Cause-effect

Dear Ms. Hillow:

You ask a short question which raises very complicated issues. Philosophers have been pondering the nature of causality and the evidence thereof for hundreds of years.

The short answer is no, data from a cross-sectional population-based surveys cannot provide strong evidence of causality. You are measuring characteristics of a population, and populations are subject to notoriously complicated influences. If you find a change after implementing some program, there may be, and probably are, many other things which could have produced the observed change or inhibited the observed change.

Evidence for causation can be strengthened by having a control group, or in this case, a control population, as similar as possible to the intervention population but not receiving the intervention. Control populations are best used by measuring the outcomes of interest at baseline in both the control and intervention populations, then measuring the same outcomes using the same methodology in the control and intervention populations at some time after the intervention is implemented. If the change is greater in the intervention population than in the control population, this change may have been due, at least in part, to your intervention.

However, this is a very simplified summary of a complex issue. A better summary of the larger epistemologic issues can be found here: http://jech.bmj.com/content/56/2/119.full. In addition, any introductory epidemiology textbook will discuss the criteria to demonstrate causality in epidemiologic studies, and other books give excellent advice on use of community studies to evaluate public health programs.

Dear Ms. Hillow:

As long as you have only baseline and follow-up measurements of indicators, I don't think you have strong evidence that any change in the indicators is due to your program.

Is it causal? Considering the BRADFORD-HILL criteria:

Whilst of course properly designed and executed intervention studies are ideal in helping us decide whether things as causally related or not, there are many situations in which this is not possible / practical / ethical. Hence it's important to make the best of observational data. In addition to approaches already listed by other respondents, I've always found it helpful to go through the "Bradford Hill" criteria to establish whether something is more likely causally or more likely non-causally associated:

See http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1291382/ for a good summary (and critique) which discusses in more details the key points below:

"1) Strength of association
- A strong association is more likely to have a causal component than is a modest association

2) Consistency
- a relationship is observed repeatedly

3) Specificity
- a factor influences specifically a particular outcome or population

4) Temporality
The factor must precede the outcome it is assumed to affect

5) Biological gradient
The outcome increases monotonically with increasing dose of exposure or according to a function predicted by a substantive theory

6) Plausibility
The observed association can be plausibly explained by substantive matter (e.g. biological) explanations

7) Coherence
A causal conclusion should not fundamentally contradict present substantive knowledge

8) Experiment
Causation is more likely if evidence is based on randomised experiments

9) Analogy
For analogous exposures and outcomes an effect has already been shown

Hope that helps.