Joumal of Family Psychology 2014. Vol.28. No. 3,421-428
© 2014 American Psychological Association 0893-3200/14/$I2.00 DOI: 10.1037/aOO36619
BRIEF REPORT
What Difference Does a Day Make? Examining Temporal Variations in Partner Maltreatment Randy J, McCarthy, Mandy M, Rabenhorst, and Joel S, Milner
Wendy J, Travis and Pamela S, Collins U,S, Air Force Family Advocacy Program, San Antonio, Texas
Northern Illinois University
Routine activities (RA) theory posits that changes in people's typical daily activities covary with increases or decreases in criminal behaviors, including, but not limited to, partner maltreatment. Using a large clinical database, we examined temporal variations among 24,460 incidents of confirmed partner maltreatment across an 11-year period within the U.S. Air Force (USAF). Specifically, we created regression models that predicted the number of partner maltreatment incidents per day. In addition to several control variables, we coded temporal variables for days of the week, month, year, and several significant days (e,g,, holidays. Super Bowl Sunday), which allowed us to examine the independent influence of these variables on partner maltreatment prevalence. While accounting for the influence of all other study variables, we observed significant increases in partner maltreatment for weekend days. New Year's Day, Independence Day, and Super Bowl Sunday, Similar results were found for partner maltreatment incidents involving offender alcohol/drug use. Furthermore, the proportion of incidents involving offender alcohol/drug use increased on New Year's Day and Independence Day, Consistent with RA theory and data from civilian samples, the current results indicate that certain days are associated with increased incidents of partner malueatment within the USAF, These fmdings should be used to inform future preventive efforts. Keywords: partner maltreatment, intimate partner violence, spouse abuse. Air Force, routine activities theory
Theories of partner maltreatment incorporate many factors, such as cultural, individual, and temporal factors. For example, at the cultural level, studies have examined country-to-country variations in partner maltreatment prevalence (e,g,, Straus, 2008) and cultural factors associated with partner maltreatment (e,g,. Archer, 2006), At the individual level, theories identify factors contributing to the
enactment of abusive behaviors (e.g., trait aggressiveness) and factors contributing to the inhibition of abusive behaviors (e,g,, dispositional self-control; Fitikel, 2007), Finally, and central to the current study, research examining temporal factors suggests that partner maltreatment is especially prevalent on days in which couples are likely to have increased discretionary time and social contact, such as weekends and holidays (Vazquez, Stohr, & Purkiss, 2005), Using a clinical database from the U,S, Air Force (USAF) containing 24,460 incidents of confirmed partner maltreatment across an 11-year period, we replicated and extended the previous findings on temporal variations in partner maltreatment. Research on temporal variations in partner maltreatment assumes that people have typical or routine daily activities. However, nonroutine days, such as weekends or holidays, may be accompanied by changes in discretionary time (e,g,, days off from work) and changes in what activities are likely (e,g,, festivities or social gatherings) or possible (e,g,, schools, government buildings, and some businesses may be closed). To the extent that changes in daily activities alter the social contact among couples, when compared with routine days, nonroutine days may be characterized by a relatively high or a relatively low prevalence of partner maltreatment (e,g,, Cohn & Rotton, 2003), This logic has been described by a routine activities (RA) theory, which proposes that partner maltreatment increases when circumstances bring couples, especially high-risk couples, into contact (e,g,, Cohen & Felson, 1979),
This article was published Online First May 5, 2014, Randy J, McCarthy, Mandy M. Rabenhorst, and Joel S. Milner, Center for the Study of Family Violence and Sexual Assault, Northern Illinois University; Wendy J. Travis and Pamela S, Collins, U,S. Air Force Family Advocacy Program, San Antonio, Texas. Mandy M. Rabenhorst is now at the Denver Veterans' Affairs Medical Center, The authors gratefully acknowledge Leasley K, Besetsny, USAF Family Advocacy Program, who provided support during this research. This study was funded by the USAF FAP through a subaward to M,M.R. and J,S,M, by Kansas State University and the U.S. Department of Agriculture, The views expressed are those of the authors and do not reflect the official policy or position of the U.S. Department of the Air Force, the U.S. Department of Defense, or the U.S. government. Correspondence concerning this article should be addressed to Randy J. McCarthy, Center for the Study of Family Violence and Sexual Assault, 125 Presidents Boulevard, Northern Illinois University, DeKalb, IL 60115, E-mail: [email protected]
421
422
MCCARTHY, RABENHORST, MILNER, TRAVIS, AND COLLINS
Eor many adults, weekends and holidays involve increased time spent with a partner. Thus, RA theory predicts that these days should be associated with increased partner maltreatment. In support of RA theory, Cohn and Rotton (2003) found increased calls to police dispatchers for incidents of domestic violence across three separate years (i.e., 1985, 1987, 1988) in Minneapolis, Minnesota, on major holidays (e.g.. New Year's Day, Thanksgiving, Christmas Eve). However, these researchers only analyzed domestic violence reports and did not provide information on whether or not these incidents were subsequently substantiated. Likewise, using the Incident-Based Reporting System for the state of Idaho across a 7-year period (1995-2001), Vazquez et al. (2005) found increases in crimes against intimate partners (e.g., spouse, boyfriend/girlfriend) on the first day of each month, weekends, holidays (e.g.. New Year's Day, Independence Day), and culturally significant days (e.g.. Super Bowl Sunday) and decreases on Valentine's Day and Christmas. Although the database that Vazquez and colleagues used included reports of crimes against intimate partners and whether the offender was arrested, the researchers did not examine whether the likelihood of offender arrests covaried with the temporal variables they studied. Eurther, the studies of Cohn and Rotton as well as that of Vazquez and colleagues implicated alcohol as a contributing factor because some holidays (e.g.. New Year's Day, Independence Day) and culturally significant days (e.g.. Super Bowl Sunday) commonly include activities involving alcohol consumption. However, their analyses did not include any data on alcohol (or any other substance use). Given the limitations of the previous research, the current study makes two major contributions. Eirst, the data used in the current study contained information on alleged incidents of partner maltreatment and whether each incident met criteria for determining whether maltreatment occurred. Although in all of our analyses we predicted incidents that met criteria for partner maltreatment, we considered the daily proportion of alleged incidents that met criteria for maltreatment (i.e., met criteria incidents/total alleged incidents). This allowed us to predict the number of met criteria incidents on a given day independent of factors that may cause changes in the likelihood of reporting incidents or the likelihood of incident substantiation. Second, whereas previous research only speculated that alcohol was a contributing factor to temporal variations in partner maltreatment, we were in the unique position to have incident-level information on whether each offender used alcohol/dmgs at the time of each incident. Thus, we examined temporal variation in the number of incidents with offender alcohol/dmg use and the proportion of incidents with offender alcohol/ drug use. Einally, our current study tests the explanatory power and generalizability of RA theory. Whereas previous research (e.g., Cohn & Rotton, 2003; Vazquez et al., 2005) examined temporal variations in partner maltreatment within nonmilitary samples, the current study involves data from the USAE. Although USAE members also may have different routines on weekends and observe similar holidays (e.g.. Independence Day) and other significant days (e.g.. Super Bowl Sunday) as the general U.S. population, it is possible that factors specific to military culture or to military members would produce observations that RA theory could not explain. Eor example, military culture has strong injunctive norms on how persons ought to behave and which behaviors
are approved or disapproved in situations. Eurthermore, within the military, behaviors such as partner maltreatment are directly tied to military members' careers. Thus, one may speculate that the predictions of RA theory may be attenuated within a population in which such strong situational factors are present. If RA theory only explained temporal changes for specific operationalizations of partner maltreatment (e.g., reports of maltreatment) and for a specific population (e.g., civilians), then the usefulness of the theory would be limited. However, although previous research suggests the existence of predictable temporal changes, observing the same pattern of effects within a different population with a different operationalization of partner maltreatment would greatly increase the confidence that researchers could have in those effects. The current study investigated temporal variations in partner maltreatment within the USAE using a database containing more incidents and covering a larger period of time than previous research, and the database included information on allegations, on whether incidents met criteria for maltreatment, and on offender alcohol/dmg use. As in previous research (e.g., Vazquez et al., 2005), we expected increases in partner maltreatment on certain days (e.g., weekends. New Year's Day, Independence Day). In addition, we tested whether partner maltreatment prevalence was associated with offender alcohol/drug use. We predicted increases in the number of incidents involving offender alcohol/dmg use and the proportion of incidents involving offender alcohol/drug use on dates when alcohol consumption may be most frequent (e.g.. New Year's Day, Independence Day, Super Bowl Sunday).
Method Before being provided to the researchers, the database used in the study presented here was stripped of identifying information, and formal approvals were obtained from the institutional review boards of the USAE and the first author's university.
Description of Datahase Information about partner maltreatment came from the USAE Eamily Advocacy System of Records. This database contained information about each allegation of partner maltreatment referred to the Eamily Advocacy Program (EAP) at USAE installations worldwide. USAE policy requires referral of all suspected partner maltreatment to the EAP (United States Air Eorce, 2009), and referrals come from various military and civil sources, both of which are mandated reporters (Linkh et al., 2008). At the time of the report, information is gathered about the alleged incident, and it is noted whether there is any evidence that the offender or victim used alcohol/dmgs at the time of the alleged incident. Each referral is then subjected to a review process during which a determination is made about whether the incident met criteria for partner maltreatment. This review process is a formal administrative, as opposed to legal, determination of whether or not an alleged incident met criteria for maltreatment and occurred within a stmctured decision-making tree format (e.g., Heyman & Slep, 2009). Within this stmctured format, members of a review board evaluated the circumstances of the allegation to determine if the incident met each criterion. As each criterion was reviewed, the board continued to the next until a "preponderance of evidence" indicated that the alleged incident of partner maltreatment "met criteria."
TEMPORAL VARIATIONS IN PARTNER MALTREATMENT Each allegation could have met criteria for physical abuse, emotional abuse, neglect, sexual abuse, or more than one type of maltreatment. Physical abuse was defined as the nonaccidental use of physical force against a partner that results in any physical injury, the reasonable potential for physical injury, or more than an inconsequential fear reaction. Emotional abuse was defined as the nonaccidental act(s) (e.g., "berating or disparaging") or threat(s) (e.g., "implying future physical harm or sexual assault") adversely affecting the psychological well-being of the victim that results in psychological harm or stress-related somatic symptoms. Neglect was defined as withholding necessary care or assistance for a partner who is physically, psychologically, or culturally incapable of self-care, although the caregiver is financially able to do so or has been offered other means to do so. Finally, sexual abuse was defined as a sexual act or physical contact in a sexual nature without the consent of the partner (see Heyman & Slep, 2006, 2009). Furthermore, when each alleged incident was reported, FAP caseworkers gathered information about offender alcohol and drug use from adults, children, other witnesses, and any police reports. When accounting for all of the available information, FAP determined whether there was sufficient evidence that the offender used alcohol and/or drugs at the time the alleged incident occurred. Offender alcohol/drug use was recorded into the Family Advocacy System of Records database as an incident-level variable. The database analyzed in the current study contained 42,997 alleged incidents of partner maltreatment involving an active-duty USAF member (as an offender or victim) from January 1, 2000 to July 11, 2011. Of these incidents, 24,460 met criteria for partner maltreatment: physical abuse (n = 21,859) and emotional abuse (n = 9,240), with small numbers of sexual abuse (n = 364) and neglect (n = 52). These sum to more than the total number of incidents because each incident can meet criteria for more than one type of maltreatment. Finally, 5,791 met criteria for partner maltreatment and involved an offender who had used alcohol/drugs at the time of the incident.
Analytic Strategy In the first two analyses, we used Poisson regression because the outcomes (i.e., number of incidents) were count data and the distributions were positively skewed. In the flrst analysis, we determined the number of incidents that met criteria for partner maltreatment on a single date. In the second analysis, we determined the number of incidents that met criteria for partner maltreatment and involved an offender who had used alcohol/drugs at the time of the incident. In a final analysis, we determined the daily proportion of partner maltreatment incidents that met criteria and involved offender alcohol/drug use. Because the proportion was a continuous variable that ranged from 0.0 (indicating zero incidents involved offender alcohol/drug use on a particular date) to 1.0 (indicating that all incidents on a particular date involved offender alcohol/drug use), we used ordinary least-squares regression. Various factors were examined as predictors of partner maltreatment prevalence. Dummy variable (0, 1) coding was used to control for temporal variables of interest. The models included six variables to control for the day of the week (with Monday being the reference day); 11 variables to control for month of the year (with January being the reference month); and variables for the
423
first day of each month. Thanksgiving, Christmas Eve, Christmas Day, New Year's Eve, New Year's Day, Super Bowl Sunday, Valentine's Day, St. Patrick's Day, Memorial Day, and Independence Day. To control for linear trend, we included a sequence variable starting at 1 and ending at 4,187 (cf., Cohn & Rotton, 2003). Finally, to control for the tendency to report incidents and for the likelihood of an alleged incident meeting criteria, the proportion of incidents that met criteria on each day was included as a control variable. For ease of comparison across different analyses, all descriptive statistics and figures report the unadjusted mean number of incidents that met criteria per day. When discussing results from the Poisson regression analyses, we report the percentage of change in the predicted number of incidents per day while controlling for the influence of all other variables. In other words, the percentage of change represents the influence of a variable (e.g., a dummy code for a particular holiday) on the predicted number of incidents for a particular day that is independent of the influence of all of the other variables. For the ordinary least-squares multiple regression analysis, the coefficients represent the predicted change in the proportion of incidents on a given day (for a one-unit increase in the predictor variable) that involved offender alcohol/drug use while controlling for the influence of other variables in the model.
Results Partner Maltreatment Incidents Year-to-year trends. Partner maltreatment incidents that met criteria ranged from a high of 7.72 (SD = 3.62) incidents per day in 2001 to a low of 3.89 (SD = 2.78) incidents per day in 2011. When controlling for other variables, the effect of year (B = -0.15) ranged from B = -0.71 to ß = 0.40 with 95% confidence (see Table 1). Thus, the effect of year did not independently predict the daily number of partner maltreatment incidents that met criteria. Month-to-month trends. Partner maltreatment incidents that met criteria ranged from a high of 6.14 (SD = 3.16) incidents per day in August to a low of 5.70 (SD = 3.20) incidents per day in December. When controlling for other variables, the coefficients for individual months ranged from B = -0.16, 95% confidence interval (CI) [-0.67, 0.35] for December to ß = -0.04, 95% CI [-0.12, 0.04] for February and B = -0.04, 95% CI [-0.15, 0.07] for March (see Table 1). The effect of month did not independently predict the daily number of partner maltreatment incidents that met criteria. Day-to-day trends. Day-to-day trends were explored in two separate ways; by day of the week and by day of the year. The mean number of met criteria partner maltreatment incidents was lowest on Wednesday (M = 5.09, SD = 2.98) and highest on Sunday (M = 7.56, SD = 3.57). While controlling for the influence of all other variables (e.g., year, month, whether a date was a holiday, etc.), the predicted number of met criteria incidents of partner maltreatment were lower on Tuesday (-8%), Wednesday (-9%), Thursday (-10%), and Friday (-6%) and greater on Saturday (-1-22%) and Sunday (-1-30%) relative to Monday (see Table 1).
MCCARTHY, RABENHORST, MILNER, TRAVIS, AND COLLINS
424
Table 1 Predicted Number of Partner Maltreatment Incidents Per Day That Met Criteria 9 5 % Cl
B
SE
RR
LL
UL
309.42 0.00 -0.15 1.56"' -0.04 -0.04 -0.05 -0.07 -0.07 -0.08 -0.04 -0.08 -0.11 -0.10 -0.16 -0.09'" -O.IO"* -0.10"* -0.06* 0.20*** 0.26*** 0.09 -0.10 -0.12 0.29' 0.48*** 0.30** 0.07 0.12 0.27* 0.35** 0.17***
569.26 0.00 0.28 0.04 0.04 0.06 0.08 0.10 0.12 0.15 0.17 0.19 0.22 0.24 0.26 0.03 0.03 0.03 0.03 0.02 0.02 0.13 0.14 0.13 0.11 0.09 0.09 0.12 0.11 0.11 0.11 0.04
2.39E-I-134
0.00 1.00 0.49 4.44 0.89 0.86 0.82 0.77 0.74 0.70 0.69 0.64 0.59 0.57 0.51 0.87 0.87 0.86 0.89 1.17 1.24 0.85 0.70 0.69 1.07 1.35 1.13 0.85 0.90 1.05 1.15 1.10
a
Predictor Intercept Sequence Year Met criteria rate'' February March April
May June July August September October November December Tuesday Wednesday Thursday Friday Saturday Sunday Thanksgiving Christmas Eve Christmas Day New Year's Eve New Year's Day Super Bowl Sunday Valentine's Day St. Patrick's Day Memorial Day Independence Day First Day of Month
1.00 0.86 4.76 0.96 0.96 0.95 0.93 0.93 0.92 0.96 0.93 0.89 0.90 0.85 0.91 0.91 0.90 0.94 1.22 1.30 1.10 0.91 0.89 1.33 1.62 1.36 1.07 1.13 1.30 1.42 1.18
.00 .50 Íi . l l
.04 .07 .10 .13 .18 .23 .33 .35 .36 .44 1.42 ().96 (5.96 (D.95 3.99
.28 1.36 1.42 1.19 1.15 1.66 1.94 1.62 1.37 1.41
.62 1.75 1.27
Note. RR = rate ratio, LL = lower limit, UL = upper limit. " Set to system missing because of overflow. '' Proportion of incidents meeting criteria on a given day. * / 7 < . O 5 . "p
© 2014 American Psychological Association 0893-3200/14/$I2.00 DOI: 10.1037/aOO36619
BRIEF REPORT
What Difference Does a Day Make? Examining Temporal Variations in Partner Maltreatment Randy J, McCarthy, Mandy M, Rabenhorst, and Joel S, Milner
Wendy J, Travis and Pamela S, Collins U,S, Air Force Family Advocacy Program, San Antonio, Texas
Northern Illinois University
Routine activities (RA) theory posits that changes in people's typical daily activities covary with increases or decreases in criminal behaviors, including, but not limited to, partner maltreatment. Using a large clinical database, we examined temporal variations among 24,460 incidents of confirmed partner maltreatment across an 11-year period within the U.S. Air Force (USAF). Specifically, we created regression models that predicted the number of partner maltreatment incidents per day. In addition to several control variables, we coded temporal variables for days of the week, month, year, and several significant days (e,g,, holidays. Super Bowl Sunday), which allowed us to examine the independent influence of these variables on partner maltreatment prevalence. While accounting for the influence of all other study variables, we observed significant increases in partner maltreatment for weekend days. New Year's Day, Independence Day, and Super Bowl Sunday, Similar results were found for partner maltreatment incidents involving offender alcohol/drug use. Furthermore, the proportion of incidents involving offender alcohol/drug use increased on New Year's Day and Independence Day, Consistent with RA theory and data from civilian samples, the current results indicate that certain days are associated with increased incidents of partner malueatment within the USAF, These fmdings should be used to inform future preventive efforts. Keywords: partner maltreatment, intimate partner violence, spouse abuse. Air Force, routine activities theory
Theories of partner maltreatment incorporate many factors, such as cultural, individual, and temporal factors. For example, at the cultural level, studies have examined country-to-country variations in partner maltreatment prevalence (e,g,, Straus, 2008) and cultural factors associated with partner maltreatment (e,g,. Archer, 2006), At the individual level, theories identify factors contributing to the
enactment of abusive behaviors (e.g., trait aggressiveness) and factors contributing to the inhibition of abusive behaviors (e,g,, dispositional self-control; Fitikel, 2007), Finally, and central to the current study, research examining temporal factors suggests that partner maltreatment is especially prevalent on days in which couples are likely to have increased discretionary time and social contact, such as weekends and holidays (Vazquez, Stohr, & Purkiss, 2005), Using a clinical database from the U,S, Air Force (USAF) containing 24,460 incidents of confirmed partner maltreatment across an 11-year period, we replicated and extended the previous findings on temporal variations in partner maltreatment. Research on temporal variations in partner maltreatment assumes that people have typical or routine daily activities. However, nonroutine days, such as weekends or holidays, may be accompanied by changes in discretionary time (e,g,, days off from work) and changes in what activities are likely (e,g,, festivities or social gatherings) or possible (e,g,, schools, government buildings, and some businesses may be closed). To the extent that changes in daily activities alter the social contact among couples, when compared with routine days, nonroutine days may be characterized by a relatively high or a relatively low prevalence of partner maltreatment (e,g,, Cohn & Rotton, 2003), This logic has been described by a routine activities (RA) theory, which proposes that partner maltreatment increases when circumstances bring couples, especially high-risk couples, into contact (e,g,, Cohen & Felson, 1979),
This article was published Online First May 5, 2014, Randy J, McCarthy, Mandy M. Rabenhorst, and Joel S. Milner, Center for the Study of Family Violence and Sexual Assault, Northern Illinois University; Wendy J. Travis and Pamela S, Collins, U,S. Air Force Family Advocacy Program, San Antonio, Texas. Mandy M. Rabenhorst is now at the Denver Veterans' Affairs Medical Center, The authors gratefully acknowledge Leasley K, Besetsny, USAF Family Advocacy Program, who provided support during this research. This study was funded by the USAF FAP through a subaward to M,M.R. and J,S,M, by Kansas State University and the U.S. Department of Agriculture, The views expressed are those of the authors and do not reflect the official policy or position of the U.S. Department of the Air Force, the U.S. Department of Defense, or the U.S. government. Correspondence concerning this article should be addressed to Randy J. McCarthy, Center for the Study of Family Violence and Sexual Assault, 125 Presidents Boulevard, Northern Illinois University, DeKalb, IL 60115, E-mail: [email protected]
421
422
MCCARTHY, RABENHORST, MILNER, TRAVIS, AND COLLINS
Eor many adults, weekends and holidays involve increased time spent with a partner. Thus, RA theory predicts that these days should be associated with increased partner maltreatment. In support of RA theory, Cohn and Rotton (2003) found increased calls to police dispatchers for incidents of domestic violence across three separate years (i.e., 1985, 1987, 1988) in Minneapolis, Minnesota, on major holidays (e.g.. New Year's Day, Thanksgiving, Christmas Eve). However, these researchers only analyzed domestic violence reports and did not provide information on whether or not these incidents were subsequently substantiated. Likewise, using the Incident-Based Reporting System for the state of Idaho across a 7-year period (1995-2001), Vazquez et al. (2005) found increases in crimes against intimate partners (e.g., spouse, boyfriend/girlfriend) on the first day of each month, weekends, holidays (e.g.. New Year's Day, Independence Day), and culturally significant days (e.g.. Super Bowl Sunday) and decreases on Valentine's Day and Christmas. Although the database that Vazquez and colleagues used included reports of crimes against intimate partners and whether the offender was arrested, the researchers did not examine whether the likelihood of offender arrests covaried with the temporal variables they studied. Eurther, the studies of Cohn and Rotton as well as that of Vazquez and colleagues implicated alcohol as a contributing factor because some holidays (e.g.. New Year's Day, Independence Day) and culturally significant days (e.g.. Super Bowl Sunday) commonly include activities involving alcohol consumption. However, their analyses did not include any data on alcohol (or any other substance use). Given the limitations of the previous research, the current study makes two major contributions. Eirst, the data used in the current study contained information on alleged incidents of partner maltreatment and whether each incident met criteria for determining whether maltreatment occurred. Although in all of our analyses we predicted incidents that met criteria for partner maltreatment, we considered the daily proportion of alleged incidents that met criteria for maltreatment (i.e., met criteria incidents/total alleged incidents). This allowed us to predict the number of met criteria incidents on a given day independent of factors that may cause changes in the likelihood of reporting incidents or the likelihood of incident substantiation. Second, whereas previous research only speculated that alcohol was a contributing factor to temporal variations in partner maltreatment, we were in the unique position to have incident-level information on whether each offender used alcohol/dmgs at the time of each incident. Thus, we examined temporal variation in the number of incidents with offender alcohol/dmg use and the proportion of incidents with offender alcohol/ drug use. Einally, our current study tests the explanatory power and generalizability of RA theory. Whereas previous research (e.g., Cohn & Rotton, 2003; Vazquez et al., 2005) examined temporal variations in partner maltreatment within nonmilitary samples, the current study involves data from the USAE. Although USAE members also may have different routines on weekends and observe similar holidays (e.g.. Independence Day) and other significant days (e.g.. Super Bowl Sunday) as the general U.S. population, it is possible that factors specific to military culture or to military members would produce observations that RA theory could not explain. Eor example, military culture has strong injunctive norms on how persons ought to behave and which behaviors
are approved or disapproved in situations. Eurthermore, within the military, behaviors such as partner maltreatment are directly tied to military members' careers. Thus, one may speculate that the predictions of RA theory may be attenuated within a population in which such strong situational factors are present. If RA theory only explained temporal changes for specific operationalizations of partner maltreatment (e.g., reports of maltreatment) and for a specific population (e.g., civilians), then the usefulness of the theory would be limited. However, although previous research suggests the existence of predictable temporal changes, observing the same pattern of effects within a different population with a different operationalization of partner maltreatment would greatly increase the confidence that researchers could have in those effects. The current study investigated temporal variations in partner maltreatment within the USAE using a database containing more incidents and covering a larger period of time than previous research, and the database included information on allegations, on whether incidents met criteria for maltreatment, and on offender alcohol/dmg use. As in previous research (e.g., Vazquez et al., 2005), we expected increases in partner maltreatment on certain days (e.g., weekends. New Year's Day, Independence Day). In addition, we tested whether partner maltreatment prevalence was associated with offender alcohol/drug use. We predicted increases in the number of incidents involving offender alcohol/dmg use and the proportion of incidents involving offender alcohol/drug use on dates when alcohol consumption may be most frequent (e.g.. New Year's Day, Independence Day, Super Bowl Sunday).
Method Before being provided to the researchers, the database used in the study presented here was stripped of identifying information, and formal approvals were obtained from the institutional review boards of the USAE and the first author's university.
Description of Datahase Information about partner maltreatment came from the USAE Eamily Advocacy System of Records. This database contained information about each allegation of partner maltreatment referred to the Eamily Advocacy Program (EAP) at USAE installations worldwide. USAE policy requires referral of all suspected partner maltreatment to the EAP (United States Air Eorce, 2009), and referrals come from various military and civil sources, both of which are mandated reporters (Linkh et al., 2008). At the time of the report, information is gathered about the alleged incident, and it is noted whether there is any evidence that the offender or victim used alcohol/dmgs at the time of the alleged incident. Each referral is then subjected to a review process during which a determination is made about whether the incident met criteria for partner maltreatment. This review process is a formal administrative, as opposed to legal, determination of whether or not an alleged incident met criteria for maltreatment and occurred within a stmctured decision-making tree format (e.g., Heyman & Slep, 2009). Within this stmctured format, members of a review board evaluated the circumstances of the allegation to determine if the incident met each criterion. As each criterion was reviewed, the board continued to the next until a "preponderance of evidence" indicated that the alleged incident of partner maltreatment "met criteria."
TEMPORAL VARIATIONS IN PARTNER MALTREATMENT Each allegation could have met criteria for physical abuse, emotional abuse, neglect, sexual abuse, or more than one type of maltreatment. Physical abuse was defined as the nonaccidental use of physical force against a partner that results in any physical injury, the reasonable potential for physical injury, or more than an inconsequential fear reaction. Emotional abuse was defined as the nonaccidental act(s) (e.g., "berating or disparaging") or threat(s) (e.g., "implying future physical harm or sexual assault") adversely affecting the psychological well-being of the victim that results in psychological harm or stress-related somatic symptoms. Neglect was defined as withholding necessary care or assistance for a partner who is physically, psychologically, or culturally incapable of self-care, although the caregiver is financially able to do so or has been offered other means to do so. Finally, sexual abuse was defined as a sexual act or physical contact in a sexual nature without the consent of the partner (see Heyman & Slep, 2006, 2009). Furthermore, when each alleged incident was reported, FAP caseworkers gathered information about offender alcohol and drug use from adults, children, other witnesses, and any police reports. When accounting for all of the available information, FAP determined whether there was sufficient evidence that the offender used alcohol and/or drugs at the time the alleged incident occurred. Offender alcohol/drug use was recorded into the Family Advocacy System of Records database as an incident-level variable. The database analyzed in the current study contained 42,997 alleged incidents of partner maltreatment involving an active-duty USAF member (as an offender or victim) from January 1, 2000 to July 11, 2011. Of these incidents, 24,460 met criteria for partner maltreatment: physical abuse (n = 21,859) and emotional abuse (n = 9,240), with small numbers of sexual abuse (n = 364) and neglect (n = 52). These sum to more than the total number of incidents because each incident can meet criteria for more than one type of maltreatment. Finally, 5,791 met criteria for partner maltreatment and involved an offender who had used alcohol/drugs at the time of the incident.
Analytic Strategy In the first two analyses, we used Poisson regression because the outcomes (i.e., number of incidents) were count data and the distributions were positively skewed. In the flrst analysis, we determined the number of incidents that met criteria for partner maltreatment on a single date. In the second analysis, we determined the number of incidents that met criteria for partner maltreatment and involved an offender who had used alcohol/drugs at the time of the incident. In a final analysis, we determined the daily proportion of partner maltreatment incidents that met criteria and involved offender alcohol/drug use. Because the proportion was a continuous variable that ranged from 0.0 (indicating zero incidents involved offender alcohol/drug use on a particular date) to 1.0 (indicating that all incidents on a particular date involved offender alcohol/drug use), we used ordinary least-squares regression. Various factors were examined as predictors of partner maltreatment prevalence. Dummy variable (0, 1) coding was used to control for temporal variables of interest. The models included six variables to control for the day of the week (with Monday being the reference day); 11 variables to control for month of the year (with January being the reference month); and variables for the
423
first day of each month. Thanksgiving, Christmas Eve, Christmas Day, New Year's Eve, New Year's Day, Super Bowl Sunday, Valentine's Day, St. Patrick's Day, Memorial Day, and Independence Day. To control for linear trend, we included a sequence variable starting at 1 and ending at 4,187 (cf., Cohn & Rotton, 2003). Finally, to control for the tendency to report incidents and for the likelihood of an alleged incident meeting criteria, the proportion of incidents that met criteria on each day was included as a control variable. For ease of comparison across different analyses, all descriptive statistics and figures report the unadjusted mean number of incidents that met criteria per day. When discussing results from the Poisson regression analyses, we report the percentage of change in the predicted number of incidents per day while controlling for the influence of all other variables. In other words, the percentage of change represents the influence of a variable (e.g., a dummy code for a particular holiday) on the predicted number of incidents for a particular day that is independent of the influence of all of the other variables. For the ordinary least-squares multiple regression analysis, the coefficients represent the predicted change in the proportion of incidents on a given day (for a one-unit increase in the predictor variable) that involved offender alcohol/drug use while controlling for the influence of other variables in the model.
Results Partner Maltreatment Incidents Year-to-year trends. Partner maltreatment incidents that met criteria ranged from a high of 7.72 (SD = 3.62) incidents per day in 2001 to a low of 3.89 (SD = 2.78) incidents per day in 2011. When controlling for other variables, the effect of year (B = -0.15) ranged from B = -0.71 to ß = 0.40 with 95% confidence (see Table 1). Thus, the effect of year did not independently predict the daily number of partner maltreatment incidents that met criteria. Month-to-month trends. Partner maltreatment incidents that met criteria ranged from a high of 6.14 (SD = 3.16) incidents per day in August to a low of 5.70 (SD = 3.20) incidents per day in December. When controlling for other variables, the coefficients for individual months ranged from B = -0.16, 95% confidence interval (CI) [-0.67, 0.35] for December to ß = -0.04, 95% CI [-0.12, 0.04] for February and B = -0.04, 95% CI [-0.15, 0.07] for March (see Table 1). The effect of month did not independently predict the daily number of partner maltreatment incidents that met criteria. Day-to-day trends. Day-to-day trends were explored in two separate ways; by day of the week and by day of the year. The mean number of met criteria partner maltreatment incidents was lowest on Wednesday (M = 5.09, SD = 2.98) and highest on Sunday (M = 7.56, SD = 3.57). While controlling for the influence of all other variables (e.g., year, month, whether a date was a holiday, etc.), the predicted number of met criteria incidents of partner maltreatment were lower on Tuesday (-8%), Wednesday (-9%), Thursday (-10%), and Friday (-6%) and greater on Saturday (-1-22%) and Sunday (-1-30%) relative to Monday (see Table 1).
MCCARTHY, RABENHORST, MILNER, TRAVIS, AND COLLINS
424
Table 1 Predicted Number of Partner Maltreatment Incidents Per Day That Met Criteria 9 5 % Cl
B
SE
RR
LL
UL
309.42 0.00 -0.15 1.56"' -0.04 -0.04 -0.05 -0.07 -0.07 -0.08 -0.04 -0.08 -0.11 -0.10 -0.16 -0.09'" -O.IO"* -0.10"* -0.06* 0.20*** 0.26*** 0.09 -0.10 -0.12 0.29' 0.48*** 0.30** 0.07 0.12 0.27* 0.35** 0.17***
569.26 0.00 0.28 0.04 0.04 0.06 0.08 0.10 0.12 0.15 0.17 0.19 0.22 0.24 0.26 0.03 0.03 0.03 0.03 0.02 0.02 0.13 0.14 0.13 0.11 0.09 0.09 0.12 0.11 0.11 0.11 0.04
2.39E-I-134
0.00 1.00 0.49 4.44 0.89 0.86 0.82 0.77 0.74 0.70 0.69 0.64 0.59 0.57 0.51 0.87 0.87 0.86 0.89 1.17 1.24 0.85 0.70 0.69 1.07 1.35 1.13 0.85 0.90 1.05 1.15 1.10
a
Predictor Intercept Sequence Year Met criteria rate'' February March April
May June July August September October November December Tuesday Wednesday Thursday Friday Saturday Sunday Thanksgiving Christmas Eve Christmas Day New Year's Eve New Year's Day Super Bowl Sunday Valentine's Day St. Patrick's Day Memorial Day Independence Day First Day of Month
1.00 0.86 4.76 0.96 0.96 0.95 0.93 0.93 0.92 0.96 0.93 0.89 0.90 0.85 0.91 0.91 0.90 0.94 1.22 1.30 1.10 0.91 0.89 1.33 1.62 1.36 1.07 1.13 1.30 1.42 1.18
.00 .50 Íi . l l
.04 .07 .10 .13 .18 .23 .33 .35 .36 .44 1.42 ().96 (5.96 (D.95 3.99
.28 1.36 1.42 1.19 1.15 1.66 1.94 1.62 1.37 1.41
.62 1.75 1.27
Note. RR = rate ratio, LL = lower limit, UL = upper limit. " Set to system missing because of overflow. '' Proportion of incidents meeting criteria on a given day. * / 7 < . O 5 . "p
