Prepared by James S. Koopman MD MPH
Dept. of Epidemiology,
University of Michigan
Directed to Various Potential Collaborators.
The purpose of this project is to identify changes in STD and
HIV surveillance which will improve disease control. Our focus
is on collection and analysis of partner histories to identify
social settings or subgroups which play key roles in transmission
dynamics and where infections are going undetected or untreated.
This proposal is stimulated by theoretical work perfomed in the
University of Michigan Dept. of Epidemiology on Social Network
Analysis and Transmission System Analysis which indicate that
new approaches to surveillance could be useful. is for initial
efforts in what is intended to become a long term collaboration
of the University of Michigan with the State Dept. of Community
Health, Local Health Departments, and local community groups.
ItSuch collaboratiions require careful construction so that the
University's efforts strengthen rather than drain local health
capacity. Time must be taken to establish common frames of reference
and joint perception of goals. The work to be done this summer
should be viewed in this light. In the next three and a half months
we should pursue a common perception of the problems facing STD
and HIV surveillance and control and we should search for ways
to unify our disparate skills so as to best address these problems.
At the end of the summer we should have created the basis for
further funding from the State, from CDC, and from other sources
such as NIH and foundations.
The rest of this document is organized as follows. After summarizing
both the long term objectives of this collaboration and the objectives
for field work over the summer, my view of the background of STD
and HIV surveillance and control will be presented. This view
is as yet inadequately informed by familiarity with current field
efforts. Presenting them is one step in establishing the communication
needed to put our collaboration on a firm basis. Then I will outline
proposed changes in surveillance and control activities and proposed
activities for the summer.
1 Establish surveillance procedures with a capacity to:
a indicate when partners of infected individuals fall into social, geographic, and behavioral categories which are not being reported as infected at a level corresponding to that which would be expected on the basis of how often they are reported as partners.
b indicate which social, geographic, and behavioral categories
of individuals are most likely to contribute to sustaining transmission
of syphilis, HIV, and other STDs.
2
3 Identify appropriate control strategies for the different types
of situations uncovered by the new methods of analysis for surveillance
data. For example, some situations might call for more intensified
partner notification, some might call for new screening and infection
detection efforts, and some might call for more opportune and
effective administration and supervision of treatment regimens.
These objectives are subject to modification as the result of
input from local collaborators.
1
Describe the surveillance and partner notification systems for
STD and HIV as they currently function in Detroit. Analyse how
these surveillance systems articulate with each other and with
other control programs. Outline other ways that these systems
could potentially relate to each other and to other control progams.
(To be completed by August 1, 1997)
2
Make a list of partner and partnership characteristics which could
be useful in the proposed surveillance system. Integrate these
into pilot data collection forms. (To be completed by July 1,
1997)
3
Evaluate the practicality and reliability of gathering data on
partner and partnership characteristics for all partners of individuals
with newly reported cases of HIV or syphilis. (To be completed
by August 15, 1997)
Community groups tend to focus on issues related to STDs in a
way that keeps STDs the Hidden Epidemic. Community groups may
identify infant mortality, complications of pregnancy or delivery,
reproductive health problems, and even cancers as major problems
without being clear about what a crucial underlying role STDs
play in these problems. The social process of addressing community
problems is often one that seeks to avoid the potential for stigma
or uncomfortable value conflicts that may arise when one deals
with sexuality. The Institute of Medicine (IOM) issued a report
in 1997 called "The Hidden Epidemic" which outlines
these issues. It points out how STDs consume more that 10 billion
dollars a year in health care costs and HIV consumes 7 billion.
One of its major conclusions is that the STD and HIV epidemics
are still on the rise because of a lack of innovative leadership.
The lack of leadership occurs at all levels. The IOM report points
out how crucial it is for community organizations to take a more
active leadership role in addressing STD problems. In addition
it calls for more creative leadership from Public Health officials
and from University investigators. Leadership is needed not only
to find more resources to address this major problem. It is needed
above all to direct the available resources more productively
and to find new alliances that bring together the diverse types
of individuals and organizations that can address this problem.
Evidence from both theoretical and applied investigations have
firmly established the concept that focusing STD and HIV control
efforts to stop key transmissions can greatly increase the cost
effectiveness of control expenditures. While STD and HIV surveillance
and control personnel recognize this need to focus their efforts,
they have not established criteria for field activities which
provide the needed focus. Most HIV and STD transmissions are part
of chains or trees of transmission which will die out on their
own without control efforts. Identification of those infections
which will sustain chains of transmission is needed to focus control
efforts more productively. A major problem, however, is that many
of the infections involved in sustained chains of transmission
go undetected. That is what enables them to be sustained.
One of the major deficiencies of current surveillance and control
efforts is that efforts are measured by the work that is done,
rather than by the work that is left undone. STDs are analogous
to immunizable diseases with regard to how control efforts need
to be measured. With immunizable diseases, counting the number
of immunizations administered does not lead to adequate control.
When the success of immunization programs is judged by the number
of immunizations administered, more immunizations will be given
to already immunized individuals and fewer to the hard to immunize
individuals. That is because it takes far less effort to administer
vaccines to the cooperative than to the uncooperative populations
and one can maximize performance standards by concentrating on
the cooperative patients. But given such programs, the unimmunized
who are in contact with each other will keep the agent circulating
in the population so that it continues to represent a threat even
to the immunized. Note that with immunizations, the first focus
is on unimmunized and then within that group the focus needs to
be upon those unimmunized individuals in contact with each other
so that they can sustain transmission.
Similarly for STDs, programs that are judged by the number of
treatments administered and the number of partners informed are
often seriously misdirected. The easiest infections to identify
and treat and the easiest partners to notify are those most likely
to be in chains of transmission that will die out on their own.
These are individuals who regularly seek medical care or who are
long term partners of infected individuals. Individiuals who do
not regularly seek care and who are not long term partners of
anyone are the individuals most likely to be involved in chains
of transmission that will not die out on their own. Within this
group of individuals, there needs to be some way to direct control
activities to those particular populations where long chains of
transmission are being sustained. We need to be able to identify
the geographic areas, the age groups, the social groups, and the
social settings where undetected infections are sustaining chains
of infection. Then we need to direct socially realistic infection
detection efforts to these groups. We need to be able to prioritize
our efforts to notify and treat partners so that instead of notifying
20 individuals who are part of transmission chains that will die
out on their own, we notify and treat that key individual who
is amplifying the chain of transmission. When we judge efforts
merely on the basis of how many partners are notified and how
many infections are detected through partner notification, it
is inevitable that partner notification efforts will be inefficiently
focused.
We have conducted theoretical analyses at the University of Michigan
which have identified a new approach to STD and HIV surveillance
that can provide the needed focus. There are two cornerstones
to this new approach. The first is to gather more information
on the partners of infected individuals than is currently gathered.
The second is to analyze this data using the interaction based
models of Social Network Analysis and Transmission System Analysis
rather than the standard risk based models of epidemiology and
biostatistics.
The theoretical work needed to identify the best approaches to
the interaction based analyses is incomplete. It is our intention
to pursue that theoretical development jointly with the pursuit
of practical field methods. That way the theoretical work can
be focused to practical issues and more useful field methods can
be designed because the theoretical possibilities will be clearer.
The basic strategy is to find imbalances between characteristics
of the partners of infected individuals and characteristics of
individuals in whom infection is detected. To take an extreme
example, if one were diagnosing and treating only heterosexual
males for an STD and never heterosexual females, there would be
an imbalance that would indicate that a population with infection
is not being diagnosed and treated, namely a population of females.
The basic insight from Social Network Analysis for expanding this
crude approach is that characteristics of the interaction between
two individuals can be very helpful in making the search for imbalances
more productive. For example, it may be that overall there appears
to be a reasonable balance between males and females being treated
with an STD, but when the balance is specified by short or long
periods of courtship, or short or long durations of partnerships,
it may be possible to identify an imbalance for a specific type
of partnership. Because the characteristics of partners that infected
individuals can specify are few, such as age group, gender, and
race, a good number of characteristics of partnerships are needed.
Characteristics of interactions between partners that will be
identically reported by both members of a partnership are needed.
These might include condom use, type of sex acts were performed,
how many different encounters occured with the individual, how
frequent encounters were, how long the courship lasted, whether
money was exchanged, where geographically or socially the partner
was encountered, etc.
But Social Network Analyses using the above type of information
is not enough. Transmission System Analysis is also needed Social
Network Analysis would be enough if the way individuals got into
the data base did not depend upon them having an infection detected
or being the partner of someone with a detected infection. If
we took a random sample of a population and gathered good partner
and partnership information, we would expect to find a balance.
If we did not find a balance, we would know that some group is
inappropriately being left out of our sample. Such a situation
occurred in a study of IVDUs in Southeast Detroit. The IVDU's
intgerviewed reported the characteristics of individuals with
whom they shared needles. In the sample there was a broad mix
of white, hispanic, and African-American individuals with a predominance
of males but many females as well. Needle sharing was characterized
by the number of individuals in the group when sharing occurred.
There were many young white females who were reported as partners
in groups of two but there were no young white females interviewed
who reported sharing in groups of two. This imbalance led to investigations
which uncovered a group of prostitute IVDUs who were not in the
sample.
But a surveillance system does not randomly sample from a population.
It samples from infected individuals who come in for care. Because
transmission is asymmetric between males and females, because
males may transmit more readily to females than vice-versa, imbalances
may arise purely on the basis of transmission dynamics and not
because some group is not getting diagnosed and treated. Of equal
importance in this regard are imbalances in the frequency with
which STDs are symptomatic. Women tend to have more asymptomatic
infections than men. Thus it may be more likely that the key classes
of individuals sustaining transmission but not getting diagnosed
and treated will be women. How detailed and accurate the transmission
system analysis has to be to distinguish informative imbalances
from those due to transmission probability differences is one
issue that is not yet completely resolved. Our impression at the
moment, however, is that the transmission system analysis can
be quite crude and still do the job.
When a true imbalance is noted, when a group of infected individuals
who are not being diagnosed and treated is correctly identified,
the issue remains of whether or not those individuals play key
roles in transmission dynamics or are peripheral to the chains
of transmission that sustain circulation. To address this issue,
more sophistication is needed in the transmission system analysis.
From the field work side, the key issue is what information to
collect and how to use this information to guide control activities.
A complex social network and transmission system analyses cannot
be used by program personnel to guide their activities. There
are two ways to get around this problem. First we can disregard
a lot of the complexity and just use our intuition to address
the problem. Second, we can use the more complex analyses to help
define simpler and more directly applicable analyses that will
do the job. We intend to take both approaches.
Intuition can get us started on the path to identifying what information
will be useful and how that information can be analyzed. Intuition
tells us that if we just monitor the social contexts where infected
indivduals have made sexual partnerships, we could see which social
contexts are always there and which ones may be causing temporary
epidemics. Intuition can help us judge which imbalances between
infected subjects and their partners might have arisen due to
differential transmission probabilities, and which might be due
to not detecting infections. But intuition is not enough. Transmission
systems are complex and non-linear. Such systems are notorious
for deceiving our intuitions. To find field practical methods
of high utility, a more formal analysis is desireable. Simulation
models should be constructed and carefully analyzed. These simulation
models should correspond as far as possible to real situations.
We at the University have been constructing simulation models
which can serve as the basis for this work. We have been using
them to help devise HIV vaccine trials and to determine where
and why standard risk factor analysis will fail to identify risk
factors for transmission of sexually transmitted infections. We
have used these models to demonstrate that vaccine trials must
examine the partners of trial subjects who become infected if
they are to identify the vaccine effects which are moste likely
to stop the HIV epidemic. We have also used them to point out
that standard epidemiological studies are likely to miss important
effects of oral sex on the circulation of HIV. Additionally our
models have helped point out that some of the most controllable
risk factors for infection are not those that increase the risk
of infection in uninfected individuals. Rather they are factors
that increase the risk of transmission from already infected individuals.
To be useful for the evaluation of different approaches to STD
and HIV surveillance, these models need considerable modification.
This is a big task and grants will have to be written to get the
funding to do it.
We propose the following process for exploring changes in surviellance
practices. The first step is to identify current activities and
goals. The next step is to conduct a group thinking process between
investigators and field personnel which will suggest changes purely
on the basis of shared intuitions. Participants in this group
process are likely to have diverse ideas about changes that might
serve to better identify classes of individuals with undetected
infection who are disseminating infection. We do not have a fixed
agenda as to what changes should be pursued. We want that agenda
to come out of a group process. But in order to clarify our thinking,
we present here some of our ideas that are as yet uninformed by
this group process. They should not be taken as dictates as to
what is needed.
Currently there is a great deal of data that is recorded in the
process of HIV and STD infection investigations. Sometimes it
may seem that the major activity in control programs is just recording
this information. Rather than finding a way to diminish the amount
of paper work, we propose a way to make that paper work more productive
for infection control.
While questions about all recent partners are part of current
syphilis and HIV case investigations, data are recorded mainly
on those partners where the possiblity of locating and notifying
that partner exists. We propose that data on partner and partnership
characteristics be standardized for all partnerships, especially
those where there is no chance of locating the partner. That is
because the key individuals whose infections are not being detected
are likely to fall in this latter group. What the minimal set
of data is, and what time frame for partnership formation should
be used are issues to be addressed. Standard syphilis histories
have concentrated on time frames where a partner might have been
a source case or where the subject might have transmitted to a
partner. Most likely that will not change. The time period for
HIV histories is another issue.
Simple tabulation of infected individual, partner and partnership
characteristics is the initial analysis of data proposed. These
tabulations will be arranged so that imbalances between infected
subject and partner characteristics within strata of partnership
characteristics become evident.
We do not propose any specific changes in control activities.
But if we believed that the current control activities were the
optimal ones, we would not feel a need to propose changes in surveillance
activities. In general we believe that there is a need to shift
control activities from intensive partner notification for Syphilis
or HIV and from broad screening for HIV infection to more focused
investigation of the social environment where infection is occurring
and to more focused screening efforts within those social environments
where there is evidence that undetected infections are sustaining
chains of transmission. Procedures to investigate the social environment
have been presented by Richard Rothenberg. These procedures require
a social work orientation to recognizing the barriers to diagnosis
and treatment. They require a shift in focus from infection to
the broader life problems facing individuals who are getting infected
and transmitting infection. They require attention to a whole
series of social and cultural issues involving housing, crime,
incarceration, welfare support, gang involvement etc. that a focus
on infection excludes. If the individuals who are sustaining chains
of transnmission and who are not currently reached by control
services are to be reached, clearly quite different approaches
must be taken than those currently taken. But until the surveillance
system can direct those control activities appropriately, we feel
that most well intentioned and most humane efforts will have a
significant chance of misdirecting resources away from the situations
where key chains of transmission can be most effectively interrupted.
The activities over the summer in pursuit of the objectives under
II B can be divided into field activities, theory development,
and grant writing.
Again this plan is just presented as one alternative that should
be discussed with State officials before proceeding further. It
is intentionally imprecise. The result of each level of field
activity should influence what is to be undertaken at other levels.
It is our intention to pursue field activities at four levels.
Activities at these four levels should be sequential to some degree
and simultaneous to some degree.
The first level is with upper Public Health and General Administration
officials at both the State and Local levels. Ideally this should
reach as high as Governor's office at the State level and City
counsel or County commissioners at the local level. Detroit and
Washtenaw county seem the most appropriate localities on which
to focus. Wayne county might also be considered. Essentially we
are proposing a process of quality improvement. As Deming has
so clearly pointed out, no process of quality improvement is possible
without a firm committment by upper administration. It is at this
level that a process of working out common perceptions and goals
is essential. We must be open to changing the objectives of this
proposal somewhat depending upon the outcome of meetings with
higher level officials.
The second level is with field operational personnel. This would
involve clinic staff, clinic administrators, DIS, and counselors
at the local level. Besides working out an understanding of objectives
at this level, a thorough understanding of what is currently being
done and why it is being done will be needed. It is hoped that
an MPH epidemiologist can work along with field operational personnel
in helping them carry out their responsibilities in order to gain
the needed understanding.
The third level is clients with syphilis or HIV. The nature of
useful partner and partnership characteristics that can be collected
needs to be explored with these individuals. This exploration
will involve two types of activities. It will involve activities
by Public Health personnel such as DIS or counselors to try out
new surveillance procedures. These procedures can be worked out
in conjunction with the University investigators; but they should
be considered part of ordinary work. Another type of activity
will involve interviews and special investigations to be carried
out by University personnel and for which human subjects approval
must be sought. The exact nature of these is yet to be determined.
They are likely to include focus groups in settings where patients
are waiting to be seen and separate interviews with each member
of pairs of individuals who have been identified in the course
of partner notification. This later type of activity will be conducted
to explore the chances that divergent perception of partnership
characteristics might affect the performance of a surveillance
system.
The fourth level involves community organizations. The social
problems that community organizations are trying to address lie
at the heart of failures to identify and control infection in
the key individuals whose transmissions sustain circulation of
syphilis and HIV. Housing, incarceration, unemployment, drugs,
gang activities, racial prejudice and conflict, and other related
problems lie behind the failures of syphilis and HIV control.
Community organizations may be able to provide key insights and
support that could enable these problems to be addressed within
the context of syphilis and HIV surveillance and control activities.
The above field activities are motivated by initial theoretical
explorations which have indicated their potential to be productive.
But those theoretical activities have not been elaborated so that
they can be presented concise formulations. The links between
Social Network Analysis and Transmission system analysis need
to be formalized. The different network analysis methods that
can be applied to this problem need to be defined. Similarly there
needs to be an initial construction of transmission system models
relevant to HIV and syphilis transmission in the populations where
investigations are to be pursued. To this end there must be some
collection and analysis of existing surveillance and partner notification
data. This work is needed in order to present a cogent and convincing
plan for further work which funders will support.
Two types of grants are anticipated. The first will be to CDC
through the State. Dr. Koopman is planning visits to CDC in Atlanta
to discuss such grants over the course of the summer. The second
will be to NIH and will involve further theoretical exploration
of surveillance system approaches.