Approved research

Mendelian Randomization analysis of the causal effect of obesity and coronary artery disease on healthcare costs

Lay summary

Accurate information on the healthcare costs associated with health conditions is required by healthcare funders. Observational studies that describe associations between health conditions and cost are prone to bias because of reverse causation, measurement error, and residual confounding. This proposal will use information on the relationship between genetic variants (single nucleotide polymorphisms) and two health conditions (obesity and coronary artery disease) to address the problems of conventional observational studies. The proposal will use genetic data linked to Hospital Episode Statistics to produce the first causal cost estimates of the effect of obesity and coronary artery disease using Mendelian Randomization methodology. The ultimate aim of the proposed research is to improve population health by increasing the quality of evidence available to decision makers responsible for the allocation of scarce healthcare resources. Evidence on the costs associated with healthcare conditions is relevant to healthcare policy evaluation, cost-effectiveness analysis, research prioritization and private sector decision-making such as in relation to the setting of healthcare insurance premiums. Moreover, a clearer appreciation of the consequences of health conditions for costs is likely to be of considerable interest to patients and to their families and carers. The effects of obesity and coronary artery disease on healthcare cost will be estimated by using data on certain types of genetic variants that are associated with these conditions. We will use data from replicated genome-wide association studies to obtain robust evidence of the associations between genetic variants and health conditions. We will obtain data on the association of variants and costs from the Biobank. Combining information on the effect of genetic variants on long-term conditions with their relationship to costs will allow important, new evidence to be generated concerning the cost impact of long term conditions. The full cohort will be eligible for inclusion in the analysis.

CURRENT SCOPE Accurate information on the healthcare costs associated with health conditions is required by healthcare funders. Observational studies that describe associations between health conditions and cost are prone to bias because of reverse causation, measurement error, and residual confounding.

This proposal will use information on the relationship between genetic variants (single nucleotide polymorphisms) and two health conditions (obesity and coronary artery disease) to address the problems of conventional observational studies. The proposal will use genetic data linked to Hospital Episode Statistics to produce the first causal cost estimates of the effect of obesity and coronary artery disease using Mendelian Randomization methodology.

EXTENDED SCOPE AGREED IN FEBRUARY 2020: We propose three additional analyses. The first will involve undertaking a genome-wide association study (GWAS) of inpatient hospital costs and rates of hospital admission, which we will create from linked records of Hospital Episode Statistics. A complementary strand of work will test the ability of any significant genome-wide significant hits from this GWAS to predict hospital costs/hospitalisation in a hold-out sample. The second analysis will apply data from Sullivan et al (https://www.ncbi.nlm.nih.gov/pubmed/21422468) to estimate quality of life and quality-adjusted life years for the Biobank cohort as a means to compare the quality of life of all individuals in Biobank, and particularly to estimate the quality of life impact and cost-effectiveness of interventions for obesity. The third and final analysis will be to conduct a GWAS of this new quality of life/quality-adjusted life years data as an outcome in its own right, and as a composite measure in conjunction with the hospital cost data.

NEW EXTENDED SCOPE: We proposed to extend previously approved analysis as follows.

We will estimate the costs associated with use of primary care from the linked primary care records.

We will use Mendelian Randomization to examine additional exposures in relation to healthcare costs, quality of life and wider outcomes already reported in UK Biobank. The proposed additional exposures are asthma, eczema, migraine, coronary heart disease, type 2 diabetes, depression, breast cancer, prostate cancer, smoking and risk tolerance.

The proposed additional outcomes to be examined are: Average household income, deprivation, current employment status, job class, degree status, owner-occupied accommodation versus renting, measures of wellbeing (reported happiness, reported loneliness), and measures of social contact (having someone to confide in weekly or more frequently versus less frequently, friend/family visits weekly or more frequently versus less frequently, cohabiting with partner or spouse versus not cohabiting, participation in any leisure/social activity versus none).

FURTHER EXTENDED SCOPE: We propose to develop measures of multimorbidity, and use this as an additional exposure to tests its causal effects using Mendelian Randomization on all of the outcomes described above, particularly healthcare costs and quality of life. 

EXTENDED SCOPE OCTOBER 2020. We propose to extend the exposures studied in relation to quality of life and healthcare cost outcomes to include breast cancer, prostate cancer, colorectal cancer, lung cancer.