clinical competence Stronger leadership in case management Process Outcome More accurate diagnosis Earlier intervention Higher throughput (shorter waiting time & procedural delay, quicker discharge & shorter length of stay) Better Monitoring Better administration of intervention Better patient satisfaction Reduced errors & adverse events Enhanced learning for junior doctors Faster decision on palliative cases Reduced mortality
- 6. Consultant presence (at weekends)Structure Intervening variables / mechanisms Higher level of clinical competence Process Outcome Prompt investigation & more accurate diagnosis Earlier intervention Higher throughput (waiting time; procedural delay; length of stay) Better Monitoring Better administration of intervention Better patient satisfaction Reduced errors & adverse events Enhanced learning for junior doctors Faster decision on palliative cases Reduced mortality Stronger leadership in case management
- 7. Causal modelling • Three key steps: - Build qualitative causal model - Populate the model o Systematic review of quantitative data Quality, quantity, relevance, heterogeneity o Elicitation of expert belief - Estimate intervention effectiveness using Bayesian approach • Over to Sam
- 8. Fielding et al. 2013 (Clin Med 13;344-8) • Consultant delivered care (n=260) vs. standard care (n=150) • 16 weeks • Length of stay (median): 4 days vs 7 days • 30-day readmission: 17% vs 14% • In-hospital mortality: 3% vs 6%