Alternative Response in Minnesota: Findings of the Program Evaluation

Full text

Protecting Children
Page 78
Volume 20 / Number 2 & 3
By L. Anthony Loman, PhD, and
Gary L. Siegel, PhD
Dr. Anthony Loman and Dr. Gary Siegel are
senior researchers with the Institute of Applied
Research in St. Louis, Missouri.
Presented here are the evaluation findings of
the Minnesota Alternative Response Project.
The Minnesota project, piloted in 20 counties
in 2001 and subsequently expanded statewide,
is an example of a new approach to assisting
families reported for child
abuse and neglect (CA/N) to
child protection services
(CPS). This approach is being
implemented in other states
under various names, such as
dual track, multiple response,
and differential response. The
term alternative response or
AR is used in this article. AR,
with many variations, is now
in effect in about 20 states
(Fluke, et al., 2003; Schene,
2001).
Alternative response defined
At the center of the AR approach is an
alternative way of responding to families
accused of CA/N. The typical, indeed the
nearly universal, mode of response of CPS to
accepted reports of CA/N has been to send an
investigator to the home. CPS investigations
are modeled after criminal investigations.
They are narrowly focused on the incident
alleged in the CA/N report and seek to
determine whether the acts or failures to act
actually occurred (a substantiation or finding),
who was abused or neglected (the victim), and
who was responsible (the perpetrator). The
approach is adversarial, and threats of
punitive actions are implicit. The atmosphere
is accusatory, and for this reason, traditional
investigations evoke anger, fear, and other
negative emotions in caregivers. Because the
object is substantiation, family members are
often approached separately to compare
different versions of incidents. When CA/N
cannot be proven and the
report is unsubstantiated,
little follow-up assistance is
typically provided to families
to address broader and
underlying problems that put
them at risk of future reports.
Under the traditional
system, formal cases are
opened primarily for
substantiated reports.
Families with substantiated
reports are a minority of all
families encountered by CPS.
There are often delays, sometimes of months,
as families are passed from investigators to
service workers. Finally, among open cases,
actual services are delivered only to a minority
of families in greatest need, usually those in
crisis. This is the pattern of the traditional CPS
system. (See Lindsey, 1994, for a description of
the traditional system and the rationale
behind it.)
AR represents a fundamental change in the
manner of responding to families reported for
Alternative Response in Minnesota:
Findings of the Program Evaluation
The atmosphere is
accusatory, and for this
reason, traditional
investigations evoke
anger, fear, and other
negative emotions
in caregivers.

Protecting Children
Page 79
Volume 20 / Number 2 & 3
CA/N. In the Minnesota project, reports were
first screened into two groups. The first group
consisted of a minority of reports that were
inappropriate for AR. These were reports with
allegations involving egregious harm or
imminent danger to children. These reports
received a traditional CPS investigation. The
second group comprised families with reports
that involved less serious threats to child
safety and that were screened as appropriate
for AR.
Changes under AR that differentiated it
from the traditional response included the
following:
1. Families received an AR family assessment
rather than a traditional investigation
(although formal child safety assessments
were conducted in response to all reports).
2. Reports were neither substantiated nor
unsubstantiated; victims and perpetrators
were not identified.
3. The entire family usually met with the
worker during initial assessment visits, and
decision making emerged from group
discussion. However, if the worker felt it
necessary, interviews of children separated
from caregivers were permitted.
4. The full array of family strengths and needs
was considered during the assessment.
Assistance to the family was a consideration
from the time of the first home visit.
Services were premised not on
substantiation of CA/N but on family
welfare and long-term child safety.
5. Through funding provided by the McKnight
Foundation, AR workers could offer
additional services to families. This feature
of the AR project may differentiate it from
alternative response programs in other
states. For example, the earlier and very
similar approach adopted in Missouri was
implemented without accompanying
service funds (Siegel and Loman, 2000).
6. Further contacts and assistance to families
were voluntary.
The AR project was implemented in early
2001. The evaluation began simultaneously,
and the first phase was completed in fall 2004.
Additional tracking of program impacts and
costs will continue through mid-2006. This
article summarizes select findings from the
first phase of the evaluation, which included
process, impact, and cost-effectiveness
components. A wide variety of formal research
questions were addressed in the 2004 report,
which is available online, along with other
ancillary findings (Institute of Applied
Research, 2004). A subset of these are
examined here in the following general areas:
child safety, family engagement, service
changes, recurrence of CA/N reports and later
child removals, family outcomes, worker
responses, and program costs.
Research design and data collection
The primary design for the impact analysis
was a field experiment, conducted in 14 of the
20 project counties, that agreed to permit
families screened as appropriate for AR to be
randomly (although disproportionately)
assigned to experimental or control
conditions. Control families received a
traditional CPS investigation, and
experimental families received an alternative
response. The following is a summary of some
elements of the research design.
• The study population consisted of 2,860
experimental families and 1,305 control
families with CA/N reports between
February 2001 and December 2002. Because

Protecting Children
Page 80
Volume 20 / Number 2 & 3
all experimental and control families had
been first screened as appropriate for AR and
then assigned randomly, it was expected that
the groups would be similar on demographic
and case variables. Comparative analyses
confirmed this.
• In the first phase of the evaluation, families
in the study population were tracked from
February 2001 through March 2004 using the
Minnesota Social Services Information
System (SSIS), from which outcome
measures were derived.
• In addition, experimental and control cases
were sampled to permit collection of
information from workers that was
unavailable in SSIS. This case-specific survey
sample consisted of 690 families: 271
experimental and 207 control, as well as 212
AR families from the six counties that did not
participate in the experimental design.
• Another and larger sample of experimental
and control families was selected for
purposes of obtaining feedback directly from
families. Families were surveyed through
interviews and questionnaires approximately
12 months after their case was closed and at
yearly intervals thereafter. Depending on
when families entered the study during the
assignment period, they were surveyed from
one to three times. Of the 3,866 families
contacted, 1,184 responded to the first
survey, 678 responded to the second survey,
and 413 responded to the third. Finally, a
third sample of 649 cases was selected for the
cost study. Cost data on sample cases
consisted of all expenditures recorded in
county accounting systems, as well as
indirect costs calculated on the basis of
worker time records.
• In 2001, and again in 2004, general surveys
were conducted of CPS social workers in
each local office to determine their attitudes
toward and appraisals of the AR approach,
determine their perceptions of the attitudes
of families in AR versus traditional CPS, and
learn of any operational changes that
occurred or problems that developed.
Responses were received from 115 workers in
the 2001 survey and from 106 workers in the
2004 survey.
• Regular site visits were made throughout the
2001-2004 period to CPS offices in counties
participating in the AR demonstration to
interview social workers and supervisors and
visit community agencies. A majority of all
CPS social workers involved in the AR
demonstration were interviewed, many on
an annual basis.
Major study findings
Child safety
Perhaps the most important finding of the
evaluation was that child safety was not
jeopardized under AR. An assumption implicit
in traditional CPS has been that adversarial
investigations are necessary to ensure children
are protected; that is, that child safety threats
are removed or controlled. A natural fear,
therefore, has been that replacing
investigations with AR family assessments
might threaten children’s safety.
An important and often-used indicator of
long-term child safety in studies of CA/N
response systems is recurrence – does CPS
intervention prevent subsequent incidents (at
least known incidents) from occurring?
Findings related to recurrence are presented
on page 85. In this evaluation, short-term
changes in child safety during the time the
family was in contact with CPS were assessed

Protecting Children
Page 81
Volume 20 / Number 2 & 3
in the initial phase of each research case. The
initial phase can be thought of as the
treatment phase of the field experiment,
during which the experimental families
received AR and the control families received
the traditional approach. Using the case-
specific survey to measure change in the
initial phase, workers were asked to assess
each sample family in 12 safety areas: food
and nutrition, clothing, personal hygiene, safe
shelter, hygienic living situation, health care
and medications, supervision, abandonment
and locking out of home, physical violence,
overly severe discipline, emotional abuse, and
sexual abuse. Workers rated problems at first
contact (mild [1], moderate [2], or severe [3])
and at the time of final
contact (not present [0], mild
[1], moderate [2] or severe [3]).
Safety changes were
calculated as the difference
between final and initial
ratings for each type of child
safety problem identified by
workers.
In some cases, only one
problem was identified. For
other families, more than one
problem was found, although
the number of safety threats rarely exceeded
three. These scores were individually analyzed.
In addition, by averaging all separate scores
for each family, a global safety change score
was generated. Generally, ratings of change in
specific safety areas ranged from no change
(0) to improvements in safety (1 and higher).
Declines in safety were reported only rarely.
While it might be assumed that workers would
be somewhat biased when judging their own
work and unlikely to indicate failures in
protecting children, the analysis was based on
relative differences in safety change
assessments between workers in experimental
cases and workers in control cases.
There was no evidence that AR resulted in
greater declines in child safety among families
in which safety problems were found. Rather,
considering all individual categories of change
in child safety, the percentage of experimental
families that ended with safety improvements
totaled 47.7% compared to 31.8% for control
families. Using the global measure of safety
change, AR experimental families with an
initial safety problem received a score of .99,
compared to .76 for control families. This
amounted to an average improvement in child
safety for both experimental and control
families, but for the experimental families
approached under AR, the improvement was
greater and was statistically significant.
Another measure of
short-term changes in child
safety was the difference
between experimental and
control families in the
number of new CA/N reports
during the initial phase,
while workers were in
contact with families. If
children were less safe under
AR, an increase in CA/N
reports on families provided
with this approach during
these early days might be expected. The
average length of the initial phase was greater
for experimental families (median of 72 days)
compared to control families (median of 52
days), primarily because more experimental
families received post-assessment services
before contact with the agency was concluded
(Figure 1). This meant that, other things being
equal, experimental families had greater
opportunity to be reported again by
caseworkers and service providers. The
difference in the proportion of new families
reported, however, was less than 1%. This
difference was not statistically significant.
Therefore it was concluded that there was no
There was no
evidence that AR resulted
in greater declines in
child safety among
families in which safety
problems were found.

Protecting Children
Page 82
Volume 20 / Number 2 & 3
difference in CA/N report recurrence during
the initial phase, indicating no decline in child
safety.
Family engagement
Engagement is an essential condition for
further cooperation and participation of
families in CPS services. Because the AR
approach was non-adversarial, family friendly,
and voluntary, the atmosphere when workers
visited families was warmer and less
threatening. Caregivers and other family
members were consequently less afraid and
more likely to respond positively to the
worker. Thus, families tended to be more
engaged. Conversely, traditional investigations
were adversarial and included the threat of
punitive consequence of a CA/N finding and
an involuntary case opening. Thus,
investigations were more likely to alienate
families. One reason for the reduction in safety
problems among AR experimental families
may have been the success of AR in engaging
families.
The Structured Decision Making (SDM)
Family Risk Assessment instrument was
completed for each experimental and control
family at the point of initial contact. One of
the items in the SDM assessment related to
caregiver cooperation. On this item, workers
rated the primary caregiver as uncooperative
in 44% of control families, but less than 2% in
experimental families. In addition, the SDM
assessment included ratings of the caregivers’
motivation as well as their assessment of the
seriousness of the report. Both primary and
secondary caregivers in the control group were
rated as less motivated and as viewing the
report less seriously than caregivers in the
experimental group.
Using a different measure of cooperation,
workers were asked to rate experimental and
control families in the case-specific sample
using an 11-point rating scale from -5, very
uncooperative, to +5, very cooperative. The
average level of cooperation of families during
the first visit was 2.4 for AR experimental
families versus 1.8 for control families. At the
last visit, the difference between the
corresponding averages was greater: 2.9 for
experimental and 1.6 for control. Both
differences, however, were statistically
Figure 1. Proportion of experimental and control families with a new child abuse and neglect report during
the initial phase of the research while in contact with the agency and median days of the initial phase

Protecting Children
Page 83
Volume 20 / Number 2 & 3
significant. Moreover, workers were more
likely to report that control parents were
hostile throughout the case (6%) compared
with experimental parents (3%).
The responses of families tended to mirror
those of workers. For example, 58% of families
in the experimental group reported being very
satisfied with the way they were treated by
workers, compared to 45% of control families;
53% of experimental families described the
worker as very friendly, compared to 41% of
control families (Figure 2).
AR increased
participation on the
part of family members,
a direct measure of
actual engagement:
68% of experimental
families said they were
involved a great deal in
decisions that were
made about their
families and children,
compared to 45% of
control families (Figure
3). Differences in
engagement and
alienation were also
demonstrated in the
reported emotional
responses of families.
Following the first visit
with a CPS social
worker, experimental
families were
significantly more likely
to report being relieved
(experimental 34%
versus control 27%),
hopeful (26% versus
20%), satisfied (29%
versus 27%), helped
(21% versus 16%),
pleased (24% versus
19%), reassured (23% versus 16%), and
encouraged (21% versus 17%). On the other
hand, control families significantly more often
reported being angry (experimental 11%
versus control 17%), afraid (9% versus 13%),
irritated (12% versus 20%), dissatisfied (5%
versus 11%), worried (16% versus 23%),
negative (4% versus 8%), pessimistic (3%
versus 6%), and discouraged (7% versus 11%).
Another element of family engagement
was participation of the entire family as a
group. Over two-thirds of AR families (68%)
Figure 3. Degree of involvement in decision making reported by
experimental and control families
Figure 2. Proportion of experimental and control families who described the
way they were treated as friendly or unfriendly

Protecting Children
Page 84
Volume 20 / Number 2 & 3
reported that one or more children were
present during the initial assessment visit
compared to slightly over half of control
families (55%). Among those who were
married, 82% of AR respondents said their
spouse has been present during the
assessment, compared to 65% among the
control group.
Services to families
Experimental families received more
services and types of services than control
families given the traditional response.
Moreover, increased participation of families
in decision making may have increased the
use of services more highly valued by families
themselves. The addition of special funding
for post-assessment services from the
McKnight Foundation was also a factor in the
increased opening of formal service cases, the
vehicle through which case management and
a variety of funded services were provided to
families.
Overall, 36% of experimental families in
the impact study had a formal case opened,
compared to 15% of control families (Figure
4). In the process, post-assessment services
were offered to a wider variety of families.
Using the risk rating on the Minnesota SDM
Family Risk Assessment
tool as a rough
measure of risk, it was
found that cases were
opened for 28% of low-
risk and 41% of
moderate-risk
experimental families,
compared to 3% of low-
risk and 9% of
moderate-risk control
families. At the same
time, more cases were
also opened for high-
and intensive-risk
families under AR (64%) compared to the
control group (57%). These findings show that
AR moved the agency to place more emphasis
on preventive services, but also maintain
emphasis on the traditional response for
higher risk cases and families in crisis.
Information about specific services
provided to families was obtained from
workers responding to the case-specific
survey. Traditional counseling and therapeutic
services were offered more frequently to
experimental families. Increases were also
evident in certain family support services that
were nontraditional in CPS. These were basic
services addressing personal, household, and
other financial needs, including employment
assistance, vocational training, transportation,
TANF and Food Stamps, emergency food,
basic household needs, housing, rent
payments, and daycare. In each of these areas,
services were offered to significantly more
experimental families than control families, in
spite of the similarity between the families in
the two groups. These services are sometimes
related to child safety but more often address
problems of general family welfare and, in this
sense, they are preventive rather than
protective services. This was further evidence
of a shift toward a more preventive approach
Figure 4. Proportion of experimental and control families with service cases
opened after the investigation (control) or family assessment (experimental)

Protecting Children
Page 85
Volume 20 / Number 2 & 3
under AR. Workers further confirmed this
during interviews as they described “services
to meet immediate needs” and “concrete
assistance.” The following quote from a worker
is illustrative:
“We can help AR families maintain
employment with daycare,
transportation, gas money, tools, and
alarm clocks. And help them with some
pretty basic things they need for their
homes and their children, like blankets,
pillows, cribs, vacuums, safety gates,
electrical plugs.”
Responses of families
generally coincided with those
of workers. AR experimental
families reported receiving
various services more often,
including food and clothing,
help with utilities, appliances
and furniture, home repair,
other financial help,
counseling for a child, respite care, and help
with employment or job searches.
Experimental families were provided more
funded services but also received more
services from workers themselves and from
unpaid community resources. Regarding the
latter, according to information received from
workers, significantly more experimental
families received services from emergency
food providers, mental health providers,
support groups, recreational facilities, youth
organizations, daycare and preschool
providers, schools, community action
agencies, job service/employment security,
and employment and training agencies.
While services were expanded to more
families, including lower-risk families, more
services were also provided to higher-risk
families. This was true both for control
families, reflecting traditional CPS, as well as
experimental families offered AR. Whether
they received an alternative or traditional
response to a child maltreatment report,
families that reported more stress in their
relationships with their children, stress in their
relationships with other adults in their lives,
concern about the general well-being of their
family, and concern about the general well-
being of their children were significantly more
likely to have reported receiving services.
However, AR families with seriously ill or
developmentally or learning disabled children
or with caregivers who were experiencing
stress associated with other
adults in their lives were
significantly more likely to
have received services than
corresponding control
families. More generally,
services under AR were also
provided to more of the
lowest income families – one
measure of risk of future
CA/N. According to
information provided by families, 60% of AR
families that received services were families
whose income was below the mean for the
group. On the other hand, among control
families that received services, 52% had
incomes below the mean.
Service increases were also seen across
families from the larger ethnic and racial
communities in Minnesota. While 52% of
white experimental families received services
compared to 36% of the control, the
corresponding percentages for African
American families were, respectively, 63% and
27%, for Hispanic families 67% and 52%, and
for American Indian families 54% and 37%.
Each of these differences was statistically
significant. Only among Asian families were
experimental and control services
comparable: 59% and 64%, respectively.
There was a modest
but statistically
significant reduction in
recurrence among
experimental families.

Protecting Children
Page 86
Volume 20 / Number 2 & 3
Recurrence during the follow-up phase
There was a modest but statistically
significant reduction in recurrence among
experimental families. The absence of new
reports of CA/N is an indirect measure of
improvement in the long-term safety status
and general welfare of children. Information
on new reports was available for all families in
the study through SSIS.
In the analysis, any new report received
anywhere in Minnesota for any child in
experimental and control families during the
period from the final family contact until the
end of data collection was counted. This is
termed the follow-up phase of the research
case. The tracking period of the follow-up
phase varied, of course, because families
entered the experimental and control groups
at different times, and the length of the initial
phase varied (see Figure 1). The statistical
analysis was able to adjust for these
differences (proportional hazards analysis)
and take into account other variables of
importance. The basic finding was that new
reports occurred significantly less frequently
among experimental families that received AR
and that families continued (statistically, they
“survived”) for longer periods before new
reports occurred. The difference in
proportions of families with recurrence was
approximately 3%: 27% of experimental
families had a new report, compared to 30% of
control families.
An additional finding was that the
reduction in recurrence was attributable both
to the new approach and to the increased
services. This finding was particularly
important because it showed that, in addition
to offering more comprehensive services,
positive benefits were achieved –
independently – by changing the way workers
approached families.
In a related analysis, it was also shown that
recurrence rates were reduced for
experimental families in each of the three
largest racial groups: Caucasian, African
American, and American Indian. These effects
of AR did not appear to be related to the racial
or ethnic identity of families.
Finally, another measure of recurrence was
examined: later placement of children in
foster care. The proportions of families in the
study population that had a child placed after
the final contact with the family in the initial
case were relatively small: 11% of
experimental families had one or more
children placed at a later time compared to
13% of control children. However, a similar
statistical analysis (proportional hazards)
showed that the difference was statistically
significant. The AR approach led to a
reduction in later removal and placement of
children. This finding has important
implications for the long-term costs
associated with families under CPS and may
account in part for the positive findings of the
study of AR costs discussed here.
Family outcomes
Other positive changes were found based
on feedback from families. Experimental
families (responding to the first follow-up
survey approximately 12 months after the
initial case had been closed) were more
positive about the services they received.
Caregivers were asked, “If you received some
services or assistance, was it the kind you
needed?” Nearly one-half of experimental
families (48%) responded yes, compared to
about one-third of control families (33%).
They were also asked, “If you received some
services or assistance, was it enough to really
help you?” Again, 44% of experimental
families responded affirmatively, compared to
27% of control families. The differences were
statistically significant and were relatively
large for family groups that were essentially
similar.

Protecting Children
Page 87
Volume 20 / Number 2 & 3
Families were asked to rate changes in the
stress level during the 12 months after the
initial phase in the following areas: financial
outlook, current job/job prospects,
relationships with other adults, relationships
with children, general well-being, well-being
of children, home, and life in general. In each,
a greater proportion of control families felt
more stress, and a greater proportion of
experimental families felt less stress. In two
areas – financial outlook and relationship with
other adults – the differences were statistically
significant. The differences on these two
variables were modest overall and seemed to
result from a sub-set of families that
experienced more intense changes in their
lives. For example, 31% of control families felt
“a lot more stress” about their financial
outlook, compared to 23% of experimental
families. On the other hand, 36% of
experimental families felt “a lot less stress”
about adult relationships, compared to 27% of
control families.
As a follow-up to the issue of reduced
financial stress, two analyses examined family
income and months worked during the
previous year. Differences in family risk were
controlled. Family responses were compared
about one year after final contact in the initial
assessment or service case. The (marginal)
mean income of the control families was
$23,762, compared to $25,497 for experimental
families, a difference just below the usually
accepted level for statistical significance
(.05 < p < .1). Looking at months worked
during the last 12, the means were 7.13 for
control families and 7.23 for experimental
families, a small but statistically significant
difference. These findings on income are
intriguing, but other research is needed,
perhaps using state wage files, to confirm
them in other CPS contexts.
Worker perspectives
With few exceptions, positive attitudes
toward AR grew stronger among workers as
they gained experience with it. A majority of
workers in both interviews and in comments
made in surveys expressed a positive attitude
toward AR. A majority indicated that the non-
judgmental, strength-based, and empowering
approach to families fostered by AR had a
positive effect on their practice. They also
indicated that AR allowed them to focus on
the family as a whole and provide support and
advocacy, as well as more immediate help and
referrals to community services and resources.
Overall, the evidence indicates that the
introduction of AR produced a shift in social
work practice and that the shift was in the
direction intended by program administrators.
This shift was recognized by many workers and
welcomed by most. Beginning in the first year of
the study, workers surveyed responded strongly
that AR had affected their practice. Among
workers who were responsible for AR only (that
is, they did not do investigations), 50% said AR
had affected their approach to families a great
deal. Three years later in the second general
survey, the response of these AR workers was
stronger still, with 69% saying it affected their
CPS practice a great deal. These findings were
reinforced in interviews of workers, none of
whom indicated a preference to abandon AR
and return to the single traditional response for
all reports. The following quotes typify the
attitudes of many workers:
“We discuss safety of the children with
families instead of trying to determine
whether maltreatment occurred or not.
And we approach families as a whole
instead of interviewing each family
member separately. With AR, families
are more involved in the decision
making.”

Protecting Children
Page 88
Volume 20 / Number 2 & 3
“Families are approached
as a unit [not split up to
interview], each person is
heard by the rest,
children's feelings are
revealed, and families
hear what they are doing
well. The approach does
not focus on blame and
wrongdoing.”
Contact between families
and workers increased with
AR. Based on data in the
case-specific sample,
traditional workers were
more likely to have one-time,
face-to-face meetings with
families than AR workers (41% versus 27%).
For those families visited more than one time,
the average number of total meetings was
higher for experimental families (5.4) than for
control families (2.9). This difference, in part,
would seem to reflect the change in family and
worker roles in these encounters and may be
another measure of greater family engagement
under AR.
Increase in family contact affects workload,
at least initially. One in five workers using the
new approach reported large increases had
occurred in their workload and paperwork. By
broadening the scope of initial assessments
beyond the narrow focus of investigations on
maltreatment allegations, encouraging
involvement of family members in decision
making, and changing the criteria for
continued work with families, the AR
approach increased the amount of time and
effort that some AR workers expended per
family. However, a majority of workers
indicated no change or only small increases or
decreases.
Overall, a sizeable minority (44%) of the
workers surveyed said the introduction of AR
made it either a little more or much more
likely that they would remain in this field of
work. Only a few (6%) said it was a little less
likely they would remain in child protection,
and none said it was much less likely they
would stay in the field.
Program costs
The cost analysis found that overall costs
associated with families were lower under AR.
Costs were collected on samples of
experimental and control families and
included service costs (reported by local
bookkeepers) and calculated staff time costs
based on workers’ logs.
Costs were examined during two time
periods for each family: 1) the period from the
initial CA/N report until the final contact with
the family after the initial report – the initial
phase – and 2) the period from the final
contact with the family until the end of data
collection – the follow-up phase (Figure 5).
The average total costs during the first period
were $1,132 for the experimental sample,
Figure 5. Combined direct and indirect costs of experimental and
control families during the initial and follow-up phases

Protecting Children
Page 89
Volume 20 / Number 2 & 3
compared to $593 for the control sample. AR
was more expensive during the early period
because, as has been shown, service cases and
various individual services increased under AR
and because contacts with families were
terminated early for many control families in
unsubstantiated investigations. During the
second period, however, the average total
costs were $804 for experimental families
versus $1,538 for control families. AR was less
expensive during the later period because AR
families had fewer later reports. Consequently,
fewer new CPS cases with their attendant
expenses were opened for AR families, and
fewer children were later removed and placed
outside their homes. The
overall mean costs were $1,936
for experimental families
under AR, compared to $2,131
for control families under the
traditional system. Savings
achieved by experimental
families later more than offset
investment costs incurred
during the initial contact
period. These findings are
encouraging but provisional,
as additional data are being collected on
sample families through 2006.
No effects, organizational issues, and size of
impact
This summary focused on the positive
outcomes of the AR evaluation. While it did
not find instances of strictly negative findings,
where better outcomes occurred among
families given the traditional approach
(among the control families), there were
findings of no difference in outcomes for
experimental and control families on a
number of measures.
Various questions and ratings of child well-
being were asked during follow-up with
families. Based on caregivers’ reports one year
or more after the final contact with CPS
following the initial incident, no consistent
differences were found between experimental
and control children on measures of overall
child well-being, child health, aggressive and
uncontrolled behavior, behavior and
relationships in school, and academic
progress. Similarly, no differences were found
in caregiver reports of improvements or
declines in their relationships with their
children, their methods of disciplining their
children, their ability to care for their children,
their home and living arrangements, or
emotional or financial support from friends
and relatives. Whether the lack of findings was
due to the absence of effects
or to parents’ reticence to
report negative effects is not
known.
There were also various
findings within the process
evaluation of difficulties and
obstacles to implementation
of the new approach. As
noted, the first step in AR
after an initial report is
received is screening. The
screening process directs some families into
traditional investigations and other families
into AR. While investigations are mandatory
when criteria indicating egregious harm or
imminent danger are met (Minnesota
Department of Human Services, 2003), other
criteria are discretionary and give latitude to
local office staff to determine whether a
traditional investigation is warranted. The
proportion of reports screened into AR from
February 2001 through December 2002 varied
from 27% to 61% across the 20 demonstration
counties (excluding the special case of
Hennepin County, where AR was limited to a
single CPS unit). The range in most counties
(14 of 20) was between 45% and 56% screened
into AR. Assuming a rough similarity in the
Savings achieved by
experimental families
later more than offset
investment costs
incurred during the initial
contact period.

Protecting Children
Page 90
Volume 20 / Number 2 & 3
types of reports received among sites, these
variations indicate two things. First, the
confidence of local staff in the AR approach
varied among offices during the early days of
the demonstration. Some counties were much
more cautious than others. A consequence
was that AR populations in counties that
screened a higher percent of reports into AR
contained more families with more intense
child safety threats (Institute of Applied
Research, 2004, 107). Given the generally
positive effects of AR across all demonstration
counties, very conservative screening that
directs fewer families into AR could be
expected to reduce the positive effects of the
approach. Second, because we also saw
variations in screening in the earlier Missouri
demonstration that were still present when a
five-year follow-up was conducted (Loman
and Siegel, 2004), greater consistency in
training of local personnel in screening,
clearer and more reliable screening criteria, or
both may be needed.
The particular way in which AR was
organized varied from office to office. This
topic goes beyond the scope of this article but
was treated in some detail in the evaluation
report (Institute of Applied Research, 2004, 11-
16). Organizational structure was influenced
by the size of county staffs and the pre-
existing organizational structure within local
CPS offices. Organizational differences
affected continuity of services; separation or
integration of the roles of CPS investigator, AR
assessment worker, and ongoing case
manager; and integration of case management
and service delivery of public workers with
that of community agencies. The
organizational choices in each of these areas
may have both positive and negative
consequences for implementation of an AR
program.
A final procedural issue should be
mentioned. The Minnesota program, like the
Missouri program that preceded it, allowed for
change of track. A report that was screened for
an AR family assessment might be changed to
a traditional investigation if the worker and
supervisor felt that the family situation
warranted such a change. Similarly, it was also
possible to change from traditional to AR if the
worker determined that the situation was less
serious than screeners had assumed. In
Minnesota, the former occurred for less than
5% of reports, and the latter in less than 1%. In
light of the screening variability among
counties, we might expect more track
switches, particularly from investigations into
family assessments. The highest percentages
would be expected in counties with the most
conservative screening proportions. If it is
assumed that investigations should be
minimized, the reasons for the low percentage
of track changes need further study.
Finally, the important differences
described in this study have been statistically
significant but modest in absolute size. The
term significant means that an observed
difference likely was not an illusion. The term
modest means a major shift in the variable of
interest was not found. This indicates that the
system did not undergo revolutionary change
but was nudged or moved slightly in a new
direction. This is sometimes seen as a negative
finding, particularly when “modest” is taken to
mean “inconsequential.” The definition of
modest, however, depends on one’s
perspective. An example of subsequent
reported maltreatment of children can be
considered. For the 2,860 families that were
being followed, there was an estimated 3%
reduction in new child maltreatment reports.
In numeric terms, this translates to about 86
families that did not have a new report but
would have had at least one new report under
the traditional approach. The number of new
child abuse and neglect incidents that were
avoided was higher because most families that

Protecting Children
Page 91
Volume 20 / Number 2 & 3
are reported two times are reported again.
However, thousands of families similar to
those in the experimental and control groups
are reported each year in Minnesota, and
hundreds of thousands are reported
nationally. In this context, the modest
difference produced by the change in
approach to families would translate into
thousands of families in which reported
children maltreatment would not recur.
Discussion
Overall, the evaluation findings in
Minnesota were positive both in instrumental
outcomes such as assistance to families and
family and worker attitudes, as well as in
measures of child safety and child and family
welfare. The results are made
more convincing by the
randomized experimental
design of the study. Two
findings may be emphasized.
First, positive results can
be achieved in CPS
interventions through greater
discrimination in how families
are approached. Family-
friendly practice produces not
only more cooperative,
engaged families, but also a
greater degree of child safety
with less cost and greater satisfaction among
both families and social workers. This may be
seen as support for proactive, strength-based,
family-centered approaches. Indeed, AR can
be thought of as a method of initiating family-
centered practice from the very first meeting
with the family.
Second, a finding of fundamental
importance in this evaluation concerns the
value of prevention. The preventive services in
the study could not be described as primary
but as targeted and as secondary or tertiary.
Nonetheless, they are preventive in that they
address fundamental risk factors in CPS
families. AR appeared to shift the CPS system
toward prevention in three ways: the number
of families that received some assistance
increased; attention to low-risk families
increased; and family support services
directed toward basic, financially related
needs increased. These findings are
encouraging because they document an
increase in positive results among families
(mostly low income) that normally would be
ignored by CPS. At the same time, the findings
are dismaying in that CPS is not currently a
family welfare agency, and CPS workers who
are generally overburdened with serious
protective services cases cannot do extensive
work in other areas of child and family welfare.
Nor does CPS, with limited
financial resources in most
states, have the capacity to
fund these services at levels
needed by families, even if
such funding pays off in the
long run. This is the dilemma
that many have referred to
regarding differential
response reforms (for
example, see the discussion
in Waldfogel, 1998, 87-93).
On the other hand, CPS is
the only agency with workers
in virtually every county and municipality in
the nation who can contact families at their
homes to offer assistance. Those who say that
dealing with dirty and dilapidated homes and
improperly clothed and fed children should
not be the responsibility of CPS (see for
example, Pelton 1991, 1998) need to answer
this question: If not CPS workers, then who?
From this perspective, serious thought should
be given to expanding (rather than restricting)
CPS workers’ roles to include other forms of
assistance to impoverished families, returning
perhaps to the more integrated approach to
child welfare employed before CPS became a
On the other hand,
CPS is the only agency
with workers in virtually
every county and
municipality in the nation
who can contact families
at their homes to offer
assistance.

Protecting Children
Page 92
Volume 20 / Number 2 & 3
separate specialized agency after 1970.
Regardless of whether CPS expands further
into child and family welfare, the findings
support the idea of partnerships with other
agencies and organizations in a concerted
effort to address child and family welfare
needs. The findings of this study suggest that a
broader service emphasis and changed
orientation to families may not only reduce
future CA/N but also be less costly in the
longer term. And, if benefits such as cost-
reductions can be demonstrated considering
only CPS service and administrative costs, a
full cost-benefit analysis that considered other
benefits, such as those accrued from increases
in family earnings and child safety, would
show even more benefits relative to costs.
References
Fluke, J., Harper, C., Parry, C., & Yuan, Y. T.
(2003, April). National study of child protective
services systems and reform efforts: Review of state
CPS policy. Retrieved from http://aspe.
hhs.gov/hsp/cps-status03/state-policy03.
Institute of Applied Research. (2004).
Minnesota Alternative Response evaluation: Final
report. Retrieved from: http://www.iarstl.org/
papers/ARFinalEvaluationReport.pdf.
Lindsey, D (1994). The Welfare of Children. New
York: Oxford University Press.
Loman, L.A. and G.L. Siegel. (2004). Differential
Response in Missouri after Five Years. Retrieved
from: http://www.iarstl.org/papers/
MODiffResp2004a.pdf.
Minnesota Department of Human Services.
(2003). DHS issues amended guidance on
Alternative Response to reports of child
maltreatment. Bulletin #03-68-02. Retrieved from:
http://www.dhs.state.mn.us/main/groups/publicati
ons/documents/pub/DHS_id_004859.pdf.
Pelton, L. (1991). Beyond permanency
planning: Restructuring the Public Child Welfare
System. Social Work, 36(4), 336-43.
Pelton, L. (1998, Spring). How we can better
protect children from abuse and neglect:
Commentary 3. The Future of Children, 8(1), 126-29.
Schene, P. A. (1998). Past, present, and future
roles of child protective services. The Future of
Children, 8(1), 23-38.
Schene, P
. A. (2001, Spring). Meeting each
family’s needs: Using differential response in reports
of child abuse and neglect. Best Practice Next
Practice: Family-Centered Child Welfare. Retrieved
from: http://www.cwresource.org/Online%20
publications/Spring%202001.pdf
Siegel, G.L. and Loman, L.A. (2000). Missouri
Family Assessment and Response impact evaluation:
Digest of findings and conclusions. Retrieved from
http://www.iarstl.org/papers/MoFamAssess.pdf.
Waldfogel, J. (1998). The future of child
protection: How to break the cycle of abuse and
neglect. Cambridge: Harvard University Press.