Studies in health economics : modelling and data analysis of costs and survival

Abstract: This dissertation consists of six essays in health economics.The first essay, “Economic evaluations in health care: Basic principles and special topics”, serves as an introduction to economic evaluations in health care, including estimations of costs, health effects, and the discount rate. Special topics of interest for the rest of the studies are also discussed, e.g. the role of modelling in cost-effectiveness analysis, and methods for dealing with incomplete observations in clinical trial data.The main theme of the second essay, “Consumption and production by age in Sweden: Basic facts and health economic implications”, is a fairly detailed compilation of consumption and production figures by age in Sweden. The purpose of this is to use the difference between consumption and production in each age group as a measure of the average costs of added years of life in the general population. In economic evaluations of health care interventions, only future costs for related ill­nesses have typically been included in the analysis. However, the health economist David Meltzer has argued that future costs for un­related illnesses and general consumption should also be in­cluded in eco­nomic evaluations. Otherwise, the analysis will not be consistent with expected utility maximiza­tion.The third essay is entitled “The possibility of predicting health care costs in the future from predicted changes in age structure and age specific mortality: The case of Sweden”. Changes in the age structure, especially the growing number of elderly people, have raised concerns about increasing costs for health and elderly care in the future. However, the number of elderly per se is not the main problem, since the growing number of elderly people is a result of better health and hence lower morta­lity. The main purpose of the study is to investigate if future health care costs can be predicted based on forecasts of future changes in age structure and mortality rates. It is shown here that at least in Sweden and in the U.S., there is a linear relationship between age-specific mortality and age-specific health care costs. When these relationships are applied retrospectively to old data, however, the predictions are underestimates of the actual costs. These results are in line with earlier studies, which show that the future age structure is not likely to have a great impact on the future health care costs.The fourth essay is called “Cost effectiveness of bisoprolol in the treatment of chronic congestive heart failure in Sweden: Analysis using data from the Cardiac Insufficiency Bisoprolol Study II” (with Niklas Zethraeus and Bengt Jönsson). Treatment of heart failure with beta blockers was introduced in Sweden already in the 1970s, but it was not until the 1990s that large-scale clinical trials established the efficacy of beta blockers in reducing heart failure mortality. The study consists of an economic evaluation of the beta blocker bisoprolol added to standard treatment of chronic heart failure, compared with placebo added to the same standard treatment. The study raises a number of methodological issues. At the forefront are the inclusion of costs of added years of life, and the question of how to model health effects that extend beyond the clinical trial on which the economic evaluation is based. The results indi­cate that treatment with bisoprolol is cost-effective. A drawback of the analysis in the fourth study was that the expected survival after the end of follow-up was modelled deterministically. This makes it impossible to assess the uncertainty of the cost-effectiveness estimate in a realistic way.The fifth essay is entitled “Assessing uncertainty in cost-effectiveness analysis by combining resampling of clinical trial data with stochastic modelling: The economic evaluation of bisoprolol for heart failure revisited”. Here, the drawback with the fourth study that was mentioned above is addressed by using resamp­ling of the clinical trial data in combination with stochastic modelling of the expected survival after the end of follow-up in the clinical trial. The methodology is inspired by the bootstrap method, which is a simulation technique whereby various statistics, like the mean and variance, can be estimated through repeated resampling from the original sample. The difference from the traditional boot­strap method is that resampling of observations from the clinical trial data is combined with stochastic modelling of the expected remaining lifetime of the patients who were alive at the end of the clinical trial. Cost-effectiveness acceptability curves for treatment of heart failure with bisoprolol were obtained as a result of the analysis.The sixth essay, “Survival analysis techniques for estimating the costs attributable to head and neck cancer in Sweden”, concerns the estimation of average treatment cost attri­butable to a disease when the data contain censored, i.e. incomplete, observations. For various reasons, censored observations are common in medical and epidemiological studies. As a result, the length of the survival time or the size of the costs for those who are alive at the end of follow-up are not exactly known. This is of course problematic if we want to estimate the average survival time or the average cost for all patients, both survivors and non-survivors included. In this study, the Kaplan-Meier sample-average estimator is used for overcoming the problem with censored observations. It is a method that has been proposed specifically for handling censored cost data.