Essay Grade

Part of the job? Workplace violence in Massachusetts social service agencies
Jennifer R. Zelnick, Elspeth Slayter, Beth Flanzbaum, Nanci Ginty Butler, Beryl Domingo and Judith Perlstein
Health and Social Work. 38.2 (May 2013): p75+.
DOI: http://dx.doi.org.ezp.waldenulibrary.org/10.1093/hsw/hlt007
Copyright: COPYRIGHT 2013 Oxford University Press
http://www.naswpress.org/publications/journals/hsw.html
Abstract:
Workplace violence is a serious and surprisingly understudied occupational hazard in social service settings. The
authors of this study conducted an anonymous, Internet-based survey of Massachusetts social service agencies to
estimate the incidence of physical assault and verbal threat of violence in social service agencies, understand how
social service agencies collect data on workplace violence, and identify disparities in who is at risk in terms of staff
education and training level and the work setting. The study gathered general descriptions of each agency and
compiled incidence data on workplace violence that were collected by agencies in fiscal year 2009. The key
findings of this descriptive study showed high rates of workplace violence against social services providers and a
pattern of risk disparity, with significantly more risk for direct care versus clinical staff. These results are based on
data routinely collected by social service agencies that typically remain unexamined. A research agenda that is
sensitive to potential occupational health disparities and focuses on maximizing workplace safety in social services
is needed.
KEY WORDS: occupational health; occupational health disparities; risk; workforce; workplace violence
Full Text:
Workplace violence is a serious and surprisingly understudied occupational hazard in social service settings
(Jayaratne, Croxton, & Mattison, 2004). In relation to other employment sectors in the United States, health care
and social assistance (HCSA) is among the most dangerous. For example, according to the most recent Bureau of
Labor Statistics (BLS) report on workplace violence, the median number of days lost as a result of assault for
private sector workers employed in HCSA was twice that of the private sector as a whole (9.7 compared with 4.9
per 10,000), and for state HCSA employees the median was more than four times that of all state employees (136
compared with 30 per 10,000) (BLS, 2010). Studies conducted since the 1990s, many in response to deteriorating
working conditions and service quality resulting from underfunding and program cuts, have consistently found high
rates of physical assault (3 percent to 30 percent) and verbal threat of assault (42 percent to 82 percent) (Ringstad,
2009). In Massachusetts, where our study was conducted, three members of the social service workforce have lost
their lives as a result of workplace violence since 2008.
Yet the topic of workplace violence in social services has received scant attention within social work and the public
health and occupational health literature. For example, a search for the temps “social work” and “workplace
violence” on EBSCO Academic Search Complete (which includes most social work- and occupational healthfocused journals) between 2001 and 2011 turned up only nine empirical studies; considering the range of
populations served by various social services and practitioners, this means that many settings where workplace
violence is a risk have not been studied at all.
Therefore, there is a need to better understand workplace violence in social services to create policies and
initiatives that make social services safer for the workforce and the people they serve. Building on previous
literature, this descriptive study has three aims: (1) to estimate the fiscal year (FY) 2009 reported incidence of
physical assault and verbal threat of assault in participating Massachusetts social service agencies; (2) to
understand how different social service agencies collect data related to workplace violence; and (3) to identify
disparities in risk for workplace violence in staff education and training level and the settings of workplace violence.
BACKGROUND
What Counts as Workplace Violence?
Studies of workplace violence among social workers typically include physical assault, verbal threat of assault,
verbal abuse, and property damage (Newhill, 1996; Rey, 1996; Shields & Kiser, 2003; Spencer & Munch, 2003;
Winstanley & Hales, 2008). Some studies include sexual and racial or ethnic harassment in their definition of
workplace violence (Guterman, Jayaratne, & Bargal, 1996; Hoobler & Swanberg, 2006; Jayaratne, Croxton, &
Mattison, 2004; Koritsas, Coles, & Boyle, 2010; McDonald & Sirotich, 2005; Pollack, 2010; Ringstad, 2005). Few
studies have used conceptually framed, validated tools to measure workplace violence (Ringstad, 2005;
Winstanley & Hales, 2008); most rely on the individual worker’s subjective definition of “workplace violence” to
account for the variation in perceptions of what is “violent” (Guterman et al., 1996; Jayaratne et al., 2004; Koritsas
et al., 2010; McDonald & Sirotich, 2005; Newhill, 1996; Pollack, 2010; Rey, 1996; Shields & Kiser, 2003). In our
study, workplace violence is operationalized as incidents of physical violence or verbal threat of violence by clients
directed at social service staff and captured by an agency’s reporting system.
The most frequently used study design to estimate prevalence of workplace violence is a survey of NASW
members (Jayaratne et al., 2004; Newhill, 1996), though this design could underrepresent minorities and those
who work in agencies and overrepresent social workers in private practice (NASW, 2008). Of the few studies that
took place at agency level or analyzed routinely collected injury data, one examined national BLS data oil time lost
as a result of physical assault (Respass & Payne, 2008), another gathered national data through local unions
(American Federation of State, County and Municipal Employees, 1999), and a third was a case study of an
agency in an urban setting with high crime and violence rotes (Bell, Mock, & Slutkin, 2002). Our study adds to the
knowledge base by reporting on incident data routinely collected by social service agencies in a geographic region.
Who Is Most at Risk for Workplace Violence?
Several studies have found increased risks for male compared with female social workers (Guterman et al., 1996;
Jayaratne et al., 2004; Newhill, 2003; Ringstad, 2009), though others have found increased risk among women
(Baines, 2005; Bell et al., 2002; Flannery, Fisher, & Walker, 2000). Other factors associated with increased risk
have included younger age (Jayaratne et al., 2004; Koritsas et al., 2010) and urban setting (Bell et al., 2002;
Shields & Kiser, 2003). The only study to look at staff educational and training level found no significant differences
between risk of being the target of workplace violence and different levels of training and education (Winstanley &
Hales, 2008). The use of physical restraint has been linked to workplace assault on staff in a residential psychiatric
setting (Flannery et al., 2000) and an intellectual disability group home setting (Hawkins, Allen, & Jenkins, 2005).
Where Are Incidents of Workplace Violence Most Likely to Occur?
In a study of violence against social workers, Newhill (2003) identified a hierarchy of risk based on primary area of
practice; criminal justice, drug and alcohol services, and child welfare were identified as “high risk” areas of
practice, and health care services and services for older people were identified as “low risk.” However, these
categories do not reveal whether elevated risk is related to the environmental setting, client population, or nature of
the services or intervention. In terms of risks associated with specific settings, Ringstad (2005) found increased risk
of workplace violence in inpatient and correctional institutions and schools. Although the dangers of providing
services in the client home are frequently discussed (Rey, 1996), no empirical study compares the likelihood of
incidents in home-based services with other social service settings.
Why Do Social Workers Fail to Report?
Underreporting of workplace health and safety incidents is a common phenomenon (Azaroff, Levenstein, &
Wegman, 2002). A recent study of U.S. workplaces estimated that 69 percent of work-related injuries are not
reported (BLS, 2010). Within social services specifically, one study found that only 18 percent of assaults on staff at
a state mental hospital were formally recorded during a 1-year period (Lion, Snyder, & Merrill, 1981). Among a
random sample of clinical social workers, 25 percent experienced an incident of workplace violence that they did
not report (McDonald & Sirotich, 2001). Reasons for failure to report included not thinking that the incident was
serious enough, a perception that violence is “part of the job,” a belief that nothing would be done, fear of being
blamed for the incident, and lack of institutional reporting policies (Lion et al., 1981; Lowe & Korr, 2008; McDonald
& Sirotich, 2001; Nobel, 2007; Rey, 1996).
How Are Data on Workplace Violence Collected in Social Service Settings?
Several studies have reported on the adequacy of safety policies in social service settings (Lowe & Korr, 2008;
Rey, 1996; Sarkisian & Portwood, 2003). The only study that evaluated compliance with Occupational Safety and
Health Administration (OSHA) safety guidelines for workplace violence prevention found that, among mental health
agencies, only half were in compliance (Lowe & Korr, 2008). There is little discussion in the literature of how
agencies collect data on incidents of workplace violence beyond the requirement of reporting an incident to a
supervisor. Public employees may not be covered by the Occupational Safety and Health Act of 1970 (P.L. 91-596)
(in Massachusetts, the setting for this study, they are not); therefore, reporting requirements vary between those
employed in the public sector and those in private or nonprofit settings (Massachusetts Coalition for Occupational
Safety and Health, 2011).
This Study’s Approach
Our aim was to describe workplace violence in Massachusetts social service agencies on the basis of routinely
collected incident data so as to aggregate and examine baseline data and identify areas that need to be explored
further. We took a literature- and field-informed approach to our inquiry that recognized existing variations in
regulatory requirements for recording workplace violence–related incidents in different agencies. This study
represents an attempt to develop a data collection strategy for gathering workplace violence incident data by using
a uniform survey to capture data from different agency reporting and recording systems. As members of a
statewide task force to maximize social worker safety, one of our goals was to make data, not typically publicized,
publically available to spark dialogue and agency- and community-level focus on workplace violence in social
services.
METHOD
Site of the Study
This study was conducted in the state of Massachusetts under the auspices of the Task Force for Maximizing
Social Worker Safety, a statewide stakeholder task force supported by the NASW, Massachusetts Chapter.
Sample and Procedure
Eligible participant agencies were agencies delivering social services in Massachusetts. The sample population of
agencies was identified through the membership lists of two statewide coalitions (the Human Services Coalition
and the Child Welfare League), all public agencies serving the state of Massachusetts, and the attendance list for a
summit on social work safety hosted by the Boston University School of Social Work in 2008. A committee of
experienced professionals in the field concluded that this approach made the survey available to the vast majority
of Massachusetts social service agencies. A letter describing the study was distributed electronically to the
executive director or chief executive officer of 200 human service agencies. Agencies who received the letter also
received a follow-up phone call to encourage participation. Data regarding social service staff were provided by
participant agencies. Eligible social service staff were employed by participant agencies during FY 2009 in the
capacity of clinical staff or direct care providers. Clinical staff were defined as employees holding a master’s degree
or higher. Direct care staff were defined as employees with a bachelor’s degree or less. The data collected by the
Internet-based survey relied on the data collection tools used in each agency. The study was conducted between
May and August 2010.
Participating agencies were asked to designate an individual with access to human resource documents to
complete an anonymous, Internet-based survey. Feedback from a pilot phase of the study indicated that, without
anonymity, agency leaders were reluctant to make workplace violence data public. As a main goal of our study was
to help make agency data public, we accepted that participation would be anonymous. A disclosure statement
included in the Internet-based survey explained the rights of the participants and the procedures for protecting the
anonymity of participating agencies. The study protocol was approved by the institutional review boards of NASW
and Salem State University (Salem, Massachusetts).
Measures
The measure used in this study was an anonymous Internet-based survey developed collaboratively with a large
research team. The survey gathered agency-level data for FY 2009 on agency characteristics (numbers of clients
served, numbers of employees, populations served, services provided, and settings where services were
provided), methods for collecting data on violent incidents, reported incidents of physical assault or violent threats
among direct care and clinical workers in the context of the use of physical restraints, reported incidents of physical
assault or violent threat among direct care and clinical workers in a non-restraint-related context, and setting of
reported incidents and perceptions of risk in each setting. Participants were provided space to give comments on
the topic of workplace violence at their agency at the end of the survey.
For the purpose of this study, workplace violence was defined as physical assault with or without injury and verbal
threat of physical assault by a client directed at a staff member. The use of restraints was not specifically defined in
the survey but was intended to capture the types of manual physical restraint of clients by staff in some social
service settings (Haimowitz, Urff, & Huckshom, 2006). Our decision to classify violent incidents as restraint and
non-restraint-related was based on feedback from a pilot phase of the study. Agency leaders indicated that
physically restraining clients could escalate violence and lead to assaults or incidents that should be differentiated
from physical assaults or verbal threats in contexts where restraints were not involved. The data collected by the
Internet-based survey relied on the data collection tools used in each agency.
Two focus groups composed of purposive samples of social service workers were conducted to contextualize the
survey findings to yield a range of attitudes and perspectives (Krueger & Casey, 2000). In addition to gathering
group viewpoints, small groups are potentially effective in drawing out nuanced and unanticipated information.
Procedurally, after the research team provided detailed information about informed consent and developing an
agreement to maintain each group’s confidentiality, the focus groups commenced. The interview guide used by the
facilitators (the first and second authors of this study) included open-ended questions addressing the central
research questions related to workplace safety. Each question was written to encourage study participants to
provide expansive responses. Polling techniques were also used to ensure that all study participants contributed to
the discussion (Vaughn, Schumm, & Sinagub, 1996). Identical questions were posed to both focus groups,
although individualized probes and clarifying questions were also used as necessary. The focus groups were
audiotape-recorded and transcribed.
Method of Data Analysis
Survey data were analyzed using SPSS, version 19 (IBM Corp., 2010). Bivariate statistical associations were
tested with chi-square tests. Odds ratios (ORs) and 95 percent confidence intervals (CIs) were calculated using
standard methods. Means were determined and compared with t tests. For the focus group transcript analysis, an
open coding approach was taken using the constant comparative method elucidated by Glaser and Strauss (1967).
RESULTS
Characteristics of Agencies and Workforce by Agency Type
In total, 40 agencies participated in this study (see Table 1). The number of clients served in these agencies ranged
from 120 to 89,000 (Mdn = 4,407; interquartile range [IQR] = 2,500-10,000) and the number of employees ranged
from 10 to 3,500 (Mdn = 117; IQR = 101-202). About half of the agencies who participated in this study primarily
served older adults (57.5 percent), but the agencies serving nonolder adult populations employed far more staff.
Results of this study represent the experiences of 2,627 (29 percent) clinical and 6,395 (71 percent) direct care
staff, with 9 percent of clinical staff and 17 percent of the direct care staff working primarily with older adults.
There were a total of 1,049 incidents reported of physical assault or verbal threat of violence. The rates of incidents
of physical assault or verbal threat among direct care or clinical staff were comparable (11 out of 100 per year in
older adult-focused services, compared with 12 out of 100 per year in nonolder adult-focused). However, a
significantly higher percentage of incidents among those working with older adults were verbal assaults (97
percent) compared with those working primarily with nonolder adults (40 percent; p < .001). Excluding verbal
assaults, the rate of incidents or injuries among those serving nonolder adult populations was significantly higher
(seven of 100 per year) than that of those working with older adults (0.31 of 100 per year; p < .001). Overall, a far
greater percentage of incidents reported in our study occurred in nonolder adult-focused services (87 percent) than
in older adult-focused services (13 percent). Also of note, there were no restraint-related injuries reported among
staff employed by agencies that primarily served older adults.
Physical Assaults and Threats among Direct Care and Clinical Staff
Data were collected with respect to both restraint-related and non-restraint-related injuries (see Table 2).
Participating agencies reported 173 restraint-related injuries for FY 2009; 143 (83 percent) of those injured were
direct care staff, who were twice as likely as clinical staff to suffer a restraint-related injury (OR = 2.02, 95 percent
CI [1.34, 3.06]). Direct care staff were 10 times more likely to experience a restraint-related injury requiring an
emergency room (ER) visit compared with clinical staff (OR = 10.13, 95 percent CI [3.09, 40.18]) and nearly four
times more likely to lose time due to their injuries (OR = 3.89, 95 percent CI [1.33, 12.81]).
Data on non-restraint-related incidents of physical assault and verbal threats were also explored. Agencies were
surveyed about four types of non-restraint-related injuries: verbal threat, physical assault without injury, physical
assault with injury, and death. Among direct care staff, 63 percent (n = 456) of non-restraint-related injuries were
verbal threats, compared with 83 percent (n = 92) among clinical staff; physical assault without injury accounted for
35 percent (n = 254) of non-restraint-related injuries compared with 14 percent (n = 16) among clinical staff; and 2
percent (n = 17) of direct care staff sustained injuries from a physical assault, whereas 3 percent (n = 3) of clinical
staff did. Overall, direct care staff were twice as likely to be threatened verbally at work (OR = 2.05, 95 percent CI
[1.62, 2.60]) and nearly five times more likely to be the victim of a physical assault (OR = 4.91, 95 percent CI [3.15,
7.73]).
Settings Where Incidents Occurred
Data were collected on the setting where incidents took place for each agency. The vast majority of restraintrelated incidents for direct care workers occurred in a group home setting (82 percent), whereas for clinical staff a
majority occurred in the psych-inpatient hospital setting (68 percent). For non-restraint-related incidents among
direct care staff, the settings where a majority of incidents occurred were either in psych-inpatient hospitals (35
percent) or day rehabilitation facilities (23 percent). The vast majority of non-restraint-related incidents among
clinical staff occurred in the psych-inpatient hospital setting (80 percent).
We also collected data on where non-restraint-related incidents were perceived as most likely to occur; the client
home was perceived to be the most risky setting for both direct care and clinical staff. This contrasts with the data
on where incidents actually occurred; 12 percent of non-restraint-related incidents among direct care staff occurred
in a client’s home, and 7 percent of incidents involving clinical staff occurred in this setting (see Table 3).
Differences in Data Collection among Agencies
Agencies responding to the survey were asked to report on how they collected information on incidents of assault
and threat. The most common methods were incident reports (18 of 40), workers compensation forms (15 of 40),
and human resources (11 of 40) (see Table 4).
Environment for Reporting Incidents of Workplace Violence
Two focus groups discussed the environment for reporting workplace violence incidents at two agencies; one group
was composed of licensed social workers and the other of program managers.
Licensed clinicians expressed a great deal of concern over safety despite few actual incidents at their agency. They
complained that a lack of safety planning left them feeling unsupported and entirely responsible for their own
safety. This lack of support was underscored by the perception that there was a “gendered” response from
management, where a mostly male administration discounted the concerns of a predominantly female clinical staff.
In addition, participants in this focus group indicated that, whereas some of the settings they worked in had safety
protocols, others did not. Even where protocols were in place, they were sometimes “forgotten” or “not
implemented” because of people “being too busy with their work.”
Program managers from a different agency discussed barriers to reporting incidents of workplace violence among
the staff they supervised and identified factors that encouraged reporting. Being seen as a “bad” social worker and
feeling as if the incident was one’s “fault” was seen as a huge barrier to reporting, particularly among newer social
workers. Being stressed for time was also seen as making a difference, in terms of both the limits on time as a
result of high caseloads and the amount of time required to complete reporting paperwork. Exacerbating these
problems was a shift in how services were billed, which left staff with no way to bill for time that was not directly
related to service delivery.
Program managers also reflected on the role they played as supervisors, both in creating an environment where
reporting and debriefing on incidents and threats were encouraged and in not minimizing their staff’s perceptions of
threats. There was broad consensus among this group of managers that peer debriefing sessions were an effective
and underused forum for evaluating workplace violence hazards.
DISCUSSION
The five key results of this study are as follows: (1) high rates of workplace violence against social service
providers, (2) a pattern of risk disparity showing significantly more risk for direct care versus clinical staff, (3)
perspectives on why staff in social service agencies might fail to report incidents of workplace violence, (4) more
incidents of workplace violence in inpatient and institutional versus other types of settings, and (5) insight on the
inconsistent approaches to collecting data on workplace violence.
Our most striking result is the risk disparity between direct care staff and clinical staff that is evidenced by their
statistically significant increased odds of nearly every type of assault or threat surveyed. Because direct care staff
have more contact hours with clients, their increased risk is logical. This logic is reflected in health service sector
data that show higher injury rates for health support staff compared with health practitioners (20.4 compared with
6.1 per 10,000; BLS, 2010). However, this inequality in risk between staff within the same work environment needs
further attention.
The study of “occupational health disparities” examines how population–level differences in health outcomes are
rooted in the work experiences of different groups, including access to employer-based health insurance (Krieger,
2010). “Two-tiered” employment in social services that divides licensed from unlicensed personnel has been
criticized for creating inequities in job quality (Baines, 2004). Being a lower paid, lower status worker in social
services appears, by the data in this study, to be associated with increased risk of exposure to workplace violence.
If direct care workers are also more likely to have poor access to health care, adverse health outcomes of exposure
to workplace violence could be exacerbated. We did not collect demographic data that would allow us to evaluate
patterns of risk inequity according to race-ethnicity, sexual orientation, socioeconomic status, or community
background, but this topic should be taken up in future research. For example, if socially disadvantaged groups are
overrepresented among those in lower status social service jobs, this raises a social justice issue that social
workers are compelled to examine.
Risk disparity was also apparent in the distribution of restraint-related incidents; direct care staff members were
significantly more likely to be injured, visit the Elk, or lose work time as a result of a workplace violence incident in
the context of physical restraint. Manual physical restraint may be defined as pressure applied by one or more staff
to restrict the movement of an individual during an episode of challenging behavior (Hawkins et al., 2005). Although
our study design does not allow us to make detailed assessment of why injuries occurred in the incidents recorded
in our study, our findings do suggest that use of restraints might be a component of the increased risk of physical
assault and injury borne by direct care staff. This finding is supported by reports that staff injuries and workers
compensation claims have dropped significantly in agencies where use of physical restraints has been eliminated
following advocacy efforts by the Child Welfare League of America (Caldwell & LeBel, 2010).
Our study also identified risk of workplace violence among clinical staff, including the death of a clinician during a
home visit. This underscores the shared interest of the social service team in making the work environment safer.
Indeed, a larger lens for analysis reveals occupational health disparities for those employed in the social service
sector compared with other industries in regards to prevalence of violent incidents (BLS, 2010) and lower pay and
benefits (Barth, 2003; Hudson, 2007).
Our qualitative results indicate that there are good reasons to expect that assaults and threats are underreported,
and they echo what has been reported in previous literature in terms of fear of being blamed for an incident, being
seen as a “bad” social worker, ignoring violence as “part of the job,” or working in an environment inhospitable to
reporting incidents. Given this, it is very likely that our study underestimated the number of actual incidents of
workplace violence in the agencies studied. Focus group participants worried that there was no requirement for
reporting “near misses,” noting that these incidents could contribute to stress in the workforce but fail to be reported
or addressed in the agency.
The majority of incidents of reported workplace violence in this study occurred in inpatient or institutional settings
rather than in a client home or community setting. This contrasts with the perception that the client home presents
greater risk. It would be important to understand whether this is a true difference or whether this phenomenon is
the result of reporting differences, as reporting protocols may be more developed or better regulated in inpatient
and institutional settings.
Our results on how agencies collect data suggest that there is a difference between collecting data for the purpose
of compensation versus using it for hazard assessment and violence prevention. Few agencies seemed to be using
their data for assessment and prevention purposes. A rare exception among our participants was the following,
reported in the Comments section of the survey:
After an incident of a violent or threatening
nature the employee files an incident report
describing the incident, setting, and who was
involved. The report is reviewed by the supervisor
and manager, a safety plan is developed
with the employee and is recorded in the incident
report. In some instances a debriefing
meeting is convened with the employee,
supervisor, and other staff to assess the impact
and for support. The report is submitted to the
central office for review, and to be recorded in
a worker safety database.
At the other extreme, another participant observed the following:
I learned that our workers’ compensation claims
for fiscal year’ 09 were just under $400,000. In
reviewing these incidents I was struck by the
violent nature of them. I am unsure whether
the agency addresses the psychological trauma
that may result from these attacks by clients.
Most of these assaults occur in our residential
programs; it seems as if the direct care workers
should be receiving combat pay.
This study has several limitations. The 40 agencies that participated in our study were self-selected; the effects of
bias limit the generalizability of the results. For example, more than half of the agencies that responded to our
survey served older adults, employed fewer staff, and experienced comparatively fewer incidents of physical
assault; given this, our description might underestimate the incidence of physical assault. However, the responding
agencies did represent a range of services, populations, sectors, and size and varied between those with few or no
incidents and those with many and between those who used restraints and those who did not. Because of the need
to protect the anonymity of the participating agencies, we are unable to characterize aspects of the services,
catchment area, and organizational setting. For instance, it would be helpful to know what region of the state was
served by each agency, because risk of workplace violence in social services has been shown to differ among
geographical regions (Green, Gregory, & Mason, 2003). For the purpose of our statistical analysis, we assumed
that each incident of reported workplace violence occurred to a different individual, which could have led to an
overestimate of incidence rates. Because our sample population of agencies included a vast range of sizes and
services offered, it reflects complex differences in working conditions related to workplace violence risk that we are
unable to consider given our study design. Although a strength of our study is that it gathers agency-level data that
explore the environments where workplace violence might occur and individual-level data on reported incidents,
because of anonymity requirements for study participation, our sample lacks the capacity to typify workplaces or
sufficiently describe individual characteristics. Because we did not gather rich data on each incident, we were not
able to analyze incidents of physical assault or verbal threat to inform prevention efforts. Future research should
capture agency, individual, and incident data that can fully characterize workplace violence hazards to individuals in
the environmental context. Doing this will require a commitment of agencies to promote safe working conditions
above concerns over negative publicity and