Identifying and Measuring Knowledge Spillovers of Health Service Delivery Interventions: A Network Analysis Approach

Health care policies may have effects other than those originally intended by policy makers. These unintended effects are called spillovers. For example, a financial penalty policy intended to reduce readmission rates for acute myocardial infarction, heart failure, and pneumonia in the US Medicare was associated with reductions in readmission rates for conditions that were not targeted by the policy.

Although the existence of such spillover effects is increasingly being recognised, there is a lack of understanding of how these spillover effects occur.

In health care, an important mechanism is the exchange of knowledge between health care professionals. Analyses of spillovers between health care professionals used to be very costly because they required a detailed mapping of the social network that could only be generated by research based on qualitative interviews and questionnaires. However, it has recently been found that social networks can accurately inferred from administrative data sources alone.

The aim of this research is to use these new methods to 1) construct social networks between health care professionals, 2) to describe the networks in terms of how closely related the network members are, and in terms of the characteristics of the members, 3) to measure the size of spillovers in two examples, and determine how the shape and size of the network and the relations between its members affect spillover effects.

The existence of spillovers can mean that the effect of health care policies is measured incorrectly. The last part of the project will consider how information about networks between health care professionals can be used to measure the effect of new policies more accurately.

Health care policies may have spillover effects beyond their intended objectives. For example, a financial penalty policy intended to reduce readmission rates for acute myocardial infarction, heart failure, and pneumonia in the US Medicare was associated with reductions in readmission rates for conditions that were not targeted by the policy, and a payment incentive scheme to improve quality of care in the North West of England was associated with improvements in the quality of care for patient groups that were not targeted by the scheme.

Although the existence of such spillover effects is increasingly being recognised, there is a lack of understanding of the mechanisms by which such spillover effects occur, and how spillovers can be measured and identified empirically.

This proposal takes a novel approach to the identification and measurement of knowledge spillovers by integrating the economic analysis of knowledge spillovers, social network analysis and state of the art methods for social network mapping that have made it possible to map social networks using administrative data sources only without the need for a costly qualitative data collection.

The proposed research develops a methodology for identifying and measuring knowledge spillovers of health service delivery and policy interventions at the level of the health care professional. Using two recurrent empirical examples of quality improvement incentive schemes, the project 1) Characterises intra- and inter-organisational networks of health care professionals using administrative data to identify testable mechanisms of knowledge spillovers. 2) Tests these mechanisms by measuring spillover effects and estimating the determinants of knowledge spillovers in social networks of health care professionals, and 3) Analyses the implications of knowledge spillovers for programme evaluation of long term effects

1. Who will benefit from this research?

Health Economists and civil servants carrying out policy evaluations of interventions aimed at changing health service organisation and delivery

Social Network Analyst working on knowledge diffusion with access to large administrative data sets

Academics, healthcare professionals, and policy makers designing new interventions to health care policy. This includes service providers in the private and third sector.

Decision-makers evaluating new evidence on the impact of proposed changes to health care policy

The wider public

2. How will they benefit from this research?

Health Economists and civil servants
Failing to account for spillover effects can lead to biased effect estimates when evaluating health policy reform. The project shows how the size and mechanisms of knowledge spillovers can be estimated leading to more accurate effect estimates. In addition, the research will show how contaminated control groups be avoided at the design stage, by considering how knowledge diffuses through social networks.

Social Network Analysts
Social Network analysts often rely on costly data collection to construct and characterise social networks. This project will provide further evidence on how administrative data can be used to construct social networks for knowledge spillovers and can be used to inform and inspire future research in this and related areas where large administrative data sources are available.

Designers of new health care policy
In this context, spillovers are defined as unplanned effects of health policy reform. Designers of new policies will benefit from awareness of how knwoledge spillovers might occur between health care professionals. Such understanding can lead to deliberate design of reform that generate spillovers onto other areas, thus providing more value for the same money, or can be considered to help avoid spillovers onto areas where they may be undesireable.

The wider public
Will benefit from more accurate estimates of the effects of policy reform, and health care policies that make better use of scarce resources by considering spillover effects at the design stage.

The PI will benefit from being enabled to further his transition towards independence and help him develop his research group. The staff working on the project will develop skills and understanding in problem structuring, econometric analysis for programme evaluation, academic writing and other professional skills such as networking and presentation skill. All these skills are transferable to other sectors, such as manufacturing, consulting and governance.