M. B. Kushel, 1 G. Colfax, 4 K. Ragland, 2 A. Heineman, 2 H. Palacio, 5 and D. R. Bangsberg 3

Transcription

1 HIV/AIDS MAJOR ARTICLE Case Management Is Associated with Improved Antiretroviral Adherence and CD4 + Cell Counts in Homeless and Marginally Housed Individuals with HIV Infection M. B. Kushel, 1 G. Colfax, 4 K. Ragland, 2 A. Heineman, 2 H. Palacio, 5 and D. R. Bangsberg 3 1 Division of General Internal Medicine, 2 Department of Epidemiology and Biostatistics, and 3 Epidemiology and Prevention Interventions Center, Division of Infectious Diseases and The Positive Health Program, San Francisco General Hospital Medical Center, University of California, San Francisco, and 4 Epidemiology and Intervention Research Section, AIDS Office, San Francisco Department of Public Health, San Francisco, California; and 5 Harris Public Health and Environmental Services, Houston, Texas Background. Case management (CM) coordinates care for persons with complex health care needs. It is not known whether CM is effective at improving biological outcomes among homeless and marginally housed persons with human immunodeficiency virus (HIV) infection. Our goal was to determine whether CM is associated with reduced acute medical care use and improved biological outcomes in homeless and marginally housed persons with HIV infection. Methods. We conducted a prospective observational cohort study in a probability-based community sample of HIV-infected homeless and marginally housed adults in San Francisco, California. The primary independent variable was CM, defined as none or rare (any CM in 25% of quarters in the study), moderate (125% but 75%), or consistent (175%). The dependent variables were 3 self-reported health service use measures (receipt of primary care, emergency department visits and hospitalizations, and antiretroviral therapy adherence) and 2 biological measures (increase in CD4 + cell count of 50% and geometric mean HIV load of 400 copies/ml). Results. In multivariate models, CM was not associated with increased primary care, emergency department use, or hospitalization. Moderate CM, compared with no or rare CM, was associated with an adjusted b coefficient of 0.13 (95% confidence interval [CI], ) for improved antiretroviral adherence. Consistent CM (adjusted odds ratio [AOR], 10.7; 95% CI, ) and moderate CM (AOR, 6.5; 95% CI, ) were both associated with 50% improvements in CD4 + cell count. CM was not associated with geometric HIV load!400 copies/ml when antiretroviral therapy adherence was included in the model. Study limitations include a lack of randomization. Conclusion. CM may be a successful method to improve adherence to antiretroviral therapy and biological outcomes among HIV-infected homeless and marginally housed adults. Case management (CM) seeks to coordinate medical care for individuals with complex medical problems [1 4]. For HIV-infected patients, CM manages complex medication regimens, coordinates primary medical care, and assists with referrals to housing, mental health, and substance abuse services [1 4]. HIV-infected people living in poverty are disproportionately affected by Received 2 January 2006; accepted 4 April 2006; electronically published 8 June Reprints or correspondence: Dr. David Bangsberg, Epidemiology and Prevention Interventions Center, San Francisco General Hospital, POB 1372 UCSF, San Francisco, CA (db@epi-center.ucsf.edu). Clinical Infectious Diseases 2006; 43: by the Infectious Diseases Society of America. All rights reserved /2006/ $15.00 unstable housing, substance abuse, and mental illness [1 8]. These factors contribute to low rates of receipt of ambulatory health care and uptake of antiretroviral therapy [9, 10] and high rates of emergency department use and hospitalization [6, 11]. Most HIV CM is funded through the Ryan White Comprehensive AIDS Resources Emergency (CARE) Act of This program provides grants to states and municipalities to promote access to health care for HIVinfected persons living in poverty. In 2002, a total of 332,377 individuals received a total of 3,689,838 CM visits funded by the Ryan White CARE Act [12]. Cross-sectional studies of CM in the general HIVinfected population found that CM was associated with decreased reports of unmet needs and higher use of 234 CID 2006:43 (15 July) HIV/AIDS

2 multidrug antiretroviral therapy (ART) regimens [4, 13]. Those studies did not find an association with use of primary care and did not assess biological outcomes. A randomized, controlled study of a brief CM intervention among persons with newly diagnosed cases of HIV infection found that CM was associated with uptake and with use of HIV primary care services [14]. In this study, we examined CM in a community-based sample of low-income, HIV-infected individuals. We hypothesized that CM is associated with antiretroviral use and adherence, increased access to primary care, decreased emergency department visits and hospitalizations, improved CD4 + cell counts, and decreased HIV load. METHODS Overview We studied participants in a prospective, observational cohort of HIV-infected homeless and marginally housed persons. We assembled our cohort by conducting 2 screenings of homeless and marginally housed persons from April 1996 through December 1997 and from April 1999 through April 2000 in 3 San Francisco neighborhoods. We set out to represent the homeless population by service-use strata using the method of Burnham and Kogel [15]. We constructed a multistage cluster sample that was stratified into shelters, free meal programs, and singleroom occupancy hotels. After each screening, in which we tested all screened participants for HIV infection, we invited all who were HIV seropositive (8.8% in and 14.5% in 1999) to join the cohort. The cohort underwent structured quarterly interviews that focused on their health status, use of and adherence to antiretroviral medications, health service use, housing and health-related behaviors. At quarterly intervals for a 15-month period, between March 2001 and July 2002, participants at the same time underwent a structured interview focused on the use of CM services. At each visit, we drew participants blood to measure their HIV load and CD4 + cell counts [16]. The study was approved the Institutional Review Board at University of California, San Francisco. Definition and Measurement of CM Study staff interviewed all participants about their use of CM. We defined a case manager as a person that (1) worked at an agency, (2) talked with participants about services, and (3) helped participants to get services. We instructed participants that they should not include money managers or doctors but could include nurses and social workers. We asked participants to note which case managers they had seen in the past quarter and at which agency they had seen them. We asked participants whether and how often they had met with each of their case managers since the prior interview. We confirmed participant reports by interviewing the identified case managers. We only considered a participant to have had a case manager when we verified this information. Independent s We categorized CM use into 3 categories (none or rare, moderate, or consistent) on the basis of the percentage of quarters in the study that the respondent reported having met with their case manager, with 25% of quarters in the study defined as no or rare CM use, 125% but 75% defined as moderate CM use, and 175% defined as consistent CM use. In our multivariate models, we separately compared participants with moderate or consistent CM use with participants with no or rare CM use. Sociodemographic s At the baseline interview, we ascertained participants sex, age, and race or ethnicity. We asked participants to complete a residential calendar, describing where they had lived since the prior interview. Participants who reported having spent at least 90% of their nights in a residential hotel or an apartment were considered to be marginally housed. Those who had spent!90% of their nights in a residential hotel and at least 1 night sleeping on the street or in a shelter were considered to be homeless. Dependent s Use of health services. We asked participants to report whether and how often they had visited their primary care physician for scheduled and acute-care visits each quarter. We considered people to have received primary care if they had a mean of 1 scheduled primary care visit every quarter. We asked whether and how often they have been to the emergency department or been admitted to the hospital. Emergency department visits that resulted in a hospitalization were not counted separately. We dichotomized emergency department use and hospitalization as none or any. Health status. We assessed health status using the Short Form with 36 Questions (SF-36 ). The SF-36 is a self-administered functional health status instrument that has been validated in a similar population [17]. It includes 2 main dimensions: physical and mental health. The SF-36 is scored from 0 to 100, with higher scores representing better outcomes. The Physical Composite Score (PCS) and Mental Composite Score (MCS) both have means of 50 and SDs of 10 in the US general population [18]. We calculated the PCS and MCS for each participant at baseline and categorized responses as 50, 35 but!50, and!35. Biological measurements. At each quarterly visit, we obtained blood samples for determination of CD4 + cell count (performed at Unilab; Tarzana, CA) and HIV RNA load (performed at Roche Amplicor; Branchburg, NJ). We determined HIV/AIDS CID 2006:43 (15 July) 235

3 Table 1. Characteristics of HIV-infected homeless and marginally housed adults at baseline. Subjects (n p 280) Sex Female 47 (16.8) Male 233 (83.2) Race/ethnicity White 116 (41.4) African-American 121 (43.2) Latino 18 (6.4) Other 25 (8.9) Age at baseline Mean years SD !40 years old 102 (36.4) 40 years old 178 (63.6) Homeless during past quarter a 79 (28.2) Marginally housed during past quarter b 201 (71.8) CD4 + cell count at baseline 400 cells/ml 122 (43.6) cells/ml 79 (28.2) cells/ml 38 (13.6) cells/ml 15 (5.4)!50 cells/ml 11 (4.2) CD4 + cell count nadir!200 cells/ml 138 (49.3) Baseline HIV load!100 copies/ml 61 (23.2) copies/ml 16 (6.1) 1400 copies/ml 186 (70.7) Antiretroviral therapy at baseline None 140 (50.0) 1 or 2 antiretrovirals 6 (2.1) 3 antiretrovirals 134 (47.9) SF-36 PCS score Mean score SD (33.1) 35 but! (39.3)!35 76 (27.6) SF-36 MCS score Mean score SD (27.6) 35 but!50 78 (28.4)! (44.0) Any illicit drug use in the past 30 days c No 166 (59.3) Yes 114 (40.7) Any injection drug use in the past 30 days No 214 (76.4) Yes 66 (23.6) Drugs used in the past quarter Crack cocaine or powder cocaine 70 (25.0) Heroin 33 (11.8) Methamphetamines 52 (18.6) Heavy alcohol use in the past month d 27 (9.6) (continued) Table 1. (Continued.) Subjects (n p 280) Emergency department use in the past quarter 56 (20.1) Hospitalization in the past quarter 27 (9.7) Primary care visits in the past quarter None 88 (31.4) (52.9) (15.7) NOTE. Data are no. (%) of subjects, unless otherwise indicated. MCS, Mental Health Component Scale; PCS, Physical Component Scale; SF-36, Short Form with 36 Questions. a Homeless was defined as at least 1 night on street or in shelter in the past quarter. b Marginally housed was defined as 90% of nights in single-room occupancy hotel or apartment in past quarter and no nights on street or in shelter. c Includes use of crack cocaine, powder cocaine, heroin, and methamphetamines. d Defined as 5 alcoholic drinks on a day when alcohol was consumed. participants lifetime CD4 + cell count nadir on the basis of medical record review. We compared CD4 + cell count values at the baseline visit with those in the follow-up period. We used the CD4 + cell count from the first and last visits to measure the percentage increase in CD4 + cell count, dichotomizing the data at the 50% increase level (i.e.,!50% vs. 50%). We calculated the geometric mean HIV load over the study period and then classified participants with a geometric mean HIV load 400 copies/ml as having undetectable HIV load at follow-up. The remainder of the participants were classified as having a detectable HIV load. Antiretroviral use and adherence measures. We determined the use of HIV antiretroviral medication on the basis of client self-report during each quarterly interview. We defined appropriate use of ART for baseline and each quarter as concurrent use of 3 antiretroviral medications for a restricted sample of the 219 participants who had a CD4 + cell count nadir of!350 cells/ml (participants with a CD4 + cell count nadir of 1350 cells/ml would not meet general criteria for use of ART) [19, 20]. We measured adherence using a standard 3-day structured interview [21]. We estimated adherence by calculating mean adherence over all of the quarters, with 0 assigned to individuals who were eligible for but did not receive therapy. Participant reimbursement. The participants received $25 for each visit. Analysis. We analyzed the data using logistic regression, except for adherence, which was measured using linear regression. s with P values of!.25 in bivariate analysis were entered into the logistic regression models. For the biological markers, we performed each model with and without adherence entered into the model. 236 CID 2006:43 (15 July) HIV/AIDS

4 Table 2. Case management use, by CD4 + cell count nadir. Case management use CD4 + cell count nadir!350 cells/ml (n p 219) No (%) of subjects CD4 + cell count nadir 350 cells/ml (n p 56) All (n p 280) None or low a 88 (40.2) 28 (45.90) 116 (41.4) Moderate b 53 (24.2) 14 (22.95) 67 (23.9) Consistent c 78 (35.6) 19 (31.15) 97 (34.7) a Any contact with a case manager in 25% of quarters enrolled in the study. b Any contact with a case manager in 125% but 75% of quarters enrolled in study. c Any contact with a case manager in 175% of quarters enrolled in the study. Propensity analysis. Because there was not random assignment of participants into CM or control groups, we conducted a propensity analysis to control for baseline differences in the intervention (CM) group and the nonintervention group. The propensity score is a model-based predicted probability of receiving the intervention of interest [22, 23]. The probability of being in the intervention group is determined for each strata of explanatory variables and is then summarized by combining probabilities weighted by the inverse of the variances to estimate the overall effect. The scores are then entered into a final model as a covariate to account for confounding by indication. RESULTS Baseline Characteristics Participant demographic characteristics. In the screening portion of the study, 411 persons had test results positive for HIV infection (table 1). Of these 411 persons, 330 (80%) agreed to participate in the cohort. Before the start of the study, 35 individuals died, 13 were lost to follow-up, and 2 dropped out, leaving a total of 280 eligible participants. Of the 280 participants, the majority (83.2%) were men; their mean age was 43 years. The majority of participants identified as white (41.4%) or African-American (43.2%). Most (71.8%) of the respondents were marginally housed; 28.2% were homeless. CM use. At the baseline interview, over one-half of the respondents reported currently having a case manager (53.1%). Of those who had a case manager, 17.6% had 11 case manager, with a mean ( SD) of case managers. Baseline substance abuse and mental illness. A total of 40.7% of the participants reported drug use in the past month; almost one-quarter (23.6%) reported having injected drugs in the past 30 days. One-quarter (25.0%) of the participants reported using cocaine (crack and powder cocaine), 18.6% reported use of methamphetamines, and 11.8% reported heroin use. Approximately one-tenth (9.6%) reported daily alcohol intake of 14 drinks per day. Baseline health status. At baseline, the mean SF-36 PCS of the overall sample was 41.2 (range, ). The mean MCS was 44.5 (range, ). The participants median CD4 + cell count was 374 cells/ml. Nearly one-half (47.9%) of the participants reported using 3 antiretroviral drugs concurrently. Nearly one-third (29.3%) of the participants had an undetectable HIV load ( 400 copies/ml) at study entry; 49.3% of those receiving ART at baseline had an undetectable viral load at baseline. Health service use at baseline. At baseline, nearly 70% of respondents reported having had a primary care visit in the previous quarter; one-fifth (20.0%) reported having had an emergency department visit, and 9.6% reported having been hospitalized. At baseline, the mean ( SD) duration of ART was similar for the 3 CM groups: no or rare CM use, months; moderate CM use, months; and consistent CM use, months. The median duration of treatment was 11.0 months, 14.0 months, and 12.0 months for the no or rare CM use group, the moderate CM use group, and the consistent CM use group, respectively. Outcomes Follow-up. Twenty-three of 280 subjects were lost to followup (defined as 12 consecutive missed interviews). CM use. Respondents had a mean of 27.7 contacts with a case manager over the 5 quarters of the study period. The number of contacts decreased over the course of the study, from a mean of 9.2 contacts in the first quarter to 3.2 contacts in the final quarter. Among those reporting having a case manager at any point in the study, 148 (52.9%) reported having 11 case manager; 114 (40.7%) reported receiving their services at 11 agency. A total of 41.4% of respondents were classified as having no or rare CM, 23.9% were classified as having moderate CM, and 34.7% were classified as having consistent CM (table 2). Health services use. During the study period, 203 (72.5%) of respondents averaged 1 visit per quarter with their primary care clinician, 112 (41.0%) had at least 1 emergency department visit, and 63 (23.1%) had at least 1 hospitalization. Antiretroviral treatment and adherence. A total of 115 (41.2%) of the participants reported using ART for at least 1 quarter, whereas 52.8% of participants with a CD4 + cell count nadir of!200 cells/ml and 44.7% of participants with a CD4 + cell count nadir of!350 cells/ml reported using ART. Among those who used ART, 46 (16.4%) reported using it in each quarter of the study. Biological markers. Of the 280 participants, 116 (41.4%) had at least 1 viral load measurement of!400 copies/ml; 46 (16.4%) had 400 copies/ml in all quarters. Of the 219 individuals who had CD4 + cell count nadirs of!350 cells/ml, HIV/AIDS CID 2006:43 (15 July) 237

5 Table 3. Factors associated in logistic regression models with receipt of health care among 280 homeless and marginally housed HIV-infected adults. Routine primary care a Emergency department care b Hospitalization c (n p 203) (n p 112) (n p 63) Rate, % Unadjusted OR Adjusted OR Rate, % Unadjusted OR Adjusted OR Rate, % Unadjusted OR Adjusted OR Sex Female Male ( ) 0.4 ( ) ( ) ( ) Race/ethnicity Nonblack Black ( ) ( ) ( ) 0.7 ( ) Age at baseline!40 years years ( ) ( ) ( ) Housing status for last quarter Marginally housed d Homeless e ( ) 0.7 ( ) ( ) ( ) CD4 + cell count nadir 200 cells/ml !200 cells/ml ( ) 1.4 ( ) ( ) 1.8 ( ) ( ) 1.8 ( ) SF-36 PCS score but! ( ) 1.0 ( ) ( ) 1.2 ( ) ( ) 3.0 ( )! ( ) 1.6 ( ) ( ) 1.7 ( ) ( ) 7.7 ( ) SF-36 MCS score but! ( ) 0.6 ( ) ( ) 1.5 ( ) ( ) 1.6 ( )! ( ) 0.8 ( ) ( ) 1.6 ( ) ( ) 1.1 ( ) Crack cocaine or powder cocaine use within last 30 days No Yes ( ) 0.8 ( ) ( ) 1.3 ( ) ( ) 2.1 ( ) Methamphetamine use within last 30 days No Yes ( ) 0.6 ( ) ( ) 1.9 ( ) ( ) 1.4 ( ) Case management None or rare f Moderate g ( ) ( ) ( ) Consistent h ( ) ( ) ( ) NOTE. Values in bold type are statistically significant. MCS, Mental Health Component Scale; PCS, Physical Component Scale; SF-36, Short Form with 36 Questions. a c Defined as an average of at least 1 scheduled, routine primary care visit every quarter of study enrollment; includes 72.5% of subjects. b One or more emergency department visits during the study period; includes 41.0% of subjects. One or more inpatient hospitalizations during the study period; includes 23.1% of subjects. d Defined as at least 90% of nights spent in a single-room occupancy hotel or apartment and no nights spent on the street or in a shelter in the past quarter. e At least 1 night spent on the street or in a shelter in the past quarter. Had contact with a case manager in 25% of enrolled quarters. g Had contact with a case manager in 125% but 75% of enrolled quarters. h Had contact with a case manager in 175% of enrolled quarters. f

6 Table 4. Factors associated in a linear regression model with increased treatment-indicated adherence among 219 homeless and marginally housed persons with CD4 + cell count nadir!350 cells/ml. Adherence Unadjusted b Adjusted b Sex, male vs. female 0.03 ( 0.11 to 0.16) Race/ethnicity, black vs. nonblack 0.06 ( 0.17 to 0.05) Age at baseline, 40 years vs.!40 years 0.16 (0.05 to 0.27) 0.15 (0.04 to 0.26) Homeless a vs. marginally housed b 0.13 ( 0.24 to 0.01) 0.04 ( 0.16 to 0.08) CD4 + cell count nadir,!200 cells/ml vs. 200 cells/ml 0.07 ( 0.04 to 0.18) 0.04 ( 0.07 to 0.15) SF-36 PCS score!35 vs ( 0.08 to 0.19) 35 but!50 vs ( 0.12 to 0.15) SF-36 MCS score!35 vs ( 0.35 to 0.10) 0.19 ( 0.32 to 0.06) 35 but!50 vs ( 0.26 to 0.01) 0.10 ( 0.23 to 0.02) Crack cocaine or powder cocaine use within last 30 days, yes vs. no 0.10 ( 0.22 to 0.03) 0.08 ( 0.20 to 0.05) Methamphetamine use within last 30 days, yes vs. no 0.04 ( 0.17 to 0.09) Case management Moderate c vs. rare or none d 0.11 ( 0.02 to 0.23) 0.13 (0.02 to 0.25) Consistent e vs. rare or none d 0.12 ( 0.02 to 0.26) 0.13 ( 0.01 to 0.26) NOTE. Values in bold type are statistically significant. a At least 1 night spent on the street or in a shelter in the past quarter. b At least 90% of nights spent in a single-room occupancy hotel or apartment and none spent on the street or in a shelter in the past quarter. c Had contact with a case manager in 125% but 75% of enrolled quarters. d Had contact with a case manager in 25% of enrolled quarters. e Had contact with a case manager in 175% of enrolled quarters. 55 (25.1%) had geometric mean viral loads of 400 copies/ ml, and 27 (12.3%) had a 50% increase in CD4 + cell count. Multivariate Outcomes CM was not independently associated with primary care use, emergency department use, or hospitalization (table 3). Treatment-Indicated Adherence to HIV ART In a multivariate linear regression model, moderate CM was associated with improved adherence ( b p 0.13; 95% CI, ), compared with no or rare CM. Consistent CM use neared but did not reach a statistically significant association (b p 0.13; 95% CI, 0.01 to 0.26) (table 4). Biological Markers Improvement in CD4 + cell count. In a multivariate model examining factors associated with a 50% increase in CD4 + cell count, we found that, compared with no or rare CM, both moderate CM and consistent CM were strongly associated with improvements in CD4 + cell counts (table 5). This relationship was true with and without adherence in the model. Removing adherence from the model slightly increased both adjusted ORs (AORs). Results were similar with a CD4 + cell count increase of 1100 cells/ml (data not show). Geometric viral load 400 copies/ml. In a multivariate model examining patients having a geometric mean viral load 400 copies/ml, the model was dominated by adherence (AOR, 16.2; 95% CI, ). When adherence was in the model, there was no significant reduction in the outcome of having consistent CM (AOR, 1.6; 95% CI, ) or moderate CM (AOR, 1.1; 95% CI, ), compared with no or rare CM. When we removed adherence from the model, there was an elevated (but not statistically significant) association between consistent CM and viral load 400 (AOR, 2.0; 95% CI, ). Propensity analysis. Adjusting factors using a propensity analysis did not change any of our results (data not shown). DISCUSSION In this study examining a cohort of HIV-infected urban poor, we found that having CM was independently associated with improved adherence to ART and improved CD4 + cell count. Having consistent CM neared but did not reach a statistically significant relationship with improved adherence to ART; it HIV/AIDS CID 2006:43 (15 July) 239

7 Table 5. Factors associated in logistic regression models with desirable biological outcomes among homeless and marginally housed HIV-infected persons with CD4 + cell count nadir!350 cells/ml (n p 219). Subjects with 50% increase in CD4 + cell count a Subjects with geometric viral load!400 copies/ml b Rate, % Unadjusted OR Adjusted OR with adherence Adjusted OR without adherence Rate, % Unadjusted OR Adjusted OR with adherence Adjusted OR without adherence Sex Female Male ( ) ( ) Race/ethnicity Nonblack Black ( ) ( ) Age at baseline!40 years years ( ) ( ) 1.0 ( ) 1.6 ( ) Housing status for the last quarter Marginally housed c Homeless d ( ) ( ) 0.7 ( ) 0.7 ( ) CD4 + cell count nadir 200 cells/ml !200 cells/ml ( ) 3.0 ( ) 3.0 ( ) ( ) SF-36 PCS score but! ( ) ( )! ( ) ( ) SF-36 MCS score to! ( ) 1.7 ( ) 1.5 ( ) ( ) 0.6 ( ) 0.5 ( )! ( ) 0.5 ( ) 0.5 ( ) ( ) 0.7 ( ) 0.4 ( ) Crack cocaine or powder cocaine use No Yes ( ) ( ) 0.7 ( ) 0.6 ( ) Methamphetamine use No Yes ( ) 0.3 ( ) 0.5 ( ) Adherence per unit 2.6 ( ) 2.2 ( ) 19.7 ( ) 16.2 ( ) Case management None or rare e Moderate f ( ) 6.5 ( ) 7.3 ( ) ( ) 1.1 ( ) 1.4 ( ) Consistent g ( ) 10.7 ( ) 11.9 ( ) ( ) 1.6 ( ) 2.0 ( ) NOTE. Values in bold type are statistically significant. MCS, Mental Health Component Scale; PCS, Physical Component Scale; SF-36, Short Form with 36 Questions. a Includes 27 (12.3%) of the subjects. b Includes 55 (25.1%) of the subjects. c At least 90% of nights spent in a single-room occupancy hotel or apartment and none spent on the street or in a shelter in the past quarter. d At least 1 night spent on the street or in a shelter in the past quarter. e Had contact with a case manager in 25% of enrolled quarters. f Had contact with a case manager in 125% but 75% of enrolled quarters. g Had contact with a case manager in 175% of enrolled quarters. was strongly associated with improvements in CD4 + cell count and neared but did not reach a reduction in geometric viral load. CM was not associated with changes in health services use; it was associated neither with an increased rate of receipt of primary care nor with reductions in emergency department use or hospitalizations. Although CM is widely used in HIV care and is believed to improve health outcomes, there are few studies looking at health outcomes with HIV CM in disenfranchised populations. Our finding that any CM was associated with improved ART adherence and improved biologic outcomes extends prior research on the effect of CM on patients with HIV infection. The effects of CM on health outcomes have not been previously documented [4]. Our findings suggest that CM may play a role in improving health outcomes among low-income persons with HIV infection. Without a randomized trial, we cannot state that there was a causal association between CM and improved outcomes. There are several means by which CM could improve biologic outcomes. Improved biologic outcomes could be explained by the increased appropriate use of and adherence to ART [24]. Of interest, increased adherence did not fully explain the as- 240 CID 2006:43 (15 July) HIV/AIDS

8 sociation between CM and improvements in CD4 + cell count. This may be because CM may have sustained CD4 + cell counts through other means, such as better use of opportunistic infection prophylaxis or other health-related behaviors. One of the primary effects of CM may be in its ability to assist clients in organization, enabling the consistent use of complicated medical regimens, including ART. It is likely that many case managers encourage their clients to adhere to medication, assist in communicating side effects to primary care providers, and check in with clients regularly to make sure medications are prescribed and refilled. Our findings that CM was not associated with primary care may be a result of high attendance at primary care facilities, allowing little room for improvement from CM. We did not find an association with emergency department use or hospitalizations, suggesting that CM neither reduced visits nor increased them (via increased surveillance). We found that rates of CM were similar to those reported for participants in Ryan White CARE programs nationally [25]. A relatively large proportion of participants used 11 case manager and 11 agency. This suggests some duplication of services. These may be explained by the different roles played by different case managers, or it may point towards a lack of coordination within the system. Our study had a number of limitations. We did not determine CM by random enrollment. It is possible that the same qualities that allowed participants to receive CM were the reason for their improved medication adherence or improved biological markers. It is also possible that those who were deemed to have greater need of CM were enrolled more aggressively. Including a propensity analysis in our multivariate model did not change our results. The CM models studied were heterogeneous: some included brokerage models and other models in which the case managers themselves provided mental health counseling or nursing CM. This heterogeneity would bias our findings towards the null hypothesis. We relied on several self-reported measures. These may be underreported or reported inaccurately. We may not have had sufficient power to note small effects, particularly when we examined the subpopulation that would benefit from the use of ART to measure ART use and adherence. We limited subjects with eligible HAART use to subjects with CD4 + cell counts!350 cells/ml. This is a conservative classification of HAART eligibility, because a common standard was to treat at CD4 + cell counts of!500 cells/ml [19, 20]. The study did not explore whether different models of CM would be more effective nor did it compare CM to other potential interventions, such as improving access to HIV testing or primary care. Finally, our study was conducted in 1 city and therefore may not be generalizable to other locations. Our study found an association between the use of CM services and improved adherence to ART and improved biological markers of HIV disease among homeless and marginally housed HIV-infected populations. This suggests that CM may be an effective way to improve health outcomes among disenfranchised HIV-infected populations. Acknowledgments Financial support. National Institute of Mental Health (RO ) and The University-Wide AIDS Research Program. M.B.K. received financial support from the Agency for Health Care Research and Quality (K- 08 HS 11415) and the Hellman Family Award for Junior Faculty. D.R.B. received support from The Doris Duke Clinical Scientist Development Program. Potential conflicts of interest. All authors: no conflicts. References 1. Laramee AS, Levinsky SK, Sargent J, Ross R, Callas P. Case management in a heterogeneous congestive heart failure population: a randomized controlled trial. Arch Intern Med 2003; 163: Ferguson JA, Weinberger M. Case management programs in primary care. J Gen Intern Med 1998; 13: Rich MW, Beckham V, Wittenberg C, Leven CL, Freedland KE, Carney RM. A multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure. N Engl J Med 1995; 333: Katz MH, Cunningham WE, Fleishman JA, et al. Effect of case management on unmet needs and utilization of medical care and medications among HIV-infected persons. Ann Intern Med 2001; 135: Tulsky JP, Pilote L, Hahn JA, et al. Adherence to isoniazid prophylaxis in the homeless: a randomized controlled trial. Arch Intern Med 2000; 160: Kushel MB, Vittinghoff E, Haas JS. Factors associated with the health care utilization of homeless persons. JAMA 2001; 285: Linn LS, Gelberg L, Leake B. Substance abuse and mental health status of homeless and domiciled low- income users of a medical clinic. Hosp Community Psychiatry 1990; 41: Martens WH. A review of physical and mental health in homeless persons. Public Health Rev 2001; 29: Celentano DD, Vlahov D, Cohn S, Shadle VM, Obasanjo O, Moore RD. Self-reported antiretroviral therapy in injection drug users. JAMA 1998; 280: Strathdee SA, Palepu A, Cornelisse PG, et al. Barriers to use of free antiretroviral therapy in injection drug users. JAMA 1998; 280: Kushel MB, Perry S, Bangsberg D, Clark R, Moss AR. Emergency department use among the homeless and marginally housed: results from a community-based study. Am J Public Health 2002; 92: Mocroft A, Brettle R, Kirk O, et al. Changes in cause of death across Europe [abstract 298]. In: Program and abstracts of the 8th Conference on Retroviruses and Opportunistic Infections (Chicago, IL) Katz MH, Cunningham WE, Mor V, et al. Prevalence and predictors of unmet need for supportive services among HIV-infected persons: impact of case management. Med Care 2000; 38: Gardner LI, Metsch LR, Anderson-Mahoney P, et al. Efficacy of a brief case management intervention to link recently diagnosed HIV-infected persons to care. AIDS 2005; 19: Burnam M, Koegel P. Methodology for obtaining a representative sample of homeless persons: the Los Angeles Skid Row Study. Eval Rev 1988; 12: Moss AR, Hahn JA, Perry S, et al. Adherence to highly active antiretroviral therapy in the homeless population in San Francisco: a prospective study. Clin Infect Dis 2004; 39: Riley ED, Bangsberg DR, Perry S, Clark RA, Moss AR, Wu AW. Re- HIV/AIDS CID 2006:43 (15 July) 241

9 liability and validity of the SF-36 in HIV-infected homeless and marginally housed individuals. Qual Life Res 2003; 12: Ware JE, Kosinski M, Keller SD. SF-36 physical and mental health summary scales: a user s manual. Boston: Health Assessment Lab, Yeni PG, Hammer SM, Carpenter CC, et al. Antiretroviral treatment for adult HIV infection in 2002: updated recommendations of the International AIDS Society-USA Panel. JAMA 2002; 288: US Public Health Service. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. Atlanta, GA: Centers for Disease Control and Prevention, Chesney MA, Ickovics JR, Chambers DB, et al. Self-reported adherence to antiretroviral medications among participants in HIV clinical trials: the AACTG adherence instruments. Patient Care Committee; Adherence Working Group of the Outcomes Committee of the Adult AIDS Clinical Trials Group (AACTG). AIDS Care 2000; 12: Heckman JJ HV. Choosing among alternative nonexperimental methods for estimating the impact of social programs. J Am Stat Assoc 1989; 84: Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: Mannheimer S, Friedland G, Matts J, Child C, Chesney M. The consistency of adherence to antiretroviral therapy predicts biologic outcomes for human immunodeficiency virus infected persons in clinical trials. Clin Infect Dis 2002; 34: McKinney MM, Marconi KM. Delivering HIV services to vulnerable populations: a review of CARE Act-funded research. Public Health Rep 2002; 117: CID 2006:43 (15 July) HIV/AIDS