To better understand the use of policy metrics, we reviewed articles published in 14 public health and preventive medicine journals. The journals chosen were broad, general public health journals and not specific to a single topic such as nutrition or disease. Journals that focused solely on policy and journal supplements were not included. We examined the following journals:

We defined a policy article as one that explicitly describes a policy, law, or regulation (including development, implementation, and evaluation). Using online archives, we conducted a systematic audit of articles published in 2008. Tables of contents were collected from each journal issue for that year. Two researchers reviewed the table of contents in each issue and compiled a list of policy-related articles. If the policy content was unclear from the title of the article, the abstract or full text was used. Any articles in question were reconciled by the research team until consensus was reached.

Once the list of policy articles was compiled, the titles were sorted by policy category. To examine policy metrics in detail, 78 articles from 2008 were analyzed. Editorials, commentaries, and reviews were excluded, resulting in 47 articles from which metrics were summarized. For articles that presented data analysis, we assessed policy metrics across several categories:

The articles examined were a mixture of both "big P" policy studies (eg, formal laws, rules, regulations enacted by elected officials) and "small p" policy research (eg, organizational guidelines, internal agency decisions or memoranda, social norms guiding behavior) (3). Articles were categorized as child health; maternal health; HIV/AIDS; drug use prevention; tobacco control; violence control; environmental and disaster preparedness and biosecurity; school health; special populations; worksite health; international health; advocacy; general policy; or health care.

The topics that were most represented were tobacco control, international health, and school health. Among international articles, health care was the most common topic. The Journal of School Health and the American Journal of Public Health published the most policy-related articles.

Most articles (74.5%) relied on a cross-sectional design (Table 1). Only 3 studies reported any economic or cost data. Fourteen studies reported on psychometric properties of the metrics. Most presented new data on psychometric testing (n = 10), while some referred to previous articles (n = 4). The testing most often reported was for reliability (eg, interrater reliability), internal consistency, or key informant validation of methods. When categorizing according to 3 levels of outcomes, most were downstream (n = 31), followed by midstream (n = 13) and upstream (n = 3). Detailed data on health disparities (eg, subgroup analysis for vulnerable populations) were available for only 2 studies. Both of these studies (10,11) explicitly investigated differences among disparate groups; 1 studied how national laws that increased tobacco prices affected smoking prevalence among different socioeconomic groups (by sex, occupation, and birth cohort), and the other investigated differences in the use of skilled birth attendants by women of varying wealth in several countries.

Most of these studies dealt with the effectiveness or evaluation of a given policy that is in effect. Three studies focused on characteristics of or influences on policies that are successfully "passed."

Policy outcomes can be monitored by accumulating evidence from many sources to gain insight into a particular topic, often combining quantitative and qualitative data to understand content and track progress. Consensus on valid and useful measures is needed (12). Successfully monitoring outcomes will also require sources beyond the usual public health data sets (eg, tax revenue, polling, and marketing data). We used the 3 domains of evidence-based policy (process, content, outcome) to present sample metrics across the 3 domains (Table 2). Metrics are quantitative (eg, the percentage of the population with a particular health behavior) and qualitative (eg, the content of a certain policy). Most studies in this review were cross-sectional; stronger study designs are needed to improve the evidence base.

Although much of the effect of public health policy occurs locally, in many jurisdictions high-quality data are lacking at the city, county, or metropolitan levels. Some attempts have been made to identify local-level indicators (13), but a set of consensus policy metrics needs to be developed for local areas, as has been done at the national and state levels.

A public health adage is "what gets measured gets done" (14). This has typically been applied to downstream endpoints; however, for policy approaches, midstream and upstream metrics are needed. A few efforts are under way to develop public health policy surveillance systems. For example, a group of federal and voluntary agencies has developed policy surveillance systems for tobacco, alcohol, and more recently, school-based nutrition and physical education (3).

Policy-relevant evidence should come from settings and organizations that reflect public health practice and policy. For example, efforts such as the Steps to a HealthierUS initiative, YMCA's Activate America, and faith-based interventions demonstrate that existing approaches for leadership development can enhance the use of evidence for promoting physical activity (15). As these efforts are documented, specific attention should be given to the key metrics for measuring progress.

Researchers and policy makers sometimes exist in parallel universes because of decision-making differences, poor timing, ambiguous findings, and lack of relevant data (16). Metrics may become relevant to policy makers when the effects of a health outcome are framed in terms of the direct impact on one's community, family, or constituents (17). An excellent example comes from the Rudd Center Revenue Calculator (www.yaleruddcenter.org/sodatax.aspx), which shows the revenue that could be generated from a 1-cent excise tax per ounce of sugar-sweetened beverages by state or municipality.

Eliminating health disparities is a policy imperative. To achieve this goal, we need to better articulate the key domains of inequality. For example, variables have included race/ethnicity, socioeconomic status or social class, geography, age, and sex (18). Our review of the existing literature showed sparse attention to metrics for health disparities and policy.

In deciding whether to take action and how to prioritize resources, policy makers often ask 3 questions: 1) Is there a problem? 2) Do we know how to fix the problem? and 3) How much will it cost? We probably have the most data for answering the first question (19), an intermediate amount for the second (20), and the least data for the economic issues (21). Studies of disease burden that use comparative units of analysis (eg, quality-adjusted life years) provide a basis for economic evaluations (22). Since much of the literature on pay-for-performance has focused on financial incentives, more work is needed to understand how the concepts apply to population-level public health policy.

Although public health research and practice are often segregated into "silos" because of categorical funding streams and interest groups (23), much can be learned across content areas. For example, several authors have examined the lessons from tobacco control that can be applied to the obesity epidemic (24,25). Similar areas in public health where policy measurement is advanced may provide beneficial insights to developing topics.