NDEP Public Survey OMB SUPPORTING STATEMENT B_29Sep2014 (2)

The National Diabetes Education Program (NDEP) contractor will procure the sample for the online survey from GfK’s KnowledgePanel® (KP). While we described the KP sample in detail in Supporting Statement A, the recruitment and empanelment methods in Attachment D, and GfK’s Best Practices for ICR Supporting Documentation for KnowledgePanel^® Surveys in G we summarize here the key points about the sample:

The KP is nationally representative, with sample demographics that compare very favorably with those reported by the U.S. Census Bureau (2003).
The GfK panel has been studied by other researchers and is generally considered representative of the U.S. population.¹
GfK’s KnowledgePanel® has sample coverage of non-Internet households by providing members that did not previously have internet access with laptops and internet access.
GfK includes predominantly Spanish-speaking households in their recruitment process and as members of their research panel and fields surveys in Spanish as needed.
The KnowledgePanel® offers another advantage not generally available to RDD surveys - information on sample characteristics including a profile of the non-responders.
The KnowledgePanel® has been used for a number of similar studies for various US government agencies over the past decade.

The recruitment rate for households participating on GfK’s research panel has increased using address based sampling (ABS) method. Of the eligible households that were contacted by GfK to become part of their research panel using ABS, 14% positively responded to GfK’s mail invitation, indicating interest in the panel. Of the households that indicated interest, 65% become participating panel members. According to GfK, these recruitment rates are higher than the slightly more than 50% achieved using RDD methods (Formerly from Knowledge Networks 2010--GfK acquired Knowledge Networks in late 2011).

The increasing number of cell phone-only households has increased coverage error associated with RDD sampling of telephone landlines. The change in the NNDS sampling strategy from RDD landline to ABS offers a reasonable solution for better inclusion of cell phone only households, solving some non-coverage bias issues. ABS also means better coverage of minority and younger adult groups, improving the representation of these key demographic groups in the sample and thereby increasing the representativeness of GfK’s research panel.² In 2009 GfK switched their sampling method for their probability based online panel from RDD to ABS, maintaining that ABS also provides a stable sampling base and flexibility in mixing modes to match the survey’s goals and target population.³

NNDS trends going forward should reflect widespread sample coverage nationwide since the GfK ABS sample now covers approximately 97 percent of the US population. Longitudinal analyses to track trends from past NNDS (RDD sampling) to future NNDS (ABS sampling) are possible, as GfK collects information from its panel members about their phone use. This information can then be used to form a subsample of landline households from the GfK ABS panel.⁴ Direct comparisons of ABS landline results to RDD landline results, using appropriate weights, can then be made to bridge differences between the two sampling strategies.

Sampling Frame

The NDEP traditionally conducted the NDEP National Diabetes Survey (NNDS) using a random digit dialing (RDD) sample frame of landline telephone households. In the face of continued decline in response to RDD landline phone surveys, the Address Based Sampling (ABS) offers comprehensive coverage of minorities, younger adults, and cell phone only household members--groups that increasingly are diminished or absent in RDD landline samples but important to the NDEP. ABS also bypasses many of the technical and operational inefficiencies often associated with dual frame landline and cell phone only populations.⁵

The sample frame for this study is a research panel, the KnowledgePanel®, recruited and maintained by GfK; it includes approximately 42,000 U.S. households (GfK, personal communication, April 30, 2014) corresponding to approximately 55,000 adult members ages 18 and older. This sample frame includes cell-phone only households, Spanish-speaking households, and households who did not previously have internet access. A comparison of the GfK’s KnowledgePanel® membership relative to demographic characteristics (i.e., benchmarks) from the U.S. Census is shown in Table B.1-1.

Table B.1-1. KnowledgePanel® Demographic Comparisons – August 2013

		Adult Panel Members [1]	Adult U.S. Population (June 2013 CPS [2] except as footnoted)
Gender	Male	48.20%	48.20%
	Female	51.80%	51.80%
Age	18–24	12.50%	12.80%
	25–34	17.40%	17.50%
	35–44	17.10%	16.70%
	45–54	17.50%	18.30%
	55–64	18.30%	16.50%
	65 or over	17.30%	18.30%
Race	White Only	76.60%	79.20%
	Black (African American) Only	12.70%	12.30%
	American Indian, Alaskan Native Only	1.00%	1.00%
	Asian Only	3.50%	5.40%
	Hawaiian or Pacific Islander Only	0.60%	0.40%
	2+ Races	5.60%	1.70%
Hispanic Ethnicity	Hispanic	15.00%	15.00%
	Non-Hispanic	85.00%	85.00%
Employment Status	In the Labor Force	67.20%	65.60%
	Employed	56.60%	60.70%
	Unemployed	10.60%	4.90%
	Not in the Labor Force	32.80%	34.40%
Marital Status	Married	52.80%	53.40%
	Not Married	47.20%	46.60%
^{Housing Ownership} [3]	Own	68.90%	68.90%
	Rent/Other	31.10%	31.10%
Level of Education	Less than High School Diploma	12.50%	12.70%
	High School Diploma or Equivalent	29.90%	29.70%
	Some College	28.50%	28.50%
	Bachelor’s Degree or Beyond	29.10%	29.10%
Household Income [3]	Under $10,000	5.30%	5.40%
	$10,000–$24,999	13.40%	13.40%
	$25,000–$49,999	23.50%	23.40%
	$50,000–$74,999	18.70%	18.60%
	$75,000 or more	39.10%	39.20%
Census Region	Northeast	18.30%	18.20%
	Midwest	21.50%	21.40%
	South	37.10%	37.10%
	West	23.10%	23.30%
^{Internet Access (Household)} [4]	Any Connection Speed	75.50%	78.90%
	Broadband	71.90%	72.50%

Active profiled adults are weighted to be representative of the U.S. population on age, gender, race, Hispanic ethnicity, language proficiency, region, metro status, education, household income, homeownership, and Internet access using post-stratification adjustments to offset any non-response or non-coverage bias.

² Estimates were calculated using June 2013 CPS microdata available at www.census.gov. The data are weighted using CPS final individual weights.

³ National housing and income statistics are from March 2012 CPS Annual Social and Economic Supplement.

⁴ National Internet coverage statistics are from July 2011 CPS Internet and Computer Usage Supplement.

Table B.1-2 shows the percent distribution of KP members for the top 10 metropolitan areas (Core-Based Statistical Areas or CBSAs) compared to CPS estimates. The KP closely matches the distribution, in that the estimated proportions of the population in the top CBSAs are not substantively different from these benchmarks.

Table B.1-2. KnowledgePanel^® Comparisons for the Top 10 CBSAs – August 2013

Top 10 CBSAs	Adult Panel Members¹	June 2013 CPS
New York – Northern New Jersey – Long Island	5.2%	6.2%
Los Angeles – Long Beach – Santa Ana	3.6%	4.3%
Chicago – Naperville – Joliet	2.6%	3.0%
Dallas – Fort Worth – Arlington	1.9%	2.1%
Houston-Baytown-Sugar Land	1.7%	2.0%
Washington – Arlington – Alexandria	1.9%	2.0%
Miami – Fort Lauderdale – Miami Beach	1.3%	1.9%
Philadelphia – Camden – Wilmington	2.0%	1.9%
Atlanta – Sandy Springs – Marietta	1.8%	1.7%
San Francisco-Oakland-Fremont	1.5%	1.6%

¹ Active profiled adults are weighted to be representative of the U.S. population on age, gender, race, Hispanic ethnicity, language proficiency, region, metro status, education, and Internet access using post-stratification adjustments to offset any non-response or non-coverage bias.

In addition to the demographic information shown, numerous other data elements are maintained for each KP member and are available for sample selection purposes and project-specific data analysis. Demographic data related to panel members sampled for this study, whether or not they participate in this study, will be made available to the NDEP. This information will be used to evaluate the representativeness of the survey respondents. GfK collects this demographic data as part of their initial panel recruitment process and is collected independent of this proposed data collection. Demographic data for panel members who are selected to participate in the survey but do not participate (i.e., non-respondents) will be provided to the NDEP. This will allow the NDEP to evaluate non-response bias in terms of possible systematic differences between panel members who are selected for this study and participate and those who are selected but do not participate.

Once panel members are profiled, they become “active” for selection for specific studies. Samples are drawn from among active members using a probability proportional to size (PPS) weighted sampling approach. Customized stratified random sampling based on profile data is also conducted, as required by specific studies.

In September 2007, Knowledge Networks was assigned a patent for its unique methodology for selecting multiple online survey samples from a panel (U.S. Patent No. 7,269,570). The selection methodology, which has been used by GfK/Knowledge Networks since 2000, assures that multiple sequential KnowledgePanel® samples from a finite panel membership will each reliably represent the U.S. population.

This sampling methodology was developed in recognition of the practical issue that different survey samples may target different panel subgroups. It is not unusual that only panel members with certain characteristics are selected for a survey. This subgroup selectivity can skew the remaining panel membership demographics and affect the representativeness of later survey samples. The patented sampling methodology was developed to correct for this⁶.

The expected precision for the overall survey can be expressed as the relative standard error (RSE), or the coefficient of variation (CV), which is the standard error of the sample as a percentage of the mean (or proportion). We used various percent estimates (P) and design effects (DEFF) to calculate the CV for different domain sample sizes (Table B.1-3). GfK uses a DEFF of about 1.3 for the general population. In the table, the CV for our intended sample size of 2500 is shown in the last column and lies between the rows for a DEFF of 1.25 to 1.5, ranging from 2 to 7, depending on the percent estimate of interest. For the Hispanic and African American subgroups, each having approximately 833 members, the CV would be in the range of about 4 to 13, based on the column of domain sample size equal to 800.

Table B.1-3. Percent Coefficient of Variation for Domains of Different Sizes

B.2. Procedures for the Collection of Information

SSS will work with GfK who will administer the NNDS via the Internet to KP members whose email addresses were sampled. Individuals will receive email notification that the survey is available for completion (Attachment E. GfK’s E-mail Notification to KnowledgePanel Participant for NNDS). Attachment F describes GfK’s process to verify respondent age. Eligible participants will include adults ages 35 years and over in the U.S. Participants will then give their informed consent for participation in the online survey. Participants will consent by selecting the appropriate link on the Web screen. The surveys will be self-administered and accessible any time of day for a designated period. Participants can complete the survey only once.

Non-respondents will receive two e-mail reminders requesting their participation in the survey. When a potential respondent is contacted but does not access the weblink to the survey within one week, a reminder e-mail will be sent. The e-mail reiterates the purpose of the survey and why the respondent’s participation is important.

B.3. Methods to Maximize Response Rates and Deal With Nonresponse

GfK expects to achieve a survey completion rate of greater than 65%.⁷ As reported by Callegaro and DiSogra (2008), the AAPOR cumulative response for GfK taking into account all phases of panel recruitment, profiling, and survey participation is approximately 10%. This is on the same order of the response rate for RDD surveys. Because this survey is panel-based, it is possible to achieve a higher response rate for several reasons:

Respondents have already consented to join the KnowledgePanel®.
They receive invitations from the KnowledgePanel®, a recognized source and address.
They may receive incentives for participation.
They are sent reminder emails as needed.

In addition, there are several ways that we will maximize cooperation and participation in this study:

The survey is a web-based (or online) survey, which KnowledgePanel® members fill in directly. GfK is able to use special features of this mode of data collection to enable respondents to understand the survey tasks and questions. The web-based mode incorporates skip patterns so that respondents are not asked to manually navigate to questions based on responses to previous questions. This reduces the time burden on respondents, and decreases the likelihood that respondents will respond to questions that they are not eligible to answer or miss questions that they are eligible to answer.
GfK uses non-survey-specific incentives. These incentives apply for any survey that is completed by a KP member and are not specific to this information collection. (See Attachment D for the empanelment methodology.) Though survey-specific incentives can be used for particular surveys such as those that exceed 20 minutes in length in order to increase completion rates, survey-specific incentives will not be used for this data collection.
GfK will make attempts to locate participants who leave the KP before the end of this study. Location efforts will include mailings of refusal conversion materials designed to persuade participants to complete the study. In addition to using mailed refusal conversion materials, GfK may also conduct telephone-based refusal conversion, contacting each non-responder via telephone.
GfK will provide a toll-free telephone number to all sampled individuals and invite them to call with any questions or concerns about any aspect of the study.
GfK data collection staff will work with SSS staff to address concerns that may arise.
Developing an appealing and understandable survey instrument is important to achieve high response rates. Experts on web survey design were consulted and assisted in the design and testing of this survey. Specific design issues that were incorporated to increase overall and item-specific response rates included:

Reducing the number of complex or open-ended questions by substituting yes/no questions, where applicable.
Streamlining the formatting of questions to be easily read on one screen without the need for scrolling down.
Eliminating all screener questions since needed information is already known about the sample panel.

These efforts promote the highest possible participation rates. While those who agree to be part of the panel may be different, their differences are measurable, and it is less likely that their responses on any particular survey will be different from the population as a whole.

The covered population for the panel recruitment is approximately 97% of the U.S. population as a result of the use of the ABS sampling. The uncovered population consists of persons with post office boxes and rural route addresses; business and institutional addresses (i.e., dormitories, nursing homes, group homes, jails, etc.); military housing; and multi-dwelling residential structures that have only a single address (called a drop point address) and for which there is no unit-level identifying information (mail is internally distributed). For a more detailed discussion of GfK’s sampling scheme, see Attachment G. Best Practices for ICR Supporting Documentation for KnowledgePanel^® Surveys.

B.4. Test of Procedures or Methods to be Undertaken

GfK will assess the technical aspects and functionality of the survey instrument with up to nine panel members and will help identify areas of the survey that were either unclear or difficult to understand. Once this is completed, GfK will create a data file for SSS analytics. This data file will contain diagnostic data on average time of survey completion, survey completion patterns (such as concentrations of missing data), and other aspects related to the proper function of the survey and information on the clarity of item wording. While we do not expect or plan to have any major changes made to the instruments, we will certainly address this need if it arises.

Analysis Plan:

Data will be analyzed for the overall sample as well as population subgroups, applying the appropriate sample weights provided by GfK. Descriptive statistics and statistical models will be analyzed using standard statistical packages such as SAS© (SAS Institute) or SPSS© (IBM), and SUDAAN© (RTI International) to allow for complex sample survey design.

Dissemination and Use of the Results

Analysis results will be reported in the aggregate in the form of a final report and PowerPoint briefings to NIH/NIDDK project staff. Eventually specific analyses will be conducted for submission to peer-reviewed journals and professional conferences related to diabetes research.

Results of the 2011 survey indicated that 50% of the U.S. population 35 years of age and older had one or more risk factors for diabetes; yet only 30% of them felt at risk of the disease. Risk factors that increase one’s risk for diabetes include being older, being overweight, having a family history of diabetes, and being African American, American Indian, Asian American, Pacific Islander or Hispanic American /Latino race or ethnicity; for women, having a history of gestational diabetes also increases risk. Results of the 2011 survey indicated that people with a family history of diabetes were more likely to perceive increased personal risk for diabetes than those without a family history of diabetes. Individuals who were overweight, older, Hispanic/Latino, or African American were not more likely to perceive increased personal risk for diabetes. In 2011, almost three-fourths of people with diabetes strongly agreed that the disease is rapidly increasing in the U.S. People with diabetes also were significantly much more aware of the terms “prediabetes” and “glycosolated hemoglobin or hemoglobin A1C” than they were in 2006. However, for these people, awareness of the message that diabetes can be prevented had not changed significantly since 2008.

These results indicate that NDEP needs to continue to increase awareness of diabetes, its risk factors, and strategies to prevent diabetes among people at risk, with special attention to racial/ethnic minorities, as well as an increased focus on assisting people with diabetes and its complications with better self-management.

An Executive Summary of the 2011 survey results and the trends results across the 3 surveys is included in Attachment B. The results of the 2006 and 2008 public surveys were published in an article in 2009⁸ as well as disseminated at various conferences. The 2011 results and the trend comparisons have been presented at various NDEP stakeholder meetings such as the mid-May 2012 meeting of the NDEP Operations Committee as well as at conference such as the 2012 and 2013 American Association of Diabetes Educators annual meetings and the 2012 Science of Eliminating Health Disparities Summit. Two journal articles are in the submission and review process with expected publication in 2014. We will continue to disseminate the findings from this information collection through meetings with key NDEP stakeholders, at professional conferences and seminars, and through publications.

B.5. Individuals Consulted on Statistical Aspects and Individuals Collecting and/or

Analyzing Data

Roshni Devchand, M.P.H., CHES

Associate Director, Strategic Planning and Research

Hager Sharp Inc.®

1030 15th Street, NW Suite 600E

Washington, DC 20005

Phone: 202.706.7457

Email: [email protected]

Mark Eberhardt, Ph.D.

Epidemiologist

National Center for Health Statistics, CDC

3311 Toledo Rd., Room 6421

Hyattsville, MD 20782

Phone: 301-458-4201

Fax: 301-458-4037

E-mail [email protected]

Martha Funnell, MS, RN, CDE

Assistant Research Scientist

Department of Medical Education

University of Michigan Medical School

1500 E. Medical Center Drive

G1208 Towsley Center, SPG 5201

Ann Arbor, MI 48109

Phone: (734) 936-9237

Email: [email protected]

Linda S. Geiss, MA

Division of Diabetes Translation

Centers for Disease Control and Prevention

MS K10, 4770 Buford Highway NE

Atlanta GA 30341-3717

Phone: 770-488-1056

Email: [email protected]

Sue Griffey, DrPH, BSN, FNP

Vice President and Director, Evaluation Center

Public Health Research Group

Social & Scientific Systems, Inc

8757 Georgia Avenue, 12th Floor, Silver Spring, MD 20910

Phone: 1.301.628.0340

Email: [email protected]

Christine Hunter, Ph.D., ABPP, CAPT USPHS

Director, Behavioral Research

NIDDK, NIH

Phone: 301-594-4728

Email: [email protected]

Lynne Doner Lotenberg, M.A.

Interim CEO

Hager Sharp Inc.®

1030 15th Street, NW Suite 600E

Washington, DC 20005

Phone: (202) 842-3600

Email: [email protected]

Christina Nicols, M.P.H., M.S., M.S.

SVP and Director, Strategic Planning and Research

Hager Sharp Inc.®

1030 15th Street, NW Suite 600E

Washington, DC 20005

Phone: 202.706.7480

Email: [email protected]

Linda Piccinino, MPS, DrPH (ABD)

Senior Researcher, Public Health Research Group

Social & Scientific Systems, Inc.

8757 Georgia Avenue, 8th Floor

Silver Spring, MD 20910

Phone: (301) 628-0244

Email: [email protected]

James Ter Maat, Group Vice President

Public Health Research Group

Social & Scientific Systems, Inc.

8757 Georgia Avenue, 8th Floor

Silver Spring, MD 20910

Phone: (301) 628-0320

Email: [email protected]

The GfK contractor representative is Mr. Joseph Garrett, Senior Vice President, GfK Custom Research, LLC. Mr. Garrett can be reached at (703) 830-0613 or [email protected].

1 Callagaro, M. et al., eds. 2014. Online Panel Research: A Data Quality Perspective. West Sussex, UK: John Wiley & Sons, Ltd.

2 GfK. 2013. GfK Knowledge Panel Design Summary. GfK: Palo Alto, CA.

GfK/Knowledge Networks. 2010. Best Practices for ICR Supporting Documentation for KnowledgePanel® Surveys. Knowledge Networks: Menlo Park, CA.

3 Link, Michael et al. 2009. Building a New Foundation: Transitioning to Address Based Sampling after Nearly 30 Years of RDD. Paper presented at the 64th Annual Meeting of the American Association for Public Opinion Research, Hollywood, FL.

4 GfK. Personal communication, September 24, 2014.

5 MSG, n.d. Pros and Cons of Address-based Sampling. MSG Whitepaper. http://www.m-s-g.com/CMS/ServerGallery/MSGWebNew/Documents/GENESYS/whitepapers/Pros-and-Cons-of-ABS.pdf

Housing Ownership

Internet Access (Household)

6 Knowledge Networks. 2011. KnowledgePanel® Design Summary (Updated November 7, 2011). Menlo Park, CA.

7 Response rate was originally designed for single survey events, not panels; survey completion rate is a way to make panels more comparable (GfK. 2013. KnowledgePanel® Design Summary).

8 Gallivan J, Brown C, Greenberg R, and Clark C. Predictors of Perceived Risk of the Development of Diabetes. Diabetes Spectrum June 20, 2009 vol. 22 no. 3 163-169.

Highlighted text shows text changes from 2012 OMB approval.

49736802 22

File Type	application/vnd.openxmlformats-officedocument.wordprocessingml.document
Author	Sue Griffey
File Modified	0000-00-00
File Created	2021-01-27

NDEP Public Survey OMB SUPPORTING STATEMENT B_29Sep2014 (2)