Ms. Donahue, Mr. Foster, and Mr. Felton are affiliated with the Center for Information Technology and Evaluation Research, New York State Office of Mental Health, Albany. Dr. Covell and Dr. Essock are with the Department of Psychiatry, Division of Health Services Research, Mount Sinai School of Medicine, New York City. Dr. Essock is also with the Mental Illness Research, Education, and Clinical Center, James J. Peters Veterans Affairs Medical Center, the Bronx. Send correspondence to Dr. Essock, Department of Psychiatry, Mount Sinai School of Medicine, 1 Gustave L. Levy Place, Box 1230, New York, NY 10029-6574 (e-mail: email@example.com). This report from Project Liberty is part of a special issue of Psychiatric Services commemorating the five-year anniversary of the September 11, 2001, attacks. Susan M. Essock, Ph.D., served as guest editor.
In response to the September 11, 2001, attacks on the World Trade Center, the Federal Emergency and Management Agency (FEMA) and the Center for Mental Health Services rapidly funded Project Liberty, an unprecedented large-scale mental health intervention. Administered by the New York State Office of Mental Health (NYOMH), Project Liberty provided short-term public education, outreach, and individual and group crisis counseling services to residents of the greater metropolitan area. This federally funded program emphasized service provision in multiple naturalistic community settings, such as churches and social clubs, through a large network of trained mental health professionals and paraprofessionals from more than 100 agencies (1).
Investigations of the short-term and longer-term effects of terrorist activities in the United States are leading to more sophisticated longitudinal conceptualizations of the development and remission of psychiatric impairments after such disasters, particularly posttraumatic stress disorder (PTSD), among survivors of terrorist attacks (2,3). Such investigations are providing more complex perspectives about interrelationships among partial (4), subthreshold (5), and delayed-onset PTSD (6) and relationships between various manifestations of PTSD and subsequent health outcomes (2,7).
Although the mental health impact of the terrorist attacks was initially severe for many individuals within commuting distance of the World Trade Center (1,8), the prevalence of PTSD symptoms in that population declined dramatically five months after the attacks, from 7.5 to .6 percent for probable PTSD and from 17.4 to 4.7 percent for subsyndromal PTSD (9). A minority of individuals, however, continued to report relatively severe chronic psychological distress. Such individuals manifested specific risk factors, such as being near or injured around the site of the attacks, losing their jobs as a result of the attacks, being unmarried, having low social support, and reporting previous trauma (9,10,11). Similar risk factors have been identified recently in assessments of survivors of the attack on the Pentagon (2), in large-scale surveys (12), and in a companion article in this issue (13).
Less is known about how service use varies as a function of demographic characteristics, particularly gender and race and ethnicity, of those affected by terrorism. Such information provides an important window into the extent to which services are being provided in culturally appropriate, acceptable ways. Estimates based on prior research and surveys conducted by the New York Academy of Medicine indicated that, over the year after September 11, 2001, a total of 422,000 individuals in the World Trade Center disaster area would meet diagnostic criteria for PTSD and 129,000 would seek assistance (14). Boscarino and colleagues (10), however, found that, compared with Caucasians, African Americans and Latinos were significantly less likely to report using mental health services or taking psychotropic medications between one and four months after the attacks.
Reasons for these disparities in service use are unclear. Recent survey research indicates that non-Hispanic blacks report lower lifetime risk of mood and anxiety disorders than Caucasians (15). In the survey Hispanics and non-Hispanic blacks, however, were more likely to report more persistent disability from mood disorders and mood and anxiety disorders, respectively (15). This report supplements such survey data by examining actual service use, determined by analysis of Project Liberty encounter logs for services delivered from a few weeks through 2.5 years after the attacks. Our intent is to profile demographic characteristics of those who used services and evaluate changes in use over time as a function of demographic characteristics.
Study participants and variables
To meet administrative requirements for reimbursement, Project Liberty providers completed service encounter log forms after each service. These logs indicated date, type of service (individual crisis counseling, group crisis counseling, or group public education), location of service (city borough or surrounding county), duration of session, and service setting (provider office or community setting, such as a school, place of employment, community center, or individual's home). Under FEMA funding conditions, all services provided under Project Liberty were free to service recipients and anonymous. For individual crisis counseling sessions, providers indicated whether the service was a first visit or a follow-up visit, whether the individual received a referral for further services, and demographic information about age, gender, ethnicity, and preferred language. Race and ethnicity were coded as observed by the counselor into one of the following categories: Caucasian, Hispanic origin, African American, Asian or Pacific Islander, Middle Eastern, American Indian or Alaska Native, other, or unknown.
Demographic data were collected for only individual counseling sessions; therefore, the analyses in this report are limited to this service type. Data were from the 687,848 individual service encounters of 753,015 total service encounters (91.3 percent) for the period September 2001 through December 31, 2003 (service encounters provided by the school-based Project Liberty program administered by the New York City Department of Education were not included in our analyses because complete information was not available when the data were extracted in May 2004). Individual crisis counseling services assisted participants in understanding their current situation and reactions, reviewing their options, addressing their emotional support, and linking them with other individuals and agencies who might assist them. Individual counseling sessions tended to be brief, with just over half (57 percent) of first sessions lasting less than 30 minutes and only 12 percent lasting longer than one hour. However, follow-up sessions were likely to be longer, with 57 percent lasting between 30 and 60 minutes and 29 percent lasting more than one hour.
Because we conducted the analyses on secondary anonymous data, the Mount Sinai School of Medicine Institutional Review Board found the project exempt.
Characteristics of service recipients were examined by using descriptive statistics and logs for first visits by individuals, which were linked to census statistics by county. Census information was dated April 1, 2000. To compare use with estimated need, logs for first visits were compared with the proportion of individuals expected to have a lifetime diagnosis of PTSD or of any DSM-IV disorder. Need was calculated by applying prevalence rates from the National Comorbidity Survey (15,16) to census information. For both comparisons, goodness-of-fit tests were applied to examine whether the likelihood of using Project Liberty services differed significantly by gender and racial or ethnic group compared with rates expected on the basis of general census data or by estimated need.
To ameliorate concern that extremely large sample sizes produce significant findings with very small differences, reporting is limited to differences that were both statistically significant (p<.05) and meaningful, which was defined as a difference between observed and expected frequencies greater than or equal to 10 percent of the subsample for that analysis—a difference we felt would be large enough to suggest changes in service delivery to target underserved populations.
Counselors completing service logs coded race and ethnicity by using a single variable. Therefore, to permit comparison of this information with census information, where race and ethnicity are coded separately, we reduced census data on race and ethnicity to a single variable. Any individual identified as Hispanic on the census form was coded as Hispanic, regardless of racial identity. Non-Hispanic individuals were coded according to race.
We used mixed regression models for nominal longitudinal data (8) to examine whether the likelihood of individual service recipients' being from a particular group changed over time in either a linear fashion (month since September 11, 2001, where month 0=September 11 through September 30, 2001, month 1=October 2001, and so on) or a curvilinear fashion (month2 [month squared] since September 11, 2001). The same statistical model was used for both first-time and follow-up visits. In these models, individual service use was predicted from the independent variables of gender, age, race and ethnicity, and preferred language and by service location (in the community or in the provider office).
Correlations were examined between clients' demographic characteristics and the proportion of follow-up visits among the 168 providers who offered individual services. We specifically examined whether more follow-up visits were related to gender, age, ethnicity or race, and preferred language.
Geographical distribution of services
Individuals receiving services in the five New York City boroughs represented most of the first visits, increasing from 85 percent in the first month to 90 percent in the second month and to 95-99 percent by six months after the disaster. Individuals in New York City also received most of the follow-up sessions over time, ranging from 60 percent in month 1 to a level of 90-99 percent from month 10 until the end of Project Liberty services in December 2003 (p<.001 for month and month2, respectively).
Distribution of services by demographic characteristics
Most Project Liberty services were provided to adults and to individuals who spoke English. The gender distribution of individual first visits was similar to population gender distribution in every county and borough and with the proportion estimated to have any DSM-IV disorder, with three exceptions. Only in Dutchess and Rockland counties did more women than expected (p<.001) access Project Liberty services. Also, in comparison with the estimated number of individuals with lifetime PTSD, fewer women than expected (p<.001) accessed Project Liberty services.
By contrast, in the five New York City boroughs and among those with estimated lifetime PTSD or any DSM-IV disorder, there were significant variations in the distribution of service use by race and ethnicity compared with expected rates. In Manhattan, more African Americans and fewer Caucasians than expected used services (46,311 observed compared with 25,982 expected African Americans and 54,507 observed compared with 77,912 expected Caucasians). In Queens, more African Americans than expected used services (33,705 observed compared with 20,790 expected). In addition, in Suffolk and Ulster counties, more Hispanics (1,156 observed compared with 535 expected in Suffolk County and 95 observed compared with 22 expected in Ulster County) and fewer Caucasians (2,979 observed compared with 4,002 expected in Suffolk County and 225 observed compared with 299 expected in Ulster County) than expected used services. Similarly, on the basis of estimated lifetime PTSD or any DSM-IV disorder (15), fewer Caucasian (137,173 observed compared with 210,273 expected) and more African American (126,388 observed compared with 80,891 expected) and Hispanic (102,960 observed compared with 75,357 expected) individuals than expected accessed Project Liberty services.
Patterns of service use over time
Figure 1 summarizes the frequencies of first and subsequent counseling visits over time. First-time users of Project Liberty services increased slowly during the five months after September 11, then rapidly escalated, peaking at eight months and remaining relatively high throughout the ensuing year. Frequency of follow-up contacts increased steadily through the first ten months, decreased for a few months, and then generally increased again until peaking in month 20. The X axis of Figure 1 indicates temporal events that may have influenced service use. These events include other disasters, such as the crash of American Airlines Flight 587 in New York City; temporal events, such as the one-year anniversary of the attacks; and changes in the ways providers were paid, from fee for service, where providers were paid at specified rates on the basis of the duration of contact, to a cost-based system, where payments were made to providers on the basis of negotiated budgets for the provision of an anticipated volume of service.
Demographic differences in service use over time
Mixed regression models for nominal longitudinal data revealed no significant differences in access by gender over time, with women accounting for 44 to 61 percent of all first visits in any month and 44 to 62 percent of all follow-up visits.
Most of the first visits (85 percent) were made by adults (90-95 percent in most months). The mixed regression models indicated that the proportion of first visits by adults varied somewhat through time (z=16, p<.001, for month since the attacks and z=-16, p<.001, for month2). For follow-up services, the proportion of services provided to adults represented 70 to 80 percent during the year after the terrorist attacks. Adult follow-up visits decreased over the second year, to a low of 57 percent at month 20, but increased to 85 percent by the two-year anniversary of the attacks (z=-25, p<.001, for follow-up visit by month and z=28, p<.001, for follow-up visit by month2).
Most of the first visits (83 percent) were provided for English-speaking individuals. Proportions ranged from 67 to 88 percent by six months after September 11, dropped gradually to about 75 percent from months 14 to 23, and then increased again to about 85 percent beginning at about month 24 (z=-40, p<.001, for month and z=40, p<.001, for month2). The proportion of follow-up visits for English-speaking individuals increased from 77 to 90 percent within six months after the attacks, where it remained more or less steady through time (z=18, p<.001, for follow-up visit by month and z=-13, p<.001, for follow-up visit by month2).
Figure 2 presents relative percentages of first-time visits over time by racial and ethnic designation. Across 27 months, Caucasians used the largest proportion (39 percent) of first-time visits, followed by African Americans (26 percent) and Hispanics (25 percent). Over time, however, African Americans came to represent the largest racial and ethnic group of new users (about 40 percent in months 5 through 8 after the attacks). Individuals of Hispanic origin represented the largest racial and ethnic group of new users 19 through 22 months after the disaster (about 35 percent). Caucasians represented the largest group (typically 50 percent or more in any given month) returning for follow-up visits (Figure 3), particularly during the months corresponding to the one-year and two-year anniversaries of the attacks.
The likelihood of having a follow-up visit was not associated with gender. However, as providers offered proportionately more follow-up visits, the proportion of first visits increased for English speakers (r=.26, p<.005) and Caucasians (r=.38, p<.001) and the proportion of first visits decreased for adults (r=-.22, p<.005), Hispanics (r=-.25, p<.005), and African Americans (r=-.26, p<.005).
Location of services over time
Aggregating the data across 27 months, we found that 73 percent of individual first-time crisis counseling services took place in community settings rather than in providers' offices. This percentage began at about 75 percent and accelerated rapidly to about 90 percent by month 6 (p<.001 for month and month2). Follow-up services were less likely to take place in the community, but use of community sites for follow-up increased from 45 to 70 percent by month 10 and then dropped off again to 35 percent by month 12. After the one-year anniversary, the likelihood of follow-up services taking place in the community increased again to 60 percent by month 14 and then gradually fell to 30 percent in months 24-27 (p<.001 for month and month2).
Project Liberty provided over 650,000 free crisis counseling encounters to individuals in the 27 months after the World Trade Center disaster. Overall use services was consistent with prior estimates of service need (14). Most services were provided to individual adults within the five boroughs of New York City. Most services were provided in community settings, rather than in practitioner offices.
A substantial number of individuals sought first visits from Project Liberty up to 27 months after the disaster. Spikes in help seeking occurred five to eight months after the attacks, and service use remained high for nearly two years afterward. In addition, follow-up visits increased curvilinearly over time, suggesting that a small but meaningful percentage of the population continued to seek ongoing counseling for several years after the attacks. These findings point out the need for long-term service availability. The findings also are consistent with other reports documenting that particular individuals experience serious symptoms consistent with PTSD, depression, anxiety, and associated functional impairment for years after experiencing trauma (2,3,5,6).
We note that in both Manhattan and Queens the proportion of African Americans seeking Project Liberty services exceeded expected rates on the basis of federal census information and estimates of lifetime diagnosis of PTSD, mood disorders, or anxiety disorders (15,16). In addition, African-American and Hispanic individuals constituted the largest groups of first-time service users between five to eight and 19 to 22 months after the attacks, respectively. Although most of the first-time and follow-up services were provided to English-speaking clients, up to 33 percent of first visits and 23 percent of follow-up services were provided to people whose preferred language was other than English. These outcomes are in contrast to survey findings (10) that members of minority groups, in particular African-American and Hispanic individuals, were less likely than Caucasians to use disaster-related mental health treatment services or psychotropic medications for up to one year after the terrorist attacks.
Also notable is the lack of gender differences in accessing services when compared with census information. When use in relation to the expected proportion of individuals with lifetime PTSD was examined, women appeared underrepresented. Reasons for the underrepresentation of women in response to this traumatic event are unclear, given that lifetime rates of PTSD are based on exposure to trauma and, historically, women are more likely to report being victims of trauma (12).
These findings must be interpreted cautiously, but they appear to suggest that provision of free counseling services in community settings may have reduced stigma and improved access, particularly for individuals from racial or ethnic minority groups and for men. Provision of services for several years after the terrorist attacks also permitted individuals who were slower to access help to obtain services. Community-oriented, longer-term service availability in response to future terrorist activities may facilitate effective outreach to traditionally underserved populations.
Minority status and disparities in access to quality health care, however, are confounded in nonlinear ways with socioeconomic differences (17). In this investigation, it was impossible to untangle whether use of crisis counseling was attributable to services' being free and readily available in naturalistic settings or to a differential impact of trauma experienced by specific ethnic minorities. Inclusion in future surveys of questions assessing indicators of socioeconomic status, such as education, job description, and family income, would help untangle these possible influences.
Relatively more Caucasians received follow-up visits. This finding may reflect the preference of service recipients from minority groups or service delivery practices of providers. This latter possibility fits with existing literature on ethnic health care disparities (18) and is bolstered somewhat by the finding that counselors who provided more follow-up visits were less likely to serve minority clients at first visits.
Our results are limited by several important constraints. The data we analyzed came from service encounter logs required of providers to obtain reimbursement. Rules for reimbursement changed from a fee-for-service to cost-based system just before the one-year anniversary of the attacks on the World Trade Center, suggesting that the trends reported herein possibly reflect both a combination of the service needs of the population of New York City and provider response to changing payment rules. These service encounter logs were completed by Project Liberty counselors, whose primary job was to provide crisis services in a highly unusual and demanding situation, rather than from self-reports of service recipients. More important, all data were recorded anonymously, per FEMA requirements, so there was no way of determining how many individuals had follow-up visits or what the average number of follow-up visits was for any given individual. Finally, race and ethnicity were coded as only one variable, and information was obtained by provider observation and not verified by client report.
Project Liberty provided free crisis counseling services in the aftermath of the attacks on the World Trade Center to people in the greater New York City metropolitan area. These services reached individuals of diverse racial and ethnic backgrounds whose demographic characteristics closely approximated those in the communities from which they came. The NYOMH strategy of contracting with a large number of agencies on the basis of the belief that these agencies would be most effective in reaching out to affected community members appears to have been effectively implemented. The findings also underscore the continuing need for intervention years after such attacks. Administrators responsible for monitoring the penetration and acceptability of services in response to disasters can benefit from information obtained from service encounter logs as simple as those used during Project Liberty.
This evaluation was funded by grant FEMA-1391-DR-NY (titled "Project Liberty: Crisis Counseling Assistance and Training Program") to New York State from the Federal Emergency Management Agency. The Center for Mental Health Services of the Substance Abuse and Mental Health Services Administration administered the grant. The authors express their appreciation to Katherine M. Shear, M.D., George Allen, Ph.D., and Carlos T. Jackson, Ph.D.