HITS: A short domestic violence screening tool for use in a family practice setting

Abstract

Domestic violence is an important problem that is often not recognized by physicians. We designed a short instrument for domestic violence screening that could be easily remembered and administered by family physicians.
In phase one of the study, 160 adult female family practice office patients living with a partner for at least 12 months completed two questionnaires. One questionnaire was the verbal and physical aggression items of the Conflict Tactics Scale (CTS). The other was a new four-item questionnaire that asked respondents how often their partner physically Hurt, Insulted, Threatened with harm, and Screamed at them. These four items make the acronym HITS. In phase two, 99 women, who were self-identified victims of domestic violence, completed the HITS.
For phase one, Cronbach's alpha was .80 for the HITS scale. The correlation of HITS and CTS scores was .85. For phase two, the mean HITS scores for office patients and abuse victims were 6.13 and 15.15, respectively. Optimal data analysis revealed that a cut score of 10.5 on the HITS reliably differentiated respondents in the two groups. Using this cut score, 91% of patients and 96% of abuse victims were accurately classified.
The HITS scale showed good internal consistency and concurrent validity with the CTS verbal and physical aggression items. The HITS scale also showed good construct validity in its ability to differentiate family practice patients from abuse victims. The HITS scale is promising as a domestic violence screening mnemonic for family practice physicians and residents.

Full text

508
July-August 1998
Family Medicine
In the United States, 8–12 million women are vic-
tims of domestic violence from current or former part-
ners.
1
Domestic violence is related to serious mor-
bidity and is a major public health problem in our
society.
1
Recent research has found, however, that
only a small percentage of victims of domestic vio-
lence are identified in medical practice.
2,3
Screening for domestic violence by physicians is
done infrequently for a variety of reasons.
2-4
We be-
lieve that one of the main reasons is that existing in-
struments are time-consuming to administer and com-
plete. For example, the Wife Abuse Inventory
5
has
40 items, and the Conflict Tactics Scale (CTS)
6
has
multiple scoring protocols and requires the purchase
of a scoring manual. Hence, physicians may be de-
terred from using these instruments in busy clinical
settings. In the present study, we developed and tested
a short domestic violence screening tool that could
be suitable for use in office practice.
Methods
Overview
Instrument development began by assembling a
focus group of family physicians
7
to discuss the type
and wording of items that would comprise a useful
screening instrument. The group decided that the items
should be few in number and focused on verbal abuse
and physical violence. The group ultimately decided
on four items that would ask a patient to indicate how
often her partner physically Hurts, Insults, Threatens,
and Screams at her. Collectively, these items can be
remembered by the acronym “HITS.”
The research was conducted in two phases. In phase
one, the reliability (ie, internal consistency) and the
concurrent validity of the HITS instrument were as-
HITS: A Short Domestic Violence Screening Tool
for Use in a Family Practice Setting
Kevin M. Sherin, MD, MPH; James M. Sinacore, PhD; Xiao-Qiang Li, MD;
Robert E. Zitter, PhD; Amer Shakil, MD
From the Department of Family Medicine, University of Illinois at
Chicago.
Background and Objectives: Domestic violence is an important problem that is often not recog-
nized by physicians. We designed a short instrument for domestic violence screening that could be
easily remembered and administered by family physicians.
Methods: In phase one of the study,
160 adult female family practice office patients living with a partner for at least 12 months com-
pleted two questionnaires. One questionnaire was the verbal and physical aggression items of the
Conflict Tactics Scale (CTS). The other was a new four-item questionnaire that asked respondents
how often their partner physically Hurt, Insulted, Threatened with harm, and Screamed at them.
These four items make the acronym HITS. In phase two, 99 women, who were self-identified
victims of domestic violence, completed the HITS.
Results: For phase one, Cronbach’s alpha was
.80 for the HITS scale. The correlation of HITS and CTS scores was .85. For phase two, the mean
HITS scores for office patients and abuse victims were 6.13 and 15.15, respectively. Optimal data
analysis revealed that a cut score of 10.5 on the HITS reliably differentiated respondents in the two
groups. Using this cut score, 91% of patients and 96% of abuse victims were accurately classified.
Conclusions: The HITS scale showed good internal consistency and concurrent validity with the CTS
verbal and physical aggression items. The HITS scale also showed good construct validity in its
ability to differentiate family practice patients from abuse victims. The HITS scale is promising as
a domestic violence screening mnemonic for family practice physicians and residents.
(Fam Med 1998;30(7):508-12.)
Clinical Research and Methods

509
Vol. 30, No. 7
sessed with a group of female patients who were vis-
iting their family physician. The CTS was chosen to
establish concurrent validity because it is the instru-
ment most widely used to measure marital violence.
8
In addition, the CTS assesses both the severity and
chronicity of that violence. Moreover, the CTS has
been used in three nationally normed studies
9-11
and
has a high level of internal consistency, concurrent
validity,
12-14
and content and construct validity.
8
The
CTS also has been found to correlate well with spouse
reports of domestic violence.
13
In phase two, the construct validity of the HITS
was tested by comparing the responses of the partici-
pants in phase one (general patients visiting their phy-
sician) with the responses of self-identified victims
of domestic violence. If the HITS is a useful screen-
ing tool, victims of violence should score higher than
the general population of patients. Moreover, it should
be possible to find a cut score that reliably differenti-
ates victims of violence from patients in general.
Instruments
The CTS. The CTS contains 15 items that measure
perception of verbal and physical violence; all 15 of
these items were used in this study. The CTS’s four
remaining “reasoning” items were not included be-
cause they are not directly related to domestic violence.
Using the response format of the original instru-
ment, patients were asked to estimate how often within
the previous year their partner committed acts toward
them such as: sulked and/or refused to talk, stomped
out of the house or room; threatened to hit or throw
something; slapped; kicked, bit, or hit with a fist; and
threatened with a knife or gun. Respondents made
their estimates using a 7-point frequency scale of
never, once, twice, 3–5 times, 6–10 times, 11–20
times, and more than 20 times. Score values could
range from a minimum of 15 to a maximum of 105.
To ensure that the verbal and physical violence items
from the CTS continued to be a meaningful scale
without the reasoning items, we conducted an inter-
nal consistency analysis; the data was collected from
the patients in the study. Cronbach’s alpha was .87
for the 15 items.
The HITS Scale. The HITS scale is a paper-and-pen-
cil instrument that was comprised of the following
four items: “How often does your partner: physically
hurt you, insult you or talk down to you, threaten you
with harm, and scream or curse at you?” Patients re-
sponded to each of these items with a 5-point fre-
quency format: never, rarely, sometimes, fairly of-
ten, and frequently. Score values could range from a
minimum of 4 to a maximum of 20.
Participants
For phase one of the study (reliability and concur-
rent validity testing), 160 female patients visiting a
family practice clinic during April, May, and June
1996 participated in this study. For inclusion, partici-
pants had to be over age 21 and had to have lived
with the same partner for at least 12 months. All par-
ticipants were patients in the Family Practice Center
of Christ Hospital Medical Center (Advocate), which
serves a population of urban/suburban patients in the
southwest Chicago area of Oak Lawn.
For phase two of the study (measuring construct
validity), 99 women, who were self-identified as vic-
tims of domestic violence, participated. Some par-
ticipants were residents of domestic violence crisis
shelters (n=54), and others presented to an emergency
room (n=45).
Instrument Administration
In phase one, the CTS and HITS were each printed
on a separate page and stapled together. To control
for presentation effects, the sequencing of the scales
was counterbalanced so that half of the participants
completed the CTS followed by the HITS, and the
other half completed the instruments in the reverse
order.
Following approval from our institution’s Medical
Investigations Committee, female patients in the fam-
ily practice population were asked by nursing staff to
participate in the present study during a normal of-
fice visit. Volunteers completed forms privately in
exam rooms, and forms were collected before leav-
ing the office. To maintain confidentiality, no identi-
fying information was recorded.
In phase two, copies of the HITS were sent to each
of the crisis shelters and the agency that attended
women who presented in the emergency room. Staff
coordinators at each site were asked to solicit partici-
pation by clients and to distribute and collect the forms
at a time that did not disrupt counseling sessions or
other important therapeutic events.
Data Analysis
Phase One: Reliability and Concurrent Validity.
For this part of the study, frequency distributions were
constructed for total scores on the CTS and HITS.
Descriptive statistics for the HITS also were com-
puted. The internal consistency (ie, reliability) of the
HITS was determined with Cronbach’s coefficient
alpha. A scatter plot of the total scores for the CTS
and HITS was constructed, and the scores were then
correlated to establish the concurrent validity of the
HITS. In addition, subscores for items that assessed
verbal abuse and physical abuse were computed sepa-
rately and correlated across instruments.
Clinical Research and Methods

510
July-August 1998
Family Medicine
Tests for presentation effects were conducted by
comparing the total score for participants who com-
pleted an instrument first in the set with those who
completed it second. The Mann-Whitney U test was
used for this comparison because the distribution of
scores for both instruments was clearly not normal.
None of the participants had any missing data on
the HITS. However, 10 participants (6% of 160) did
not provide complete data on the CTS. Nine partici-
pants had one missing item, and one participant had
two missing items. To use all subjects in the analysis,
missing values were imputed with the mean value of
the existing CTS items.
Phase Two: Construct Validity. To measure the con-
struct validity of the HITS, scores for the respondents
at the shelters and emergency room were compared
with scores of the 160 female patients in phase one.
If the screening tool is efficacious, HITS scores of
self-identified victims of domestic violence should
be significantly higher than those of general patients
who are visiting their physician. HITS scores were
therefore compared using Student’s t test for inde-
pendent samples.
In addition, the Optimal Data Analysis® program
(ODA 1.0, Optimal Data Analysis for DOS, Chicago,
Optimal Data Analysis, Inc) was used to find a cut
score that reliably differentiated the two groups of
participants. The intent of this was to find a score
above which domestic violence probably has occurred.
In evaluating the cut score to distinguish victim-
ized and non-victimized respondents, a leave-one-out
validation method was used. To do this, one score
was held out while the other 258 (ie, 160 subjects in
phase one plus 99 subjects in phase two, minus one)
were used to find the cut score. This cut score was
then used to classify the holdout score in terms of it
belonging to the office or victimized group. After this
was done for all 259 scores, the results were merged
to examine the overall percentage accuracy classifi-
cation. The leave-one-out methodology allows one
to classify HITS scores that are not used to derive the
cut score, thus rendering an unbiased account of clas-
sification accuracy.
Results
Phase One: Reliability and Concurrent Validity
The frequency distributions for the CTS and HITS
scores from phase one are shown in Figure 1. As can
be seen, both distributions are L-shaped, indicating
that the majority of respondents scored in the low
(non-victim) range of the scale. The lowest and high-
est HITS scores were 4 and 18, respectively. The mean
was 6.13, the median was 5, and the standard devia-
tion was 2.75. Cronbach’s alpha was .80 for the four-
item scale. The analysis further showed that deleting
the item about being physically hurt would leave al-
pha unchanged. However, alpha would notably de-
crease if any of the other items were dropped from
the scale.
Figure 2 shows the scatterplot of HITS and CTS
scores. The lower left portion shows a higher density
of points due to the L-shaped nature of the score dis-
tributions. However, the relationship is positive and
linear. A correlation of .85 was found between HITS
and CTS total scores. Subscores on both instruments
that measured respondents’ experience of physical vio-
lence showed a correlation of r= .82. The same was true
for items that measured verbal violence, r=.81.
Figure 1
Frequency Distributions of CTS and HITS Scores
CTS—Conflict Tactics Scale
HITS—acronym for Hurts, Insults, Threatens, and Screams

511
Vol. 30, No. 7
Presentation Effects. Presentation effects were not
found. The median total HITS score was 5 for those
who completed the instrument first, as well as for
those who completed it second, z=.23, P=.815. The
median CTS scores were 19.64 and 20 for those com-
pleting the instrument first and second, respectively,
z=.26, P=.794.
Phase Two: Construct Validity
The mean HITS scores for the victimized and of-
fice groups were 15.12 and 6.13, respectively. This
difference was statistically significant, t=24.12, P<
.0005. Computations showed that 69% of the vari-
ance in HITS scores was attributable to group mem-
bership.
ODA
®
analysis revealed that the score of 10.5 re-
liably discriminated the two study groups
(P<.05).Table 1 shows the cut score classification per-
formance summary. In terms of actual group mem-
bership, 96% (95/99) of the victimized participants
and 91% (146/160) of the office participants were
classified correctly using this cut score. This is analo-
gous to sensitivity and specificity, respectively. In
terms of making predictions, 87% (95/109) of those
predicted to be victimized by domestic violence and
97% (146/150) of those predicted to be office patients
were accurate. This is analogous to the positive and
negative predictive values, respectively.
Discussion
The HITS scale is not the first short domestic vio-
lence screening tool to be developed for outpatient
clinical settings. Other short instruments, such as the
Abuse Assessment Screen,
15
have been developed for
the same purpose, but the HITS instrument is shorter
than others.
15-17
HITS has only four items, two each
that address verbal and physical aggression. The brev-
ity of the HITS is rivaled only by the three-item Part-
ner Violence Screen developed by Feldhaus et al.
18
However, the latter was designed for use in an emer-
gency room, and the items do not form an easily re-
membered acronym.
The results from phase one indicate that the HITS
has good internal consistency and concurrent valid-
ity with the CTS. Although the four reasoning items
were not used in the CTS, there is no reason to be-
lieve that this affected the ability of the scale to mea-
sure perception of physical and verbal violence. An
internal consistency analysis of the 15-item CTS with
our office sample revealed an alpha of .87.
The results from phase two of the study provide
two important findings to demonstrate the construct
validity of the HITS. First, the group of self-identi-
fied victims of abuse scored significantly higher than
family practice patients. Second, ODA
®
revealed that
the score of 10.5 reliably differentiated the two groups
of respondents. These findings are consistent with an
effective screening tool.
Figure 2
Scatter Plot of HITS and CTS Scores
Table 1
Cut Score Classification Performance Summary
Predicted Group
Membership
Victimized Office Total Accuracy
Victimized 95 4 99 96%
Office 14 146 160 91%
Total 109 150
Accuracy 87% 97%
Note: A HITS score of >10.5 classified someone as a victimized
respondent.
HITS—acronym for Hurts, Insults, Threatens, and Screams
Actual Group
Membership
CTS—Conflict Tactics Scale
HITS—acronym for Hurts, Insults, Threatens, and Screams
Clinical Research and Methods

512
July-August 1998
Family Medicine
In practice, the cut score of 10.5 is not directly us-
able because the HITS scoring procedure does not
allow for fractions of points. We, therefore, suggest
that clinicians suspect domestic violence when their
patients have a HITS score greater than 10.
Despite the statistical findings of the HITS cutoff,
physicians should investigate domestic violence
whenever they believe such a problem might exist.
The HITS is not used, nor should it be used, in lieu of
good clinical judgment. One must keep in mind that
our sample of office patients was compared with a
group of women for whom the experience of domes-
tic violence led them to seek professional help. We
suspect that there are many more women who cope
with a violent home life, yet, for whatever reason, do
not want to bring up the issue with a health care pro-
fessional. As always, clinical acumen should outweigh
test scores if there appears to be a discrepancy be-
tween the two.
Hopefully, a verbal form of the HITS with a yes-
no response format would have similar accuracy as
the written instrument used in this study. If so, physi-
cians could screen for domestic violence during a
conversation with a patient, thus obviating the need
for a paper-and-pencil instrument (albeit a short one).
Given the positive results from this study, additional
work should be done to explore the characteristics of
the HITS. For example, concurrent validity with other
normed instruments, such as the Index of Spouse
Abuse,
19
should be examined. HITS scores also could
be correlated with the incidence of violence that is
reported in medical records.
20
In addition, the utility
of the HITS should be studied with other women who
are known to be at high risk of violence.
Acknowledgments: This research was supported by a grant from the Med
Fund of Christ Hospital Medical Center (Advocate). The Med Fund has
sole rights for the reprint or use of the HITS questionnaire herein de-
scribed. The authors extend their appreciation to the following individu-
als and organizations that participated in this study: Sr. Cathleen Moore
and Pat Meneghini from Good Shepherd, Jenny Miller from Crisis Cen-
ter Advocates, Sharon Harold from South Suburban Family Shelter, and
Theresa DuBoise from Family Rescue. A portion of this work was pre-
sented at the 1997 Society of Teachers of Family Medicine (STFM) An-
nual Spring Conference in Boston and the 1998 STFM Annual Spring
Conference in Chicago. The work also was presented at the 1997 Ameri-
can Academy of Family Physicians Scientific Assembly in Chicago.
Corresponding Author: Address correspondence to Dr Sherin, Univer-
sity of Illinois at Chicago, Christ Family Medicine Residency Program,
4400 W 95 Street, Suite 207, Oak Lawn, IL 60453. 708-346-5324. Fax:
708-346-2003. E-mail: ksherin@uic.edu.
REFERENCES
1. American Medical Association. Diagnostic and treatment guidelines
on domestic violence. Chicago: American Medical Association, 1994.
2. Hamberger LK, Saunders DG, Hovey M. Prevalence of domestic
violence in community practice and rate of physician inquiry. Fam
Med 1992;24(4):283-7.
3. Tilden VP, Schmidt TA, Limandri BJ, Chiodo GT, Garland MJ, Love-
less PA. Factors that influence clinicians’ assessment and manage-
ment of family violence. Am J Public Health 1994;84(4):628-33.
4. Parsons LH, Zaccaro D, Wells B, Stovall TG. Methods of and atti-
tudes toward screening obstetrics and gynecology patients for do-
mestic violence. Am J Obstet Gynecol 1995;173(2):381-6.
5. Lewis BY. The wife abuse inventory: a screening device for the iden-
tification of abused women. Soc Work 1985;Jan-Feb:32-5.
6. Straus MA. Physical violence in American families. New Brunswick,
NJ: Transaction Publishers, 1990.
7. Brown JR, Sas C. Focus groups in family practice research: an ex-
ample study of family physicians’ approach to wife abuse. Fam Pract
Res J 1994;14(1):19-28.
8. Straus MA. Measuring intrafamily conflict and violence: the con-
flict tactics scales. Journal of Marriage and the Family 1979;4:75-
88.
9. Aldorondo E, Straus MA. Screening for physical violence in couple
therapy: methodological, practical, and ethical considerations. Fam
Process 1994;33(4):425-39.
10. Straus MA, Gelles R. Societal change and change in family violence
from 1975 to 1985 as revealed by national survey. Journal of Mar-
riage and the Family 1986;48:465-79.
11. O’Leary KD, Barling J, Arias I, Rosenbaum A, Malone J, Tyree A.
Prevalence and stability of physical aggression between spouses: a
longitudinal analysis. J Consult Clin Psych 1989;57:263-8.
12. Cronbach LJ. Essentials of psychological testing, third edition. New
York: Harper and Row, 1970.
13. Bullcroft RA, Straus MA. Validity of husband, wife, and child re-
ports of intrafamily violence and power. Durham, NH: Family Re-
search Lab, University of New Hampshire, 1975:16.
14. Kramer JJ, Conoley JC. The 11th mental measurement yearbook.
Lincoln, Neb: University of Nebraska Press, 1992.
15. Norton L, Peipert J, Zierler S, Lima B, Hume L. Battering in preg-
nancy: an assessment of two screening methods. Obstet Gynecol
1995;85(3):321-25.
16. Gendron C. The development of a questionnaire for identification of
women who are victims of violence in the conjugal milieu. Can J
Nurs Res 1991;23(3):21-32.
17. Attala JM. Risk identification of abused women participating in a
women, infants, and children program. Health Care Women Int
1994;15(6):587-97.
18. Feldhaus KM, Koziol-McLain JK, Amsbury HL, Norton IM,
Lowenstein SR, Abbott JT. Accuracy of three brief screening ques-
tions for detecting partner violence in the emergency department.
JAMA 1997;277(17):1357-61.
19. McFarlane J, Parker B, Soeken K, Bullock L. Assessing for abuse
during pregnancy: severity and frequency of injuries and associated
entry into prenatal care. JAMA 1992;267(23):3176-8.
20. Sanders DG, Hamberger LK, Hovey M. Indicators of woman abuse
based on a chart review at a family practice center. Arch Fam Med
1993;2:537-43.