OMB_reply_FNSfollowup

1. OIRA Stipulation from Part I: “Specific analysis of response patterns and/or response bias for all “check all that apply” questions with more than 6 response options, as well as any web tables requiring horizontal scrolling on the survey instruments used in Part 1.”

FNS Response: “FNS is currently conducting the analyses for the conditions stated above and will update OMB as soon as the result of such analyses is available.”

OIRA Response: The terms of clearance document in ROCIS provides the frequencies of “check all that apply” items fielded in phase 1; however, we do not see an analysis nor recommendation based on these frequency distributions. Please describe how your analysis of response patterns informs survey item content and format for phase 2.

FNS response:

The response patterns to the web survey of State WIC agencies used in Phase 1 do not inform survey item content or format for Phase 2 because the State WIC agency web survey used in Phase 1 will not be used in Phase 2. In fact, we will not be administering any web-based survey in Phase 2. Instead, as described in the supporting materials, the Baseline and Follow-up surveys of WIC participants (Appendices A1 and A2) will be conducted by telephone using Computer Assisted Telephone Interviewing (CATI) software and neither of these surveys includes any item with 6 or more response options. There are some items on these phone-administered surveys in which the interviewer categorizes an open-ended response from the respondent – but these items do not present the respondent with a list of 6 or more options (that is, the interviewer does not read aloud the categories she will use to classify the open-ended response).

Inspection of the frequency distributions that FNS provided previously for “check-all-that-apply” items with 6 or more options in the Phase 1 web survey demonstrates that the distributions were not dependent on the number of response options. FNS would be happy to provide additional analysis if OIRA can specify what kind of further analysis would inform the instrumentation for Phase 2; note, however, that the web-based survey of State WIC agencies will not be repeated in Phase 2, and none of the surveys to be administered in Phase 2 contain any items with 6 or more “check all that apply” response options. Moreover, no web surveys will be administered in Phase 2—thus, response patterns to the web survey used in Phase 1 do not seem relevant to the content or format of surveys in Phase 2.

OMB reply

1. Ok. Please reference in the supporting statement that, since a web survey will not be used for phase 2, the stipulation described in the terms of clearance for phase 1 do not apply.

FNS: Please see pages 36-37 of Supporting Statement Part A

2. The proposal would collect responses from 2-3  local WIC agency (LWA) employees at two time points at 3.5 hours per respondent. A response rate of 100% is anticipated.

a. We see the LWA director, breastfeeding coordinator, peer counseling coordinator, and administrator for each LWA would be invited to complete the LWA interviews. If we understand this correctly, it would seem that much of these responses would be repetitive within the LWA, if questions are asked about agency features and 2-3 employees per LWA would be asked to respond. Would it be preferable to collect one response per LWA, thus reducing costs and burden? Split the LWA instruments so that the relevant questions are only asked of the most knowledgeable person?

b. A response rate of 100% may be ambitious, given the length of the proposed instruments; attrition, and item nonresponse might become an issue. Please provide a justification, citing similar studies, for the anticipated response rate.

c. The LWA instruments and the Peer Counselor Background Instrument seem very long, which may affect both item and unit nonresponse. We recommend streamlining further. Which sections of the proposed instruments might be streamlined further to reduce burden and support response rates?

FNS response:

2a. FNS intends to collect one response per LWA, with 2 to 4 employees providing answers only to those questions for which each is the most knowledgeable respondent. The estimated burden for each employee is as follows:

	Interview during Demonstration Period	Interview during Study Period	Total per respondent
LWA Director	0.5 hours	0.5 hours	1 hours
Breastfeeding Coordinator	0.5 hours	0.5 hours	1 hours
Peer Counseling Coordinator	2 hours	2 hours	4 hours
Administrative Assistant	0.5 hours	0.5 hours	1 hours
Total per LWA	3.5 hours	3.5 hours	7 hours

The total burden per LWA is 3.5 hours (for each of two interviews), but this burden is allocated across different employees as shown above (as shown also in Exhibit A12.1 in Supporting Statement, Part A, p. 34). For example, the LWA Director is only asked to respond to interview questions for which she/he is the most knowledgeable respondent; those questions answered by the LWA Director are not repeated for other respondents. The interview requires one-half hour of the LWA Director’s time. In contrast, the Peer Counseling Coordinator is typically the most knowledgeable respondent for a greater number of questions; as a result, the interview requires two hours of her time, but items which the Peer Counseling Coordinator has answered are not repeated for other respondents. In this way, the burden per respondent is minimized.

Since roles vary greatly by WIC agency (for instance, in some cases the breastfeeding coordinator and peer counseling coordinator are the same person), it is not possible to split up the guide a priori by role. Instead, when scheduling the site visits during which interviews will be conducted, Abt research staff will ask who is the most knowledgeable about each module of the instrument and arrange to collect the information accordingly.

2b. The agencies that will be interviewed will have agreed to participate in the study, have received resources and will have signed a memorandum of understanding outlining roles and responsibilities, which will include participating in these interviews. Therefore, declining to participate in the process study would be greatly unexpected. We have done several similar interviews. In addition, we have done interviews with very high response rates among LWAs that were recruited and not volunteered as is the case for Phase 2. LWA staff interview protocols proposed for Phase 2 were adapted from similar protocols used in Phase 1 in 40 local WIC agencies and with 16 LWAs in the Evaluation of the Birth Month Breastfeeding Changes to the WIC Food Packages (OMB control no. 0584-0551). In both studies, the response rates were 100%. None of the staff members interviewed at any of these 16 LWAs declined to participate, withdrew from the interview or declined to answer any item.

2c. We acknowledge OMB’s concern about the length of the LWA instruments, and we have identified items within each instrument that could be shortened, and items that could be dropped because they were of lower priority than other items or because item content overlapped somewhat with similar items. To streamline the LWA instruments we propose the following (revised instruments are attached):

LWA Staff Interview Guide 1 (Appendix E1):

We have streamlined the LWA Staff Interview Guide 1 further by: eliminating 14 items that were deemed lower priority or somewhat redundant with information collected in another interview item; streamlining 11 items by reducing the number of response options, the number of columns of information to be collected in tables, narrowing the amount of information requested, or replacing an table of specific information requests with an open-ended item; and reducing the burden on respondents by re-ordering the sequence of 14 items so that similar topics are covered within same module of the survey.

LWA Staff Interview Guide 2 (Appendix E2):

From this instrument, we have eliminated 12 items and/or columns of information requested. Half of these these items were redundant with other items in the interview protocol or with data collected from another source. The other half were deemed to be items of lowest priority. In addition we have streamlined (i.e., shortened) 8 other items.

Peer Counselor Interview Guide 3 (Appendix E3):

We have eliminated 10 items from the Peer Counselor Interview Guide and shortened 6 items. Eight items were dropped because they were redundant with other items in the interview guide or because other data sources in the study would provide sufficient information; two were dropped because they were deemed to be of lower priority than all others.

Peer Counselor Background Questionnaire (Appendix B): This one-time questionnaire includes 19 closed-ended items with an estimated maximum time per response of 15 minutes. (Note that Item 1 is included to ensure that respondents have completed the consent form before completing the rest of the self-administered questionnaire.) If OMB believes that this length is overly burdensome, we propose to eliminate two items (# 4and #10), reducing the instrument’s length by 11 percent and reducing the estimated maximum time per response to 12 minutes:

4. Are you paid to work as a Breastfeeding Peer Counselor? (Pay includes wages or a salary for work that you do)

10. Do you have a nursing degree, certification, or license? Mark all that apply.

• Certified nursing assistant (CNA) or nursing diploma

• Licensed practical nursing degree (LPN) or licensed vocational nursing degree (LVN)

• Associate’s degree in nursing (ADN), usually a 2-year degree

• Bachelor of science in nursing (BSN), usually a 4-year degree

• Registered nurse (RN) license

• Nurse practitioner (NP) or clinical nurse specialist (CNS)

• Nurse-midwife (CNM)

A revised Peer Counselor Background Questionnaire (Appendix B) with these two items removed accompanies this response.

OMB reply:

2. 2a.-2c. Ok. Please reference in supporting statements A and B and the instruments. We anticipate that the burden table will also reflect these proposed changes.

FNS: Please see Supporting Statement Part A pages 32-35 which includes revised burden estimates in Exhibit A12.1 (re: OMB item 2a); Supporting Statement Part B pages 31-32 (re: OMB item 2b); and Appendices B, E1, E2, and E3 (re: OMB item 2c).

3. Peer Counselor Interview Guide 3: Study Period proposes either one-on-one or focus group method to collect information.

a. If we understand correctly, the interview guide indicates that it would take about 90 minutes to complete; is this estimate based on the one-on-one method or the focus group method?

b. Why would selection of just one method be preferred for analysis of peer counselor data? Which of the two approaches would you recommend, given the intent of the study?

FNS response:

3a. The estimate of 90 minutes for the Peer Counselor Interview Guide 3 is based on the focus group method.

3b. We have used the focus group method to estimate the maximum possible burden per respondent but intend to use either the one-on-one or focus group method depending on how each LWA implements their existing Loving Support Peer Counseling program and the proposed intervention. In some LWAs, peer counselors are not often in the same location at the same time; under these circumstances we would interview up to 4 peer counselors using the one-on-one method with each peer counselor participating in the interview for 20 minutes; the total burden across all 4 such respondents in an LWA using the one-on-one method would then be 1.33 hours (1 hour, 20 mins). In other LWAs, it may be possible to meet with a group of peer counselors in one location at the same time (e.g., if they are already scheduled to attend an LWA staff meeting the day of the site visit and interview. In sum, neither the one-on-one or focus groups method is preferred; each approach may be used depending on circumstances at each participating LWA. We recommend using the approach that best fits the LWA’s circumstances when the site visit during the Study Period is scheduled.

OMB reply:

3a.-3b. Thank you. Given that any given program may not be able to facilitate a focus group format for the reasons described, and given that the method of information collection may affect comparison of participant responses, we recommend that a one-on-one format be used for all peer counselor interviews.

FNS: For the study, we will conduct one-on-one peer counselor interviews with up to 4 peer counselors per LWA. Please see pages Supporting Statement Part A page 23 and revised burden estimates in Exhibit A12.1.

4. We believe that the Study Enrollment Form is the informed consent material for WIC study participants.

a. Will there be documentation of informed consent? If written consent would not be administered, please provide the rationale for waiver of written consent, given vulnerability of study population?

b. How are WIC study participants informed which arm of the study they will be joining—treatment or control? Assuming the study will be blind, how will information about services received be relayed to the participant?

FNS response:

4a. There will be documentation of informed consent. The informed consent document for WIC study participants was included as Appendix F6 in our original submission. Due to an error, reference to Appendix F6 was omitted from page 13 of our Supporting Statement, Part B and should have been included where shown below (underlined):

“To obtain informed consent for all study activities, each WIC-BPC participant invited to the study by the LWA will be asked to read and sign a form giving their consent to participate in the study (Appendix F6). This consent form will be reviewed by the evaluation contractor’s Institutional Review Board to ensure that it meets the requirements for obtaining informed consent from human participants in research. In addition, verbal consent from each respondent will also be requested prior to the Baseline and Follow-up Surveys.” (page 13)

4b. After the evaluation contractor has completed the Baseline Survey (or completed the maximum number of attempted contacts) with a WIC participant, Abt will randomly assign the WIC participant and inform the local WIC agency of her treatment status. The WIC participant’s assigned peer counselor will then notify the participant of her treatment status—whether she has been assigned to the treatment group or to the control group. Thus, WIC participants will not be blind to treatment status – in fact, part of the intervention includes knowing that their peer counselor will contact them when they are in the hospital for delivery and that once they deliver their baby they will have receive in-person peer counseling within their first 10 days post-partum.

The Consent Form (Appendix F6) specifically describes the random assignment procedure to be used in the study.

OMB reply:

4a.-4b. Ok. Please reference in supporting statements A and B.

FNS: Please see Supporting Statement Part B, page 11 and Appendix F6.

5. The proposal anticipates a response rate of 85% for WIC study participants.

a. We understand that loss due to follow up on account of incomplete contact information would be minimized in the study design. We also acknowledge the proposed nonresponse bias analysis and proposed weighting adjustment would be important where actual response rates are less than 70%. Please cite justification for an expected response rate of 85% from the similar studies.

b. Does your expected response rate account for attrition at time 2 (between consent and interview)? Nonresponse at the LWA level? How would a response rate of 70% affect power calculations to observe 6 percent difference in outcome (one-tail test, p.05, 80% power)?

FNS Response

5a. The Evaluation Contractor (Abt Associates and their subsidiary AbtSRBI) has achieved high response rates with low-income populations similar to the WIC Participant population to be surveyed in this study, and in studies with a design similar to this one (i.e., a randomized control trial or quasi-experimental study where baseline data and follow-up data were necessary to compare two groups). Examples include:

• Assessing Quality of Life Issues in FEMA’s Alternative Housing Pilot Program (AHPP) completed in 2007 for the US Department of Housing and Urban Development. A Household Outcomes Survey was conducted with individuals receiving an AHPP housing unit in four Gulf states with large numbers of residents diplaced by Hurricane Katrina (study design differed by state and was either a randomized control trial or a pre-post design). Among individual leaseholders in AHPP housing, the response rate was 83 percent.

• Impact Evaluation of Upward Bound's Increased Emphasis on Higher-Risk Students (2006) for the US Department of Education. In this randomized control trial, at-risk high school students (living in a low income household was one of several factors defining “at-risk”) were assigned to either a Treatment and Control group and administered a Baseline Survey. Response rate for this survey was 94 percent (the study was subsequently canceled so no follow-up data collection was conducted).

• Impact Evaluation of the U.S. Department of Education's Student Mentoring Program (2005) for the US Department of Education. Students were randomly assigned to a mentoring program or a waiting list. Baseline surveys were collected from 1,300 students across 21 different sites with a response rate of 97 percent; a 93 percent response rate was achieved on the follow-up survey.

• Moving to Opportunity, completed in 2002 for the US Department of Housing and Urban Development. As part of a randomized control trial demonstration project, collected participant data from in 4,248 individual heads of households for a response rate of 89 percent. Participants in the experimental group received vouchers for use only in low-poverty neighborhoods along with agency counseling. Members of a comparison group received Section 8 vouchers for housing in the locations of their choice but did not receive counseling. Members of the control group continued to live in public housing or project-based assisted housing and received no vouchers.

Like individuals who participated in the above studies, participants in Phase 2 of the WIC Peer Counseling study will be low-income individuals who have applied for a benefit or resource available from a social service agency and who have volunteered to participate in a random assignment study. Moreover, women who agree to participate in the study will also have volunteered to participate in breastfeeding peer counseling. Thus, this population is likely to be interested in the survey topic, a significant factor in stimulating survey participation.¹ In addition, researchers will have recent contact information for study participants provided by the participant themselves to the LWA staff who recruit them into the study (i.e., the evaluation contractor will receive contact information at most 7 days after a participant enrolls in the study, and attempts to conduct the Baseline Survey would commence within 8 to 10 days of enrollment).

Although peer counselors maintain contact with women during the interval between the Baseline and Follow-up survey (women in both the Treatment and Control conditions receive regular contact from their peer counselor), upon further consideration of the two to three-month lag between birth of the infant and the administration of the Follow-Up Survey, there is some risk that mothers may lose interest in the study, in peer counseling, or could move out of the service delivery area of their local WIC agency. We have reduced our anticipated response rate to the Followup Survey to between 80 and 85 percent , resulting in an overall response rate of 72 to 76 percent. Our response to Part (b) of this question (below) demonstrates that a response rate as low as 70 percent would still provide a minimum detectable difference (MDD) of 7 percentage points.

5b. The expected response rate assumed attrition of 10 percent between consent and the WIC Participant Baseline Survey (a response rate of 90 percent) and additional attrition of 10 percent between the WIC Participant Baseline Survey and Follow-up Survey (a response rate of 90 percent), providing an overall response rate of 81 percent (.90 x .90). However, because the study design calls for conducting a Followup Survey with WIC Participants who did not complete a Baseline Survey we anticipated recovering 4 to 5 percent of the sample that had not participated at Baseline, leading to a higher anticipated response rate.² As described above in our response to 5a, however, we have revised this expectation and now anticipate an overall response rate among WIC Participants of 72 to 76 percent.

Nonresponse at the LWA level is considered extremely unlikely. All LWAs in the study will have demonstrated a commitment to participation in three ways: first, they will have participated in a grant competition to participate in the study; each LWA will have signed a Memorandum of Understanding (sample provided in Appendix F3) in which they agree to follow study procedures, including the provision of study enrollment information on study participants on at least a weekly basis; and they will have completed a two-month Demonstration Period during which they attempt to deliver the peer counseling intervention to a target number of WIC Participants. In addition, the evaluation contractor will have assigned “site liaisons” who will follow-up with each LWA to ensure the timely receipt of accurate data on study participants. The participation rate of local WIC agencies in two recent studies have been 100 percent (Phase 1 of the WIC Peer Counseling Study, and FNS’s Evaluation of the Birth Month Breastfeeding Changes to the WIC Food Packages (OMB control no. 0584-0551

If the response rate were 70 percent instead of 85 percent, then of the 1,800 WIC Participants consenting to participate, 1,260 would be retained in the analysis sample. With an analysis sample of 1,260 the minimum detectable difference would be 7 percentage points, assuming the following:

i) a balanced design (nC=nT) in which half of the sample (n=630) is assigned to each group;

ii) desired statistical power is 80 percent;

iii) a significance criterion of alpha=.05 will be used to test one-tailed hypotheses;

iv) expected breastfeeding exclusivity and breastfeeding intensity rates in the control group sample of 50 percent;

v) the amount of variation in the outcomes explained by covariates will be 10 percent (R2 = .10).

OMB reply

5a.-5b. Ok. Please reference in supporting statement B with updated power analysis.

FNS: We have incorporate the above discussion of response rates (5a) on pages 24-25 of Supporting Statement B; and an updated power analysis is shown in Supporting Statement B, pages 17-23.

6. Infant and maternal health issues that may affect ability to breastfeed may be more common among WIC participants and their children.

a. How does power analysis and desired sample size take into account the rate in which WIC mothers may not be able to participate in breastfeeding activity. We acknowledge that this outcome is independent of treatment or control, but may affect sample available for analysis.

FNS Response:  

As OMB has noted, health issues affecting ability to breastfeed will be independent of treatment or control. Many infant or maternal health issues affecting breastfeeding among WIC participants likely derive from WIC participants’ (women and infants) low-income status and not necessarily their selection into WIC participation. Further, research suggests that prenatal participation in WIC improves birth outcomes such as low birth weight, a risk factor for not breastfeeding (see Oliveira and Frazao, 2009).³ And in fact, some of the most common reasons why low-income women do not breastfeed include lack of knowledge about the health benefits of breastfeeding. Peer counseling that provides education and support by knowledgeable peers can overcome these barriers.

The power analyses presented on page 18-19 of Supplemental Part B discusses the expected outcomes in the control group of WIC participants (who are subject to the same health issues as the treatment group) for breastfeeding intensity and exclusivity. Exhibit B4 shows the sample sizes required using estimates of 20 and 50 percent rate of exclusive breastfeeding in the control group. We chose to base sample size calculations on a conservative estimate of 50 percent, in order to ensure that the sample was of sufficient size to provide meaningful minimium detectable differences. Therefore, the power analyses and desired sample sizes should not be affected should there be a higher prevalence of health issues among the WIC population that affect breastfeeding outcomes.

OMB reply:

6. Thank you. Please reference in supporting statement.

FNS: Please see page 21 of Supporting Statement B.

7. If control contents vary, effect of treatment may be especially hard to detect.

a. Given the acknowledgement that controls will vary, and that response rates (especially over time), may be lower than anticipated, has the agency considered engaging more than 8 LWAs to participate in this evaluation? Would the anticipated reduction in burden and cost (described above) support adding an additional LWA?  

FNS Response:

The proposed study design is a within-site randomized control trial in which WIC Participants served at each of up to 8 local WIC agencies are randomly assigned to receive either the intervention or the “business as usual” peer counseling services offered by the LWA, where the business as usual condition is a priori defined as one that does not include either of the two enhancements that comprise the intervention. That is, any LWA that already provides either contact with WIC participants when they are in the hospital for delivery and/or an in-person peer counseling visit during the first 10 to 14 days post-partum is not eligible to participate in the study. Other than insuring that the control condition is distinct enough from the intervention condition, variation among different “business as usual” controls is not a concern for the study and typically varies in studies using a similar design (that is, a multi-site RCT where the control condition is explicitly defined as “continue with business-as-usual.” )

The minimal savings obtained from lower response rates to the telephone survey with WIC Participants does not offset the cost of adding an additional LWA. Even for non-respondents, the fixed costs of tracking and attempting to reach all WIC Participants in the study at the time of consent do not change; only the marginal cost of completing two 20 minute surveys with 15% fewer study participants would be saved.  This modest savings does not cover the costs of including an additional LWA, which are substantial; these costs include providing direct funding to the LWA, providing high levels of technical assistance, and conducting additional site visits during the Demonstration and Study Periods to document implementation of the intervention.

OMB reply: 7. Thank you. Please describe how the studies cited (and the current proposed study) will insure that the control condition is distinct enough from the intervention condition. Please modify supporting statement language to this effect (seems like control contents will not be permitted to vary beyond certain parameters that will be described in the supporting statement).

FNS: To ensure that the control condition is distinct from the intervention condition in each site, the evaluation will take three steps:

1. Selecting sites that are not implementing the intervention;

2. Establishing a formal agreement with each site not to implement the intervention with those assigned to the control condition; and

3. Monitoring the delivery of services to treatment and control group members to catch violations of random assignment as they occur—that is, to identify instances where control group members have received intervention services— so that a member of the evaluation team can talk with the site to avoid future violations of random assignment.

We have added discussion of these three elements to Supporting Statements Part A (pages 8 and 10) and Part B (pages 3, 5-6, 15-17).

8. Outcome variable: exclusivity and intensity is measured as “within the past 24 hours.”

a. Given the importance of the measure, consider adding a second question, “within the past 7 days, that is [NAME DATE, DAY OF WEEK].” You may gain analytical leverage with greater variation in responses.

b. We assume both dichotomous and ordered logistic regression models will be used.

FNS Response:

8a. While keeping the “within the past 24 hours,” question, we will add the following item to the WIC Participant Follow-up Survey:

A6a. Thinking back over the last week, that is, since last [DAY OF WEEK], which of the following best describes the kind of milk you fed your baby?

• Breastmilk only [GO TO A6a_2]

• Mostly breastmilk with some formula [GO TO A6a_2]

• Breastmilk and formula about equally [GO TO A6a_2]

• Mostly formula with some breastmilk [GO TO A6a_2]

• Formula only [GO TO A6a_1]

• REFUSED

• DON'T KNOW

The revised WIC Participant Follow-up Survey is enclosed.

8b. OMB’s assumption is correct: Both dichotomous and ordered logistic regression models will be used.

OMB reply:

8a and 8b. Ok. Please fold into instruments.

FNS: We have incorporated this item into the WIC Participant Followup Survey (Appendix A2); the revised instrument is attached. Use of both dichotomous and ordered logistic regression models was previously discussed in Supporting Statement Part B; we have clarified this discussion on pages 29 and 30 of Supporting Statement Part B.

9. The proposal indicates the importance of “matching” peer counselors to WIC participants to support response to program.

a. The proposal seems to indicate that while the degree and rate of congruence between the peer counselors and WIC participants are important to ensure program success, and would be evaluated at a descriptive level. However, these factors would not be integrated in impact analysis for the enhanced treatment. Would it be possible to create a composite variable to capture this information (and reduce parameters in the model), thus accounting for its impact, increasing any variance explained, and possibly reducing the sample size required to detect impact at proposed levels?

FNS response:

The degree of match between peer counselors and WIC participants is of interest to FNS independently of the proposed impact evaluation. Creating the type of composite variable OMB suggests would be difficult because it is not clear how much relative weight to assign to the degree of congruence between various matching factors, such as WIC Participants’ and Peer Counselors’:

• Language;

• Race/ethnicity;

• Highest educational level;

• Household income;

• Number of children living in household; etc.

More importantly, however, there is no evidence on which dimensions of matching are most important to ensuring favorable outcomes for WIC participants or how much explanatory power (proportion of variance) any particular match variable would have. Therefore, we would not recommend reducing the sample size requirements for this study based on the hope that one or more match variables would help explain variation in breastfeeding outcomes.

However, if OMB requests it, we could add an item to the Follow-up Survey that would permit us to calculate the correlations between congruence on each of the matching factors listed above and WIC participants’ perception of their peer counselor. The matching factor with the highest correlation to WIC Participants’ perception of their peer counselor could be included in the impact model. That is, we could add to the WIC Participant Follow-up Survey, the following item (after the current Item B5):

Bx. How strongly do you agree or disagree with these statements about your Breastfeeding Peer Counselor? Do you Strongly Agree, Agree, Disagree, Strongly Disagree, or Don’t know:

1. My breastfeeding peer counselor and I were alike in a number of ways.

2. My breastfeeding peer counselor and I saw things in much the same way.

3. My breastfeeding peer counselor and I had similar values and attitudes.

These 3 items from Turban and Jones’ (1988) scale of perceived similarity⁴ have been used to predict the success of mentor/protégé relationships with Cronbach’s alpha coefficent of.85.⁵ A single “Perceived similarity” score would be calculated as the average rating across the three items (a, b, and c). For each Peer Counselor-WIC Participant pair, a congruence score for each of several demographic characteristics could be calculated (e.g., if peer counselor and WIC participant shared a common language then Congruence_language = 1, else =0). Finally, we would determine the correlations between the congruence scores and perceived similarity. The match factor where congruence yielded the highest correlation with perceived similiarity could be included in the impact model.

OMB reply:

9. Ok. Please fold into the supporting statement and instruments.

FNS: Please see Supporting Statement Part B pages 27 and 31-32. Item B7a has been added to the WIC Participant Follow-up Survey (Appendix A2):

B7a. How strongly do you agree or disagree with the following statements about your Breastfeeding Peer Counselor from WIC? Do you Strongly Agree, Agree, Disagree, Strongly Disagree, or Don’t know that:

	You and your breastfeeding peer counselor…	Strongly Agree	Agree	Disagree	Strongly Disagree	DK	REF
i.	…were alike in a number of ways.						
ii.	…saw things in much the same way.						
iii.	…had similar values and attitudes.						

1 Groves RM, Presser S, Dipko S. The role of topic interest in survey participation decisions. Public Opinion Quarterly. 2004;68(1):2–31

2 The purpose of the Baseline Survey is to improve precision of the impact estimate, but missing data on the Baseline Survey would not interfere with impact estimate because this is a random assignment study.

3 Oliveira and Frazão (2009). The WIC Program: Background, Trends, and Economic Issues, 2009 Edition, Economic Research Report No. 73, U.S. Department of Agriculture, Economic Research Service, April 2009.

4 Turban, D. B., & Jones, A. P. (1988). Supervisor-subordinate similarity: types, effects and mechanisms. Journal of Applied Psychology, 73, 228−234.

5 See Ensher, EA & Murphy, SE (1997). Effects of race, gender, perceived similarity, and contact on mentor relationships. Journal of Vocational Behavior, 50, 460-481; and Polander, E & Schneider, T (August, 2010). The importance of perceived similarity within faculty-faculty mentoring dyads. Paper presented at the 118th Annual Convention of the American Psychological Association, San Diego, CA.