National Institute on Drug Abuse
National Institutes of Health
"The Measurement of HIV Risk Behaviors for Treatment
Research Studies"
October 21, 1997
SUMMARY
Debra S. Grossman
Background
NIDA funded researchers with clinical research projects are strongly
encouraged by NIDA staff to provide HIV risk reduction interventions. A
number of different risk assessment questionnaires and interviews have been
used to measure the effects of these efforts. However, many of the existing
instruments have problems, including: 1) insufficient coverage of such issues
as the relationships between sex and drug-related risk, (2) extremes of
complexity/simplicity and length of time for administration, especially
in regards to suitability for re-administration for ongoing risk monitoring
and (3) lack of information regarding reliability and validity. In addition,
the advantages and disadvantages of the two forms of administration, self-administration
versus interviewer-administration, have not been well specified.
When we queried our grantees about their methods of measuring HIV risk
behaviors, we found that several were using the same few instruments. Many
were unaware of the lack of psychometric data available for these tests,
and equally unaware of other available instruments. In light of the fact
that NIDA staff is encouraging HIV risk reduction activities, it is essential
to measure the outcome of such activities in a valid and reliable manner.
Thus, we felt that an examination of these instruments and their psychometric
properties was necessary for meaningful research on HIV risk reduction in
the drug abuse treatment setting.
Purpose of Workshop
In order to advise NIDA staff of existing instruments and their psychometric
properties so that they may be able to make informed recommendations, the
Treatment Research Branch sponsored a 2-day workshop on September 11-12,
1997. Ultimately, it is hoped that information gleaned from this workshop
will facilitate the integration of HIV risk measurement in treatment research
studies so that researchers can accurately assess the efficacy of HIV risk
reduction activities.
Structure of Workshop
Four types of researchers were invited to the meeting:
- ten individuals who have developed or who were involved in the development
of HIV risk behavior instruments for use in treatment research studies
presented their instruments to the group and discussed psychometric data,
or lack thereof;
- four drug abuse treatment researchers who have practical experience
using the available instruments commented on the pros and cons and practical
utility of the most widely used instruments;
- three sex risk researchers shed light on measurement of sexual risk
behaviors of specific populations; and
- three psychometricians commented on methodological issues in the measurement
of risk behaviors.
Measurement Principles (common themes):
Several issues were discussed repeatedly during the workshop and were
generally agreed upon by the majority of the participants:
- Collect psychometric data. This was one of the most important
recommendations. If the instrument is not reliable and valid, the data
collected may be considered worthless.
- Questions must be sensitive to gender and cultural differences.
HIV risk is different for men and women, and different cultural groups
and thus, questions should be gender and culture specific. Consider
a core instrument with alternative items. The group recommended that
perhaps a brief core instrument could be identified with add-on questions
for different populations, i.e., women, gay men, cultural groups, etc.
- Wording is important and terms should be defined. Many subjects
may not self-identify as gay or bisexual, although they may engage in sexual
behavior manifested by these groups.
- Reading level should be considered. Valid data will not be
collected if the questions are incomprehensible to the subjects.
- Perceived risk should be considered. A strong relationship
exists between behaviors and perceptions, therefore perceived risk as well
as attitudes, beliefs, and self-efficacy should be assessed, as well as
behavior.
- Time frame assessed should not exceed six months. Recall
is greatest within a six month time frame. Assessing more that a six month
time frame produces estimates, not exact numbers. Additionally, recall
is facilitated by assessing risk by partner (casual, regular, once
only), one at a time.
- Most sensitive questions should be self-administered. Not only
do self-administered questionnaires eliminate effects of social desirability,
but they eliminate interviewer bias and are more efficient. Alternatively,
interviewer-administered questionnaires enable probing for clarification.
- Computer-administered interviews should be evaluated. Use of
the computer to collect HIV risk information reveals higher rates of divulgence
of risky behaviors.
- Some overall basic recommendations of the group when developing an
instrument was to keep it brief but comprehensive, focused and simple,
reliable, valid, easy to score, and sensitive to change. Avoid the
issue of assessment becoming the intervention.
Action Plan:
- An expanded executive summary of the workshop is currently being prepared.
The summary will be posted on the NIDA website and distributed to the
field. The summary will be helpful to researchers conducting drug abuse
treatment and AIDS risk reduction research, by informing them of the currently
available instruments and associated psychometric properties of the instruments.
Researchers will thereby have more information when making a choice.
- A follow-up meeting of DSCR and Office on AIDS staff is planned to
discuss the possibility of the following:
- convening an expert panel to determine which of the available instruments
are worthy of recommendation now.
- providing administrative supplements to grantees to (1) conduct psychometric
testing of instruments they are using; (2) compare two different instruments;
or to (3) compare a brief version and an extensive version of an instrument.
- hiring a contractor to develop a new instrument, perhaps a core instrument
with add-on questions for specific populations.
|