gms | German Medical Science

Background: Evaluation in Health services research (HSR) inherently takes place within complex social contexts, such as clinics, certain regions, or cases. Additionally, HSR desires to provide causal findings on health services processes for health policy makers. Both of which increase the evaluation demands in HSR. While conservative quantitative approaches can be challenged by the complexity and heterogeneity of the field in HSR, qualitative approaches avoid causal findings. Within this spectrum, HSR seems caught between unmet evaluation demands. However, according to Pearl [1] drawing causal inference needs the combination of data, methodology and causal knowledge and assumptions (theory).

Objective: This contribution aims to bridge the gap between qualitative and quantitative research in HSR, while strengthening theory and causal knowledge. Taking common scenarios in HSR as examples, such as an intervention (X) aiming at reducing adverse outcomes (Y) in different clinics (C), we will demonstrate how a causal model can be evaluated with the integrated inferences approach [2], which utilizes qualitative as well as quantitative data.

Methods: The integrated inference approach dissolves the qualitative and quantitative methodological distinction through appreciating the importance of model-based causal inference. Under the lens of integrated inference, causal models are evaluated based on the probability of finding these empirical observations. In detail, we will show how:

1.

directed acyclic graphs (DAGs) can be used to translate theory and contextual knowledge into causal models;

2.

The integrated inference approach can be applied based on these causal models and Bayesian statistics;

3.

The quantitative and qualitative data are to be combined to provide (multi-)case causal effects instead of average causal effects;

4.

The estimated models can be defended (and tested) against alternative causal models with prior and/or prospective evidence.

Implication for research and/or (healthcare) practice: In Health Services Research (HSR), it is recommended to apply integrated inference and causal modelling before data collection. This allows for observation, updating, and evaluation of causal models based on the probability of observing them in reality. When making causal claims, such as 'Does Intervention X improve Outcome Y?', HSR should consider causal modelling prior to data collection. The integrated inference approach can advance evaluation standards in HSR by making theories and assumptions in causal models explicit. Integrated inference is especially useful, when collecting data during times of disruption or crisis and parametric demands of quantitative models are not met. This alternative is not dependent on the size of N, but rather on theory and testing of causal models.