Randomized trials are the gold standard method for inferring causality. In health services research, however, it is often neither practical nor ethical to use randomized trials. However, it may be possible to infer causal effects from observational data by making use of so-called natural experiments, exploiting random variations in exposures of interest. This project has four work packages.
The first package will study the effects of practice variation, such as differences in referral practices between general practitioners, and differences in the availability of specialized primary health services between geographical areas. To date, studies have been published on the effects of general practitioner discontinuity, and the effects of capacity pressure on surgery for hip fracture patients. In package 2, we will look at the effects of centralization of hospital services. Here, the first studies will be ready for publication in 2022. The third package aims to increase our knowledge of vulnerable groups' needs for health services by studying the causal effects of mental health problems and socioeconomic status on the use of health services. Here, genetic markers are used as a source of exogenous variation in exposure. For example, we have published several studies on the consequences of body mass index, such as the effect on personal income, health service use and work participation. The fourth package will ensure user participation through collaboration with health personnel in hospitals, among GPs, administrative managers in the health sector and patient organizations. We have, for example, arranged seminars, as well as participated in various settings with lectures, presentations and media contributions.
Questions regarding the health care system often revolve around cause and effect: Are births safer at larger hospitals than smaller? Are differences between physicians important for patient safety? Contrary to the introduction of new medications and treatment methods, organizational aspects of changes in health care services are rarely justified by randomized controlled trials, which is the primary method for assessing causal effects. Assessment of composite exposures, often involving patients with complex problems, is difficult to conduct within the rigorous frames required in such trials. Randomization is also difficult due to ethical or practical reasons associated with system level interventions. Because of this, health services research is essentially based on observational data.
We apply for funding to establish a research environment to improve causal inference in health and welfare services. The project group aims to establish a better foundation for knowledge-based decision for clinical management and policy evaluation. Utilizing state-of-the art research methodology for causal inference with observational data achieves this.
The research environment will be organized within the frame of four work packages: Work package (WP) 1: Practice variation between physicians and municipalities, WP2 centralization vs decentralization of the provision of specialized health services, WP 3 differences in health care needs for vulnerable groups and WP4 user involvment.
In this proposal, we present research designs which are receiving increased attention in health services research and epidemiology internationally. We aim to connect with other Norwegian groups with the aim of strengthening the national competence on causal inference with observational data. A strong environment for causal inference in health and welfare research at NTNU and nationally will ensure high quality evidence for decision-making and makes it possible to study a broad field of issues.