Please click here to review our image reproduction and use pol icies.

(Downloading may take up to 30 seconds.
If the slide opens in your browser, select File -> Save As to save it.)
Click on image to view larger version.

FIGURE 3—— Suggested diagrams of complex causal systems, (a) linking health outcomes to transportation, (b) linking changes in health outcomes to transport policies, and (c) showing the predicted health impact of controlling traffic speed.

Note. As a convention, the health outcomes are shown as being negative (harmful). This can be criticized as embodying a medical model of health and ignoring positive health, but that is not the intention. For example, if mental health is likely to improve as a result of an increase in physical activity, the chart registers this in terms of a reduction in impaired mental health (without use of clinical categories). There are 2 reasons for adopting this convention. First, it helps with reading the charts if the outcome always carries the same type of implication. Second, while it would have been possible to introduce a positive convention instead, in practice most recorded health outcomes are problems—deaths, diseases, injuries—and arguably these also are more likely to influence policy makers than more general considerations of well-being and positive health, however important we may consider these to be.

The causal direction for each link is specified with directional arrows. We use "+" for an increasing function (more of the item in the source box leads to more of that in the destination box), "–" for a decreasing function, and no sign indicator if mixed. If color is available, we prefer to use a color code: blue, red, and black, respectively. For any particular chain of causation from a specific policy intervention to a specific health outcome, the overall impact is positive (health gain) if there is an odd number of "–" (red) arrows, and negative (harmful) if there is an even number; however, the presence of a mixed (black) arrow makes the overall effect of the chain indeterminate. For this reason, it is best to avoid arrows of mixed sign, as with a hump-backed function that has rising and falling elements, and replace them with separate arrows that have unambiguous polarity.²

No attempt has yet been made in these charts to quantify the relationships shown. The thickness of the arrows can be used to show the strength of the causal association. The nature of the line (e.g., continuous, dashed, or dotted) can show the degree of confidence in the judgment that the causal link exists (i.e., that it is different from zero, so that the null hypothesis is rejected, which is approximately equivalent to a P value in statistics). The length can be used to represent duration, because causal relationships require time to take effect (e.g., latency in the causation of a disease such as cancer); in practice, this makes drawing diagrams complicated, and an alternative method is to place a delay box within the arrow.² Finally, it would be simple to create an electronic version that enables the reader to click from an arrow to text that provides evidence for its existence, strength, etc., and back again.