Use of Research Evidence in State Policymaking for Childhood Obesity Prevention in Minnesota

Abstract

Objectives. We describe how scientific evidence about obesity has been used in Minnesota legislative materials to understand how research evidence might more effectively be translated into policymaking.

Methods. We selected 13 obesity-related bills introduced from 2007 to 2011 in Minnesota. Using state archives, we collected all legislative committee meeting materials and floor testimony related to each bill. We used a coding instrument to systematically analyze the content of a sample of 109 materials for their use of research evidence and non–research-based information.

Results. Research evidence was mentioned in 41% of all legislative materials. Evidence was often used to describe the prevalence or consequences of obesity or policy impacts but not to describe health disparities. In 45% of materials that cited evidence, no source of evidence was indicated. By contrast, 92% of materials presented non–research-based information, such as expert beliefs, constituent opinion, political principles, and anecdotes.

Conclusions. Despite an abundance of available research evidence on obesity, less than half of legislative materials cited any such evidence in discussions around obesity-related bills under consideration in Minnesota.

Increased recognition of childhood obesity as a public health crisis has spurred numerous legislative and regulatory proposals at the national, state, and local levels. Experts agree on the need for better alignment of the evidence base for obesity prevention with the evidence needs of decision makers.^1,2 Unfortunately, the pace of research translation is slow relative to the scope and urgency of the problem, with 17% of US youths currently clinically obese,³ although small declines in obesity rates have recently been observed.^3,4 A comprehensive evidence-based policy approach is needed to continue to make a meaningful impact.^5,6 However, researchers have shown that policymaking is infrequently supported by research evidence; instead, lawmakers balance personal experience, constituent concerns, anecdotes, values, and political expediency to make policy decisions.^5,7–9 Although these themes of policymaking are well known, little information is available to quantify how frequently research evidence—in comparison with other types of non–research-based persuasive appeals—is cited in policy-relevant discourse. We designed this study to fill this gap.

Many researchers have observed that researchers and policymakers live in different universes,⁷ posing major challenges to the process of effectively translating research evidence to policymakers.⁸ For instance, researchers work on different time horizons and have different incentives than policymakers, leading to research that is ill timed for the legislative cycle and synthesized in language that is appropriate for academic settings rather than being accessible to most policymakers.^7,10 Yet, literature in the growing field of knowledge translation has highlighted recommendations to promote more consistent use of evidence by policymakers. For instance, research on state policymakers’ perceptions of the role of evidence in health policy has offered generic tips about how to better disseminate evidence to have a more meaningful impact on the policy process, such as ensuring that the evidence is locally relevant and timely, identifies specific policy implications, and is presented in easy-to-grasp written formats or, especially, conveyed in person.^7,11–13

Although these generic recommendations have value, other research has suggested that the specific political and economic context of the particular policy domain affects the ways in which evidence is used in that domain,^11,14,15 suggesting that obesity-related policy demands specific scrutiny. In addition, although many previous studies on translating research to policymakers have a national scope,^13,16–18 much policy action on childhood obesity prevention occurs at the state level.¹⁹ Thus, a focused analysis of state-level legislative discussion can help inform understanding of the evidence translation process for obesity policymaking.¹⁸ Developing actionable recommendations for improving the use of evidence in policymaking will require considering the specific context of childhood obesity and attending to variation in the types of policy arenas encompassed (e.g., nutrition policy, physical activity policy) under the broad domain of obesity prevention.

RESEARCH OBJECTIVES

The primary objective of this research was to quantify the extent to which research evidence and other non–research-based information have been used in materials related to childhood obesity prevention bills introduced in Minnesota between 2007 and 2011. To understand opportunities to better tailor research evidence to the particular policy environment in Minnesota, we also examined the most frequent sources of research evidence cited, the types of research evidence used, and differences in research evidence citation by characteristics of the proposed legislation. Specifically, we asked whether presentation of research evidence varied for legislation concerning the nutrition policy area compared with active living or physical activity–related policy, whether presentation of research evidence differed by whether the information was presented orally or written or whether the goal of the document or testimony was to provide descriptive background information or to recommend a specific policy, and finally whether materials associated with bills that passed the legislature used evidence more or less than materials for bills that did not pass. We conducted a content analysis of legislative materials to answer these research questions.

METHODS

The content analysis process proceeded in multiple stages: sampling policy events, collecting materials, sampling materials, developing a coding scheme, applying the coding scheme to collect data on each material, and analyzing the data.^20,21

Sampling Policy Events

To identify key obesity legislation in Minnesota, the research team conducted 11 in-person key informant interviews with obesity prevention advocates in the state during the summer of 2011, selected through study team guidance and snowball sampling on the basis of their familiarity with legislative events. The interview topics concerned interviewees’ recollections of salient obesity prevention policy events since 2007 as well as their reflections on how evidence was used in those legislative debates (as preparation for a follow-up study). Interviews lasted about 45 minutes.

After reviewing interview notes, transcripts, or both and information on the Minnesota state legislature and National Conference of State Legislatures Web sites, we constructed a timeline of all obesity-relevant legislation introduced in Minnesota between 2007 and 2011, the last year for which we had complete information. We then purposively sampled 13 of the bills for analysis (Table 1), selecting bills to be roughly divided according to principal policy area: 6 bills targeted nutrition or general obesity (school meal reimbursement, community grants for obesity prevention, body mass index measurement and nutrition education in schools, farm-to-school programs, liability against food companies), and 7 bills targeted active living environments or physical activity (physical education standards in schools, policy surrounding school siting, Complete Streets, Safe Routes to Schools). For 1 bill (Safe Routes to Schools), no supporting materials were available because it was introduced in 2011 with little legislative discussion, so no materials associated with this proposed legislation were included in the final analytic sample.

TABLE 1—

Minnesota Obesity-Related Proposed Legislation Selected for Study Inclusion: 2007–2011

Year and Bill Topic	Bill Status
2007
School meal reimbursement increase	Passed, signed by governor
Physical education standards and school wellness policies	Passed, vetoed by governor
2008
State Health Improvement Program (community grants)	Passed, signed by governor
BMI screening and nutrition education in schools	Did not pass
2009
Physical education standards	Did not pass
School meal reimbursement increase	Did not pass
Farm to school program funding	Did not pass
School siting	Passed, signed by governor
2010
Physical education standards	Passed, signed by governor
Complete Streets	Passed, signed by governor
2011
Personal responsibility in food consumption act (reduced liability to food producers)	Passed, vetoed by governor
Joint use agreements for reduced liability for school facilities	Passed, signed by governor
Safe Routes to School state fundinga	Did not pass

Note. BMI=body mass index. We selected these 13 bills from a timeline of 2007–2011 legislative events that we compiled through key informant interviews and assessments from the National Conference of State Legislatures. The full timeline consisted of a total of 33 bills introduced in these 5 years concerning health care reform, agriculture (e.g., promotion of local foods), food production, active transportation, and school-based health promotion. We selected included bills on the basis of bill timing (at least 2 bills in each year), distribution of success and failure, and diversity in approach to addressing obesity.

^aAlthough originally part of our original study design strategy, there were no archived legislative materials associated with the introduction of Safe Routes to School funding in 2011 (it was reintroduced in 2012), and so no materials associated with this legislative event ended up in our analytic sample of materials.

Collecting and Sampling Materials

We used the archives at the Minnesota Legislative Library and the Web site of the Minnesota state legislature to identify every material associated with each selected legislation. We defined materials as any document circulated at bill hearings (including letters, media articles, fact sheets, legislative research analyses, reports) as well as every archived audio or video testimony from hearings. We next created a database of every document and testimony associated with each bill, labeled by date and with every material assigned a unique identifier. The total sample size of all materials collected after eliminating duplicate materials, corrupt audio files, and materials clearly unrelated to the obesity prevention aspect of included bills (i.e., nonrelated aspects of omnibus or otherwise comprehensive bills) was 200. Given the time intensity of analyzing the oral testimony, the sheer volume of testimony relative to other types of materials (130 of the 200 materials collected were testimony), and our interest in comparing print and oral modes of legislative discourse, we took a 50% random sample of testimony, yielding 65 testimony materials in the final sample and 70 nontestimony materials (n = 135). Finally, we excluded the text of the 13 bills (including 13 House and 13 Senate versions, or 26 materials) from analysis because, per Minnesota legislative customs, no preamble or introductory text is used in bills (so content was strictly legal with no additional descriptive information that would meet study requirements for coding). The final sample size was 109 materials.

Constructing a Coding Instrument

The next stage in the process was to develop a coding instrument to collect data on each document and testimony material in the sample. The coding instrument consisted of 3 parts: (1) descriptive information about the material (e.g., length, date, policy issue area, type of material, whether the material included descriptive information or a specific call to support or oppose the bill), (2) information about use of evidence in the documents (e.g., type of evidence cited such as prevalence or disparities data, source of evidence cited, if available; format of evidence citations), and (3) information about use of non–research-based information (e.g., anecdotes, mention of political values, mention of policies in other states). See the Appendix (available as a supplement to the online version of this article at http://www.ajph.org) for the complete coding instrument. Most of the variables collected in the instrument were dichotomous: the information was either present in or absent from the material.

Based on team consensus, we defined research evidence as data (both qualitative and quantitative) that came from a systematic investigation designed to develop or contribute to generalizable knowledge. The investigations could include randomized controlled trials, cohort or case-control studies, observational studies, and general epidemiological surveys. We also defined specific indicators to identify use of research evidence, including citing a research study or researchers; footnoting a research study; using specific numbers or time spans or comparative terms such as “odds,” “likelihood,” “rates,” “risk,” or “significantly” to describe those specific numbers; using the word “data”; and paraphrasing a finding from a well-known research report, even if unattributed (such as “children in this generation may for the first time have a shorter life expectancy than their parents”²²). After establishing the key variables of interest, we created a coding definition book that included examples and definitions for all variables. Through an iterative process of pilot coding and team deliberation, we established the reliability of our coding instrument. Documents were divided between 2 coders, with a sample of 30 documents double-coded, and our reliability in the double-coded sample was sufficiently high (κs > .60²³).

Collecting and Analyzing Data

After filling out paper coding instruments for each of the 109 materials in the sample, we entered all data into a Web-based database designed for the study. The complete data set was exported into STATA version 10.0 for quantitative analysis (StataCorp, College Station, TX). The analysis included descriptive statistics for all variables in the data set, as well as tests for differences by policy issue area, type of document (oral vs written), intent of document, and success or failure of bills by using the Pearson χ² test. Finally, we estimated a multivariable logistic regression model to examine whether the use of research evidence in documents varied by bill passage, adjusting for potentially confounding variables including policy issue area and obesity focus of bill.

RESULTS

Table 2 presents descriptive information about the sample of 109 documents. By design, almost 60% of the sample consisted of testimony (n = 65), of which 68% was available as audio and 32% was available as video. The remaining 40% of the sample was print documents, including news articles, fact sheets, policy briefs, and reports; other documents included letters to officials from stakeholders, local school district policies, legislative research reports, and lists of grants awarded to local organizations. Testimony length varied from 1 minute or less (25% of the sample) to 27 minutes, and about 80% of testimony was less than 6 minutes. Most printed documents were 2 pages or less (73%), and the longest document in the sample was 17 pages. More materials were drawn from 2009 (n = 49) than in the other years because 4 of the selected bills were introduced in 2009 (Table 1).

TABLE 2—

Descriptive Information About Minnesota Obesity-Related Proposed Legislation Analyzed: 2007–2011

Item Analyzed	No. (%)
Type of material
Testimony	65 (59.6)
News articles	9 (8.3)
Fact sheet	7 (6.4)
Policy brief	2 (1.8)
Research report	1 (0.9)
Meeting minutes	1 (0.9)
Othera	24 (22.0)
Year
2007	20 (18.4)
2008	13 (11.9)
2009	49 (45.0)
2010	16 (14.7)
2011	11 (10.1)
Printed document length (n = 44)
1 page	19 (43.2)
2 pages	13 (29.5)
3 page	2 (4.5)
4–6 pages	9 (20.5)
≥ 7 pages	1 (2.3)
Testimony length,b min (n = 60)
≤ 1	15 (25.0)
1–3	14 (23.3)
4–6	18 (30.0)
6–9	6 (10.0)
≥ 10	6 (10.0)
≥ 20	1 (1.7)
Author
Nonprofit	29 (26.6)
Legislator	22 (20.2)
Legislative research	13 (11.9)
State agency	9 (8.3)
Journalist	9 (8.3)
School or education representative	8 (7.3)
Coalition	7 (6.4)
University or academic institution	6 (5.5)
Private sector	3 (2.8)
Federal agency	1 (0.9)
Celebrity	1 (0.9)
Unknown	1 (0.9)
Purpose of material
Descriptive	34 (31.2)
Explicit policy recommendation	75 (68.8)

Note. The sample size was n = 109.

^a“Other” included letters, school district policies, list of funded grants, and House or Senate research reports.

^bThere were 65 testimony files, but 5 files were missing time-stamp information.

Figure 1 displays the use of research evidence and non–research-based information for the sample over time (use of research evidence and non-research-based information was not mutually exclusive). Overall, 41% of all materials cited research evidence, and 92% cited non–research-based information. We found some variation by year, with more research evidence cited in the last year of the study compared with the earlier years (not statistically significant). The use of non–research-based information declined from 100% of materials in 2007 to 82% of materials in 2011.

FIGURE 1—

Proportion of sampled obesity-related legislative material from Minnesota citing research evidence and non–research-based information, by year: 2007–2011.

Note. Differences by year for research evidence citation and non–research-based citation were not significant (based on the Pearson χ² test).

Although 41% (or 45 materials) offered research evidence, the specific types of obesity- related evidence they offered varied (Table 3). More research evidence (50%) was cited in materials concerning nutrition policy than in active living materials (36%), although this difference was not statistically significant (P = .16).

TABLE 3—

Types of Research Evidence Cited in Minnesota Obesity-Related Proposed Legislation: 2007–2011

Evidence Type	Overall (n = 109), No. (%)	Nutrition (Nn = 40), No. (%)	Active Living (n = 69), No. (%)	Diff. by Policy Type,a P
Is any research evidence mentioned?	45 (41.3)	20 (50.0)	25 (36.2)	.159
Describes the magnitude of the obesity or a risk factor	23 (51.1)	11 (55.0)	12 (48.0)	.641
Describes a cause of obesity
Individual	5 (11.1)	2 (10.0)	3 (12.0)	.832
Environmental	6 (13.3)	3 (15.0)	3 (12.0)	.769
Describes the impact of a policy, program, or behavior change	21 (46.7)	5 (25.0)	16 (64.0)	.009
Mentions a specific policy or program	12 (57.1)	4 (80.0)	8 (50.0)	.237
Describes any consequence of obesity	28 (62.2)	11 (55.0)	17 (68.0)	.371
Mentions health consequences	13 (28.9)	8 (40.0)	5 (20.0)	.141
Mentions health care costs	11 (24.4)	6 (30.0)	5 (20.0)	.438
Mentions impacts on workforce or productivity	2 (4.4)	2 (10.0)	0 (0.0)	.106
Mentions academic outcomes	10 (22.2)	0 (0.0)	10 (40.0)	.001
Mentions quality of life or psychosocial consequences	6 (13.3)	1 (5.0)	5 (20.0)	.141
Mentions impact on defense or national security	3 (6.7)	0 (0.0)	3 (12.0)	.109
Target groups described
Mentions children	21 (46.7)	9 (45.0)	12 (48.0)	.841
Mentions seniors	4 (8.9)	3 (15.0)	1 (4.0)	.198
Mentions racial or ethnic groups	0 (0.0)	0 (0.0)	0 (0.0)
Mentions socioeconomic groups	0 (0.0)	0 (0.0)	0 (0.0)
Mentions gender differences	2 (4.4)	0 (0.0)	2 (8.0)	.196
Mentions geographic differences/ urban versus rural	0 (0.0)	0 (0.0)	0 (0.0)
Format of research evidence
Uses a graph or figure	5 (11.1)	3 (15.0)	2 (8.0)	.722
Uses a table	9 (20.0)	3 (15.0)	6 (24.0)	.469
Uses a map	0 (0.0)	0 (0.0)	0 (0.0)
Uses bullet points	10 (22.2)	4 (20.0)	6 (24.0)	.512
Uses footnotes or in-text citations	8 (17.8)	4 (20.0)	4 (16.0)	.725
Geographic relevance of research evidence
National	15 (33.3)	8 (40.0)	7 (28.0)	.396
MN state	22 (48.9)	9 (45.0)	13 (52.0)	.641
MN locality	8 (17.8)	2 (10.0)	6 (24.0)	.222
Non-MN state (adjacent)	1 (2.2)	0 (0.0)	1 (4.0)	.366
Non-MN state (nonadjacent)	4 (8.9)	1 (5.0)	3 (12.0)	.412
Unknown	8 (17.8)	2 (10.0)	6 (24.0)	.222

^aBased on the Pearson χ² test.

Most evidence was used to describe the magnitude of obesity or a risk factor (i.e., citing the prevalence of unhealthy diets or the prevalence of obesity) or to describe the impact of a behavior, policy, or program. Significantly more evidence was used to describe the impact of a program in active living–related materials (64%) than in those concerning nutrition issues (25%; P = .009). Evidence was also used to describe the consequences of obesity, with 10 materials—or 40% of all active living materials that cited evidence—mentioning information on the association between weight and academic outcomes. Research evidence concerning children was mentioned in almost half (47%) of all documents that cited any evidence. No materials presented any research evidence on racial or ethnic groups, groups defined by socioeconomic status, or differences by geography.

When presenting research evidence, materials used graphs or figures (11%), tables (20%), or bullet points (22%); some (18%) used footnotes or other citations. Much evidence was reported at the state level, although national data and more local-level data were also presented in materials.

Table 4 offers a summary of the types of non–research-based information cited in the sample. The most common types of non–research-based information cited were expert opinion (48%), cost information not generated from a research study (37%), mention of specific politicians who support or do not support an issue (36%), public opinion (33%), political principles or ideology (32%), and narratives or anecdotes (24%). Among the narratives, the most common narrative concerned a community, neighborhood, or school. Few differences existed between materials describing nutrition-related legislation versus active living legislation, although the former more frequently cited analogies to other policy arenas, particularly tobacco policy.

TABLE 4—

Types of Non–Research-Based Information Cited in Minnesota Obesity-Related Proposed Legislation: 2007–2011

Evidence Type	Overall (n = 109), No. (%)	Nutrition (n = 40), No. (%)	Active Living (n = 69), No. (%)	Differencee by Policy Type,a P
Is non–research-based evidence mentioned?	100 (91.7)	39 (97.5)	61 (88.4)	.10
Describes public, stakeholder, or constituent opinion	33 (33.0)	9 (23.1)	24 (39.3)	.092
Describes expert opinion or testimony	48 (48.0)	17 (43.6)	31 (50.8)	.48
If yes, identifies a professional society or formal group recommendation	22 (45.8)	5 (29.4)	17 (54.8)	.091
Mentions cost or cost-effectiveness	37 (37.0)	17 (43.6)	20 (32.8)	.275
Describes a narrative or anecdote about an identifiable individual	24 (24.0)	11 (28.2)	13 (21.3)	.431
If yes, identifies about self, family, friend, or colleague	7 (7.0)	2 (5.1)	5 (8.2)	.557
If yes, identifies about constituent or community member	2 (2.0)	2 (5.1)	0 (0.0)	.074
If yes, about some other individual	4 (4.0)	2 (5.1)	2 (3.3)	.645
If yes, mentions a community, neighborhood, or school	14 (14.0)	6 (15.4)	8 (13.1)	.75
If yes, mentions a local business	1 (1.0)	1 (2.6)	0 (0.0)	.209
Appeals to emotions or nostalgia	3 (3.0)	2 (5.1)	1 (1.6)	.318
Appeals to political principle or ideological orientation	32 (32.0)	13 (33.3)	19 (31.2)	.819
Discusses particular political strategies
Mentions legislators, other politicians who support an issue	36 (36.0)	12 (30.8)	24 (39.3)	.384
Mentions policy outcomes in other states	11 (11.0)	4 (10.3)	7 (11.5)	.849
Mentions analogy to another policy arena (e.g., tobacco)	4 (4.0)	4 (10.3)	0 (0.0)	.011

^aBased on the Pearson χ² test.

Next, we examined the sources of research evidence cited. Materials most commonly did not cite a source at all for particular research evidence mentioned (in 45% of materials that cited evidence) or just indicated some generic reference to research such as that it came from a “study” (in 30%). Few (5 materials, or 13%) referred to specific peer-reviewed studies in journals, although more cited federal government sources (23%; e.g., the Centers for Disease Control and Prevention) or state agency sources (20%; e.g., Minnesota Department of Health). Three materials cited a national research institution (such as the National Institutes of Health), 1 material cited a health care organization as source, and 10 cited multiple sources for the same type of evidence. No materials cited local or nonlocal academic institutions, local nongovernmental organizations, industry, the news media, or a state policy assistance organization (e.g., National Conference of State Legislatures) as the source of research evidence.

Finally, we explored differences in the frequency of citation of research evidence or non–research information by 3 key variables: document type, intent of document, and bill success or failure. First, we found that oral testimony was less likely to cite research evidence (34%) than printed materials (52%; P = .055), but we found no differences in the likelihood of citing non–research-based information. Next, we found no significant differences in citation of evidence or non–research-based information for materials that presented descriptive information with no policy recommendation compared with those that included a specific call to action or policy recommendation. Comparing the materials affiliated with the bills that passed (Table 1), we found that materials for bills that passed used evidence less (32%) than those that did not pass (58%; P = .01) although this was confounded with policy type, because active living policies passed more often than nutrition- or obesity-related policies in this time period. In a logistic regression model predicting use of evidence or not, we found no significant relationship with bill passage once controlling for the policy issue area and obesity focus.

DISCUSSION

Research evidence was cited in just less than half of policy-relevant materials in Minnesota from 2007 to 2011, suggesting that policy stakeholders—advocates and legislators—may see the value of describing evidence to support policy decisions, but a large opportunity exists for research to be cited more frequently. Evidence on policy impact and prevalence of obesity was particularly likely to be cited in these policy discussions. However, other types of information—including stories, political principles or values, and expert beliefs without evidence—were much more likely to be included in policy documents. This supports previous work examining the influences on the policy process: although evidence is recognized as important, other factors particularly germane to the political context of policymaking are major influences on policy decisions.^7,24

Previous research examining use of evidence in policymaking has used surveys and interviews to capture policymakers’ perspectives on the extent to which evidence is valued in policymaking.^7,11–13 Other recent work has examined the content of 1 type of potentially influential material, obesity-related policy briefs, offering recommendations for improving their design.²⁵ In examining all types of documents and testimony used in 1 particular state context, this study goes beyond previous work. We were also able to disaggregate use of research evidence across different policy types and, in fact, we found some differences by type of policy intervention. Research evidence was more often cited (although not statistically significantly so) in nutrition policy documents than in documents for bills related to active living. This could be because the research evidence connecting nutrition to obesity was more available, better disseminated, or more readily used by the types of advocates and lobbyists involved in nutrition-related interventions than by those involved in community-level active living or physical activity policy. In other work, we found that research evidence presentation differed within the active living policy domain, with physical activity–related policies more commonly citing research evidence than community design– or urban planning–related policies.²⁶

Our research has indicated that documents utilized in the legislative process use some of the best practices endorsed in the research literature on translating health research to policymaking.^7,11–13,25 For instance, the majority of written documents used to support legislative discussions were 2 pages or less. About one fifth used bullet points, and a smaller proportion used tables or figures to convey research evidence. Among the non–research-based information used in the materials, about a quarter used stories or anecdotes. Other researchers have identified the importance of narrative forms of communication in the policy process, suggesting that evidence should be incorporated as much as possible in story form to have more of an impact on policymakers’ attention and interest.¹³

Given the focus of several pieces of legislation included in this analysis on children and school environments, evidence related to children was often reported in policy materials. However, in spite of the huge proliferation of evidence on the differences in obesity rates by race, ethnicity, socioeconomic status, and geography,²⁷ none of this research evidence entered the policy discourse we observed. This could suggest that research evidence about disparities was not considered to be the appropriate evidence to motivate stakeholders in Minnesota. Understanding why policy conversations are not engaging with research on health disparities is an important direction for future research.

Limitations

Although content analysis methods have value in that they can capture in an objective fashion the prevalence of particular messages and formats in documents and oral testimony, these methods only revealed the observable use of evidence in the policy process (and, indeed, we were further limited to only those materials available in existing archives). Thus, this study does not account for the frequent back-room discussions between legislators and among legislators, aides, staff, advocates, and lobbyists that are quite influential in garnering support for bills. Research evidence may be a salient part of these discussions, but this study does not capture this. In subsequent work, we are conducting qualitative interviews with policy stakeholders to better understand perceptions of use of evidence beyond the observable setting of legislative hearings. In addition, although our content analysis documented citation of research evidence and type of research evidence, we did not evaluate the quality of research evidence cited. Although we were able to capture whether evidence was incorporated into the material (and whether the source of the evidence was a peer-reviewed study), we cannot ascertain whether this evidence was generated from a well-designed, high-quality study. Yet we know that great variation exists in the quality of research evidence available about obesity.²⁸ Finally, these results are specific to Minnesota and so may not be generalizable to other state legislative contexts.

Another limitation concerns the small sample size in our final analytic sample of materials (n = 109), which precluded the application of complex statistical analyses. Given that our materials are nested within a particular legislative event, robust statistical analyses—beyond what we present here—ought to account for clustering and multilevel design. Future research using a larger sample size and capturing additional variation across years or state contexts might leverage these more sophisticated design elements. In addition, we document only descriptive associations between key variables of interest. We are not able to infer any causal relationships between evidence presentation and policy outcomes using this descriptive study design.

Conclusions and Policy Implications

The growing body of empirical literature on knowledge transfer has suggested that the success of a knowledge translation system depends on designing that system to match the local policymaking context.²⁹ This study indicates that the policy discourse surrounding obesity prevention in Minnesota attends to research but to a limited extent, possibly suggesting additional opportunities for incorporating evidence into the discussion alongside other potentially persuasive strategies, such as presenting public opinion and stories. Of course, just because a document or person testifying presents research does not mean that this evidence is persuasive on its own merit or that it is a quality study devoid of bias or conflict of interest. Future research must attend to the nuances of evidence translation—how it is communicated, when it can be persuasive, and how its users judge its quality.³⁰ By documenting the use of evidence in policy discourse over a 5-year period in 1 state, this study offers baseline data that can be used to design locally tailored evidence communication vehicles and then serve as a benchmark for comparing the extent to which research is part of policy communication in the future.