2013 Colloquium

Presented by: Professor Peter Davis

Most social science and public policy research is carried out in natural settings. Yet such research can rarely generate inferences of a plausibly causal status sufficient to inform policy interventions. However, advances have been made in helping researchers develop and draw more credible inferences from such data. These advances have come particularly from logicians and philosophers, who have generalised to observational work, a variant of the model of causal inference based on the experiment (potential outcomes, counterfactuals). Also from applied statisticians, particularly those working in econometrics and in educational and applied social research, who are concerned with drawing conclusions about policies and interventions. The presentation reviewed this work and asked the question of whether this model of causal inference can help the “policy sciences” make the case for intervention.

Presented by: Dr Barry Milne

Better access to primary health care has been shown to be associated with lower rates of Ambulatory Sensitive Hospitalisations (ASH), that is, hospitalisations for conditions that are thought to be preventable by timely and effective primary health care (e.g. asthma, cellulitis, hypertension, gastroenteritis). The introduction of the “Primary Healthcare Strategy” in New Zealand in 2001 led to an improvement in access to primary health care, and reductions in socio-economic and ethnic disparities in accessing primary health care.

In this presentation, Dr Milne presented data from 2001-2009 on whether these improvements led to reductions in rates of ASH, and to reductions in inequalities in ASH admissions. A novel method was described for creating population health data by combining health datasets with population tables.

Presented by: Mr Roy Lay-Yee

This presentation reported on a dynamic microsimulation model of the later life course (ages 65 years and older) focusing on two strategic areas with major policy implications: (1) the impact of the increasing prevalence of chronic disabling conditions on older people's use of health and social care, and (2) the impact of changing the balance of care for people in need (across a range of modalities). The model was built on data from two New Zealand series of repeated cross-sectional surveys on health and disability respectively.

Mr Lay-Yee described the construction of the model and how the model can be used to test policy-relevant scenarios. For example, by changing levels of disability or the balance of care and observing the impact on downstream outcomes.

Presented by: Professor Peter Davis

Professor Davis provided an outline of presentations for the after lunch session and an introduction to KIWI: a modeling focus.  

Presented by: Dr Barry Milne

Micro-simulation models require rules to determine how individuals transition from one stage to another. For our micro-simulation model of the early life-course, we derived these rules by analysing data from New Zealand's rich array of child longitudinal studies.

In this presentation, Dr Milne described how we have integrated data from four New Zealand longitudinal datasets for the purposes of analyses. He also described methods to weigh these datasets to represent the ethnic distribution of New Zealand.

Presented by: Mr Roy Lay-Yee

This presentation demonstrated an approach that uses a microsimulation model, based on real data, and counterfactual reasoning to test the differential impact of changing selected determinants for disadvantaged groups on a range of child outcomes. The focus was on health service use with a comparison to outcomes in non-health domains, namely educational attainment and antisocial behaviour, as a pointer to where policy initiatives might be the most effective.

Presented by: Dr Barry Milne

The analysis of synthetic data is often favoured when the release of ‘real’ data is not possible because of privacy and confidentiality concerns. Ideally, the synthetic data should mimic the properties of the real data but not contain information that would enable any ‘real’ data units (i.e. individuals) to be identified. To establish a representative starting file for our simulation of early life-course development, we created synthetic dataset of newborns by creating ‘composites’ of similar newborns from the 2006 Census.

In this presentation, Dr Milne described the procedures and how this method creates realistic data without identifying any individual in the Census.