important and what stakeholder actions should achieve. A CHIP and the community coalition that is at its core should provide a framework within which a community can reach agreement on the values and expectations to be represented in performance indicators.
Indicator selection should also reflect a strategic consideration of the value of individual indicators and of the collection of indicators to be used in connection with a specific health issue (Sofaer, 1995). Individual indicators become more valuable to the extent that they are effective proxies for multiple dimensions of performance. A set of indicators will usually be needed to cover a range of relevant performance areas and must be assembled carefully to assure that, together, the indicators effectively serve the process of improving the community's health. The set should appeal to many stakeholders and reflect broad consideration of the domains of the field model, but it should be limited to a comprehensible number of indicators. Too many indicators become distracting and, in practical terms, could make collecting and analyzing data prohibitively burdensome (Sofaer, 1995).
Operational implications and costs of data collection and analysis also must be considered in selecting indicators and indicator sets. Even though indicators may be formulated with the intention of promoting actions that will have positive effects on community health, they must be based on an accurate understanding of their effect in the setting in which they will be applied. For example, reducing the number of cigarette vending machines as a way to limit youth access to tobacco will not have the anticipated impact if teenagers buy most of their cigarettes in convenience stores. It also is possible to frame indicators in a way that creates "perverse incentives" for action, which produce a "better" measured result but do not achieve intended health goals. For example, lower rates of sexually transmitted diseases might be "achieved'' through less complete reporting rather than through true reductions in disease rates.
A reasonable balance must be struck between the information value of an indicator and the cost of collecting the necessary data. A conceptually appropriate indicator will not be helpful if a community cannot afford to obtain the data it requires. Costs of data generation may include designing data collection instruments, collecting or locating data, analyzing and summarizing the results, and reporting information to the community. In some communities and for some indicators, these activities may require new resources. In other cases, it may be possible to apply existing