Periodic Reporting for period 1 - SPEM (Semi-Parametric Econometric Models: Health, Obesity and Patient Expenditures)
Reporting period: 2018-01-01 to 2019-12-31
Econometric modelling of healthcare costs serves many purposes: to obtain key parameters in economic evaluation of new technologies based on their cost-effectiveness; to implement risk adjustment in public and private insurance systems; and to examine the health care costs attributable to risk factors such as smoking and obesity. Modelling healthcare costs is challenging because the distribution of cost data are typically non-normal, with heavy tails and a highly skewed distribution. It becomes even more challenging when the relationship between key predictors and costs is complex. Econometric methods that can accommodate these features are scarce and often strong assumptions need to be made to facilitate the data analysis. This may lead to biased results.
Why is it important for society?
To model the relationship between key predictors and costs (as well as other health outcome variables) reliably is crucial for policy making in areas such as economic evaluation and risk adjustment. Biased or imprecise estimates can lead to misleading evidence and hence incorrect decisions and subsequently loss of social welfare.
What are the overall objectives?
The primary objective of SPEM is to develop semi-parametric methods that build upon Generalised Linear Models (GLMs). GLMs are the benchmark approach for analysing medical costs and many other types of health data. SPEM will apply the new methods to produce accurate and robust estimates of the relationship between childhood obesity and healthcare costs, which are influential in the design and evaluation of government programmes aimed at treating and preventing childhood obesity.
1.2.1 WP1: Methodology development.
Several meetings were organised at the start of the project with supervisor Professor Andrew Jones and advisory panel members. Two research visits to U of Bath were organised to learn the method of sieves, a new methodology used in the project, from Dr Bin Peng. I also sat in two courses related to the project: Health Economics and Evaluation of Health Policy.
During this phase, my work was focused on learning and developing semi- and non-parametric methods for analysing healthcare data at patient level. Outcomes of this phase resulted in two research papers. One explored a large number of econometric models including GLMs and finite mixture models to analyse health-related quality of life. It has been published at Quality of Life Research. The other used the latent class method (or finite mixture model) to analyse patient choice data. The manuscript has been uploaded to the website. Details on completion of the seven tasks are provided in Section 1.1 of the Technical Report.
1.2.2 WP 2: Application of the new methods to investigate how childhood obesity impacts on healthcare costs.
A review of the literature on the relationship between childhood obesity and healthcare costs was undertaken using the snowball/citation search method (based on key papers). In summary, the results on the excess costs of childhood obesity are mixed. It was also found that the literature is dominated by collapsing the BMI z-score into discrete categories, as in Au (2012), which may lose substantial information. This gives our new method an advantage as it models BMI z-score directly. The initial results of the application were presented at a HEDG seminar at U of York for feedback. The seminar also covered the knowledge transfer regarding the LSAC survey and its linkage with Medicare data. A working paper using the new method to explore the association between childhood BMI z-score and healthcare costs has been completed and published as a HEDG working paper.
1.2.3 WP 3: Dissemination of findings.
A summary of the research findings was published on the project website. Related codes were also uploaded for public use. Andrew Jones organised the 28th European Workshop on Econometrics and Health Economics (Leuven, September 2019) which included papers on new econometric methods for analysing health data. I presented the project results and Dr Eugenio Zucchelli was the discussant. A HEDG research away day workshop was organised focusing on dissemination of the results of the project and knowledge transfer of health economics and health econometrics among health economics researchers based at U of York. Project results were also presented at various situations and for different types of audience. One paper from the project has already been published at a peer-reviewed journal. The main output has been published as a HEDG working paper and submitted to Journal of Applied Econometrics. The other paper will be submitted to a peer-reviewed journal in 2020.
I was also invited to visit the Health Technology Assessment research group at Shanghai Health Development Research Centre, under the Ministry of Health of Shanghai Government. This group is currently developing the cost-effectiveness threshold for HTA in China which requires the estimation of the impact of healthcare expenditure per capita on health outcome, which is a perfect opportunity to apply the new methodology developed in this project. I have been invited to join this important project and I will also be appointed as an external advisor for the Centre.
The SPEM project has provided a platform for my work to be recognised by the research community. I was recently appointed as an editorial board member for Journal of Hospital Management and Health Policy, a rising journal in the area of health policy research:
I will make use of this opportunity to further promote the development and application of advanced econometric methods for more informed and accurate policy making, the core message of SPEM.