Attachment 15 - HRSA Response to OMB Comments_Final

25 June 2009
TO:

Karen Matsuoka

FROM:

Amanda Cash, HRSA Reports Clearance Officer

SUBJECT:

HRSA Bureau of Primary Care Patient Survey – Response to Comments

The following responses are based on a phone call that took place on Tuesday, June 16, 2009.
1.

Categorizing respondents who are eligible for more than one group.
All patients will receive three questions during the screening process that will
classify them based on their farmworker status, their homeless status, and their
public housing status (yes or no for each). (A revised Patient Screening Form is
attached.) The survey has a target number of interviews for each type of patient
(farmworker, homeless, public housing resident and general community patient).
The allocation for the targets are 826 interviews for farmworkers and their
dependents served by Migrant Health Centers, 826 interviews for individuals
experiencing homelessness served by Health Care for the Homeless clinics, and
660 interviews for public housing residents served by public housing primary care
clinics. The public housing has fewer interviews because the public housing
population is much smaller than the health care for the homeless and migrant
health center populations. In the 2007 Uniform Data System (UDS), there were
776,191 patients from the Health Care for the Homeless program, 775,106
patients from the Migrant Health Center program, and only 133,518 patients from
the Public Housing Primary Care program.
The targets were established to allow the comparison of important survey
outcomes between the different types of patients and allow more in-depth
comparison between different sub-groups within general community patient with
reasonable statistical power. These target figures are equally distributed among
the participating grantees so that each grantee also has target(s) for the type(s) of
patients to be interviewed at its site(s). Each patient interviewed will be counted
against the target for only one patient type. Based on the UDS data submitted by
grantees for 2007, we know that the health center patient population has more
general community patients than homeless patients, more homeless patients than
farmworker patients, and more farmworker patients than public housing resident
patients. Therefore, we assume that it will be hardest to meet the target for the
public housing residents, next hardest to meet the target for the farmworker
patients, next hardest to meet the target for the homeless patients, and easiest to
meet the target for the general community patients.
Only patients that are not classified as homeless, farmworker or public housing
resident by the screening process will be counted toward the target for general
community patients. In order to maximize the likelihood of meeting all targets,
patients classified by the screening process as more than one type of special

population (homeless, farmworker or public housing resident) will be counted
toward the target that is considered hardest to meet. The interviewer has the
targets for each site where he/she will be working. When a patient agrees to do the
survey, the interviewer will go through the screener questions to determine which
target the patient should be counted toward. For example, if a grantee has funding
to serve farmworker patients, homeless patients and public housing residents, the
process of determining which target to count the patient toward will work as
follows. First, the interviewer will go through the screener to determine if the
patient can be classified as a public housing resident, can be classified as a
farmworker patient, and/or can be classified as a homeless patient. If the patient
is a public housing resident and that target has not yet been met, the patient will
be counted toward that target. Otherwise, if the patient is a farmworker patient
and that target has not yet been met, the patient will be counted toward that target.
Otherwise, if the patient is a homeless patient and that target has not yet been met,
the patient will be counted toward that target.
2.

Remuneration to respondents
As discussed in the supporting statement, respondents will be provided with
remuneration valued at $25 for taking part in the interview. Project staff will
consult with the site staff to determine their preferred form of remuneration,
which may include cash or one of the following alternatives to cash: visa gift
cards, food vouchers, telephone cards, personal hygiene bags, and movie tickets.
We have modified the recruitment script, brochure, and consent forms to clarify
this. For proxy interviews for child respondents aged 12 and younger, the
remuneration will be provided to the Parental/guardian who responds on behalf of
the child. Revised documents are attached.

3.

Documentation of Current IRB Approval
RTI’s IRB approval was renewed in March, 2009. Documentation is attached.

4.

Collecting data on race
We have updated the questionnaire to be consistent with OMB guidelines for
asking about race; the interviewer will no longer read “Other” as one of the
response categories. In addition, the question about race has been removed from
our recruitment guidelines, as this information is not needed for sampling or data
collection planning purposes. Revised documents are attached.

5.

Response rates from 2003 Health Care for the Homeless and 2002
Community Health Center User Surveys
Final Response Rate for 2003 HCH Survey

Sample Category
Total sample selected

Number of
Users

Refusals, breakoffs, and other nonresponders
Total completed interviews

% of
Eligibles

1,444

100.0%

11

0.8%

1,433

99.2%

100.0%

416

28.8%

29.0%

1,017

70.4%

71.0%

Ineligible cases
Eligible cases

% of Total
Sample

Final Response Rate for 2002 CHC Survey
Sample Category
Total sample selected
Ineligible cases
Eligible cases
Refusals, breakoffs, and other nonresponders
Total completed interviews

Number of
Users
3,465
678

% of
Total
Sample

% of
Eligibles

100.0%
17%

3,120

82.2%

100.0%

991

26.2%

31.8%

2,129

56.1%

68.3%

6.

Cognitive Interviewing
The cognitive interview process identified a number of issues that have been
resolved in the current version of the questionnaire. These revisions largely fall
into the following categories:

Errors in skip logic were corrected;

Questions that adolescents were not able to answer were identified, and
the questionnaire was revised to skip over these for adolescents;

Some series of questions that seemed overly long were shortened;

Questions that seemed repetitive were eliminated;

For some terms that patients did not understand; explanations were added;

Show cards were added for questions that needed them;

Lengthy lists of response categories were shortened.
The reports from the Round 1 and Round 2 cognitive interviews are attached.
7.

Geographic eligibility criteria for health centers
We will exclude sites that are more than 100 miles away from the central location
in order to have all sites within a reasonable distance for one field interviewer to
cover. (The central location is usually the site that serves the largest number of
patients.) We believe that removing such sites from site sample frame should
have very minimal impact on the patient sample because sites tend to be clustered
within a small area. So far, we have recruited 68 grantees for the survey which
have a total of 406 sites. Only 4 of the 406 sites are more than 100 miles away

8.

from the central location. When a site is determined to be more than 100 miles
away it does not render other sites from that grantee ineligible, only that particular
site.

Patient Sampling Procedures
Selection of patients will be facilitated by the site receptionist. The receptionist
will tally all patients who register for services during the time that the interviewer
is at the site to conduct interviews. When the interviewer signals he or she is
ready to meet a new patient, the receptionist will select the next patient who
registers for services provided that they meet the following two criteria: 1) has
previously received services within the past 12 months and 2) is not an
unaccompanied minor. The receptionist will read a brief script, hand the patient a
study brochure, point them in the direction of the interviewer, and tally that
patient as both “registered for services” and “referred to interviewer”. If the
patient chooses to approach the interviewer, the interviewer will take them to a
private location to conduct a screening. If the patient screens in as eligible and
additional interviews are needed at that site for a patient type that the patient
represents (i.e. H, PH, M, or CHC), the patients will be invited to participate in an
interview. The interviewer will only screen patients who have been referred by
the receptionist and then choose to approach the interviewer.
A limitation of the study is that the patients’ selection method is not entirely
random. A selection method that selects every nth patient for an interview is
probably not feasible for those sites in which the volume of targeted patients is
very low (such as a site serving multiple types of patients, including a very low
volume of public housing patients or homeless patients). At a “low-volume” site,
sampling patients at an interval could result in wait times of several days before
the nth patient arrived. In addition, this is the selection method that was used for
the 2002 Community Health Center and the 2003 Health Care for the Homeless
User Surveys.

9.

Power analysis
We recognize that some of the subgroup analyses have very low power to detect
differences of 10%. Due to budget and sample size limitations, we are not able to
increase power to those subgroups. However, the contractor will work closely
with the Project Officer to ensure the analyses are meaningful and there is
sufficient power to detect differences. We will evaluate subgroups or areas where
larger MDE (minimum detectable effects) are useful and where the power is
above 80%. We will strongly caveat areas where low power or large MDEs exist.

10.

Plan for addressing non-response bias
We will address potential non-response bias; however, we recognize that we will
not be able to analyze those potential respondents who say no initially to the
receptionist and do not provide the information collected through the screener.
We will collect patient characteristics, such as age, race and gender during the FI

screening process. (Revised screener attached.) These patient characteristics are
likely to be related to the survey outcome variables. Thus, using the patient
characteristics available both for respondents and non-respondents, the
respondents and non-respondents can be compared across various attributes to
approximate non-response bias.