Work package 5

Cost-effectiveness analysis of behavioural programs and public campaigns

Work Package leaders: Prof. Joanna Coast, University of Birmingham,
Prof. Richard Smith, London School of Hygiene & Tropical Medicine              

Objectives
To generate economic evidence to inform policy and practice about tools to change the behaviour of the general public and professionals with regard to unnecessary antibiotic use.  Specifically:

  1. To review the literature associated with the costs and cost-effectiveness of tools for behavioural changes of the general public and professionals aimed at containing prescription and use of antibiotics in the community
  2. To estimate the economic impact of tools to effect behavioural change in the public and professionals in relation to the impact on current and future use of antibiotics using a decision-analytic model using data from the literature review and other secondary sources.
     

Description of work
The following investigations relevant to the specific objectives have been conducted:

1.  Review of literature on costs and cost-effectiveness
The literature on the costs and cost-effectiveness of tools to effect behavioural change with regard to antibiotic use has been systematically reviewed.  Search for literature was based both on consultation with relevant experts, including those involved in GRACE, and electronic bibliographic database search. A matrix of search terms has been developed for use in searching the major electronic databases. These included, for example, antibiotic, resistance, education, prescription, control, prevention economic and cost.  Databases searches included Medline, ISI, EMBASE, Grey Literature, The York Databases, OPA and the Cochrane Library Online.  Abstracts were downloaded for initial review by at least two reviewers and categorised according to whether they were full economic evaluations or cost studies.  Standardised forms were used to extract data.  These had been developed for previous work and were derived from quantitative checklists already in existence and from standard economic criteria.  The validity of each paper was assessed using the ‘risk of bias criteria’ suggested in the Cochrane Handbook.  The reliability of the data review process was monitored, with the first ten studies being independently assessed by all those involved in the project to ensure that the data extraction sheets were suitable.  Random samples of additional studies were then subject to an inter-rater reliability check. 
 

2.  Modelling the economic impact of tools to effect behavioural change
In collaboration with the other work packages a set of intervention tools covering both public campaigns and behavioural interventions for professionals have been identified.  Information about the effectiveness of these tools was derived primarily from the literature reviews undertaken in work packages 2, 3 and 4. Information about the costs of these tools was derived from the literature review undertaken here where possible, but was supplemented by direct collection of cost data from agencies that had undertaken in the past, or were at that moment undertaking, such campaigns, as well as other cost information from existing sources, such as government official statistics.  Important categories of cost were likely to include staff time for training, printing of materials, media costs and communication costs, as well as prescription and drug costs, etc.  This cost and effectiveness information has been combined within a decision-analytic model.  The summary statistic for each intervention was the incremental cost-effectiveness ratio (ICER).  Given the level of uncertainty in this area, and the need for modelling this, an estimate of the level of uncertainty was assessed through re-sampling the data using a non parametric bootstrap approach.  Here the data for each ICER were resampled 1,000 times to generate a mean ICER, with uncertainty around this mean represented by the cost-effectiveness acceptability curve (CEAC).  This stated the proportion of re-sampled datasets for which the ICER was below a given threshold, and was interpreted as the probability that the ICER of the intervention was below that threshold (the ICER thus provided the probability of cost-effectiveness).

Cost-effectiveness analysis of behavioural programs and public campaigns

Work Package leaders: Prof. Joanna Coast, University of Birmingham,
Prof. Richard Smith, London School of Hygiene & Tropical Medicine              

Objectives
To generate economic evidence to inform policy and practice about tools to change the behaviour of the general public and professionals with regard to unnecessary antibiotic use.  Specifically:

  1. To review the literature associated with the costs and cost-effectiveness of tools for behavioural changes of the general public and professionals aimed at containing prescription and use of antibiotics in the community
  2. To estimate the economic impact of tools to effect behavioural change in the public and professionals in relation to the impact on current and future use of antibiotics using a decision-analytic model using data from the literature review and other secondary sources.
     

Description of work
The following investigations relevant to the specific objectives have been conducted:

1.  Review of literature on costs and cost-effectiveness
The literature on the costs and cost-effectiveness of tools to effect behavioural change with regard to antibiotic use has been systematically reviewed.  Search for literature was based both on consultation with relevant experts, including those involved in GRACE, and electronic bibliographic database search. A matrix of search terms has been developed for use in searching the major electronic databases. These included, for example, antibiotic, resistance, education, prescription, control, prevention economic and cost.  Databases searches included Medline, ISI, EMBASE, Grey Literature, The York Databases, OPA and the Cochrane Library Online.  Abstracts were downloaded for initial review by at least two reviewers and categorised according to whether they were full economic evaluations or cost studies.  Standardised forms were used to extract data.  These had been developed for previous work and were derived from quantitative checklists already in existence and from standard economic criteria.  The validity of each paper was assessed using the ‘risk of bias criteria’ suggested in the Cochrane Handbook.  The reliability of the data review process was monitored, with the first ten studies being independently assessed by all those involved in the project to ensure that the data extraction sheets were suitable.  Random samples of additional studies were then subject to an inter-rater reliability check. 
 

2.  Modelling the economic impact of tools to effect behavioural change
In collaboration with the other work packages a set of intervention tools covering both public campaigns and behavioural interventions for professionals have been identified.  Information about the effectiveness of these tools was derived primarily from the literature reviews undertaken in work packages 2, 3 and 4. Information about the costs of these tools was derived from the literature review undertaken here where possible, but was supplemented by direct collection of cost data from agencies that had undertaken in the past, or were at that moment undertaking, such campaigns, as well as other cost information from existing sources, such as government official statistics.  Important categories of cost were likely to include staff time for training, printing of materials, media costs and communication costs, as well as prescription and drug costs, etc.  This cost and effectiveness information has been combined within a decision-analytic model.  The summary statistic for each intervention was the incremental cost-effectiveness ratio (ICER).  Given the level of uncertainty in this area, and the need for modelling this, an estimate of the level of uncertainty was assessed through re-sampling the data using a non parametric bootstrap approach.  Here the data for each ICER were resampled 1,000 times to generate a mean ICER, with uncertainty around this mean represented by the cost-effectiveness acceptability curve (CEAC).  This stated the proportion of re-sampled datasets for which the ICER was below a given threshold, and was interpreted as the probability that the ICER of the intervention was below that threshold (the ICER thus provided the probability of cost-effectiveness).

Richard Smith