Family support predicts psychiatric medication usage among Mexican American individuals with schizophrenia

Transcription

1 Soc Psychiatry Psychiatr Epidemiol (2006) 41: DOI /s ORIGINAL PAPER Jorge I. Ramírez García Æ Christina L. Chang Æ Joshua S. Young Æ Steven R. López Æ Janis H. Jenkins Family support predicts psychiatric medication usage among Mexican American individuals with schizophrenia Accepted: 24 March 2006 / Published online: 29 May 2006 SPPE 69 j Abstract Background Support provided by family caregivers to persons with schizophrenia is a viable intervention focus to improve psychiatric medication usage. However, little is known about the relation between medication usage and family support as well as other key caregiving factors. Method Family support and Expressed Emotion (EE) dimensions were tested as predictors of medication usage during a 9-month period following psychiatric hospital discharge in a sample of 30 individuals of Mexican descent with schizophrenia. Results Family instrumental support predicted higher medication usage (Odds Ratio = 4.8) in multivariate analyses that statistically adjusted for the impact of emotional support, family EE, and psychiatric status (e.g., positive symptoms) on medication usage. Conclusions Findings suggest that efforts to improve medication usage among Mexican American individuals with schizophrenia should take into account social supportive factors such as instrumental or directive, hands-on assistance from family caregivers. j Key words medication usage or medication compliance schizophrenia family support expressed emotion Mexican Americans J.I. Ramírez García (&) Æ J.S. Young Dept. of Psychology University of Illinois, Urbana-Champaign 603. E. Daniel Street Champaign (IL) 61820, USA Tel.: / jramirez@uiuc.edu C.L. Chang Æ S.R. López Psychology Dept. UCLA Los Angeles (CA), USA J.H. Jenkins Dept. of Anthropology University of California-San Diego La Jolla (CA) USA Introduction Poor psychiatric medication usage among the seriously mentally ill has long been known as one of the top risk factors for psychiatric rehospitalization [1]. However, the role that social support may play in medication usage is poorly understood. This major gap in the literature is critical for at least two reasons. First, consistent adherence to psychiatric medication regimens among individuals with serious mental illness such as schizophrenia is a major challenge, with around 40% of these individuals not taking their medications regularly [2, 3]. In fact, a recent review estimates that this figure jumps to 75% when a 2-year window is used to assess medication usage [4]. Clearly, there is plenty of room for improvement. Second, some variables that have been consistently shown to predict medication usage have minimal or low utility for intervention purposes. For example, a short duration since illness onset is one of the best predictors [5], but this predictor is historical or static [6]. That is, it is not amenable to modification by practitioners. Hence, it becomes crucial to study and understand medication usage predictors that are viable foci for intervention. In this paper we examined family support as a predictor of psychiatric medication usage. Family caregivers are a great resource for potential interventions to increase medication usage and prevent unfavorable course of mental illness. For example, family-based interventions have been empirically shown to significantly impact the course of mental illness [7, 8]. Moreover, family interventions cover educational components of medication usage and utilize family caregiver problem solving sessions to target relapse prevention. Hence, improved understanding of what family caregiving factors are linked with medication usage can inform these family intervention programs.

2 j Family factors and medication usage Reviews of the literature on predictors of psychiatric medication usage and family variables have yielded scarce and inconclusive findings. Lacro et al. [5] concluded that family involvement yielded inconsistent prediction results. However, this trend was based on a handful of studies and on global indices of family involvement and living arrangements. On the other hand, in a prior review, Fenton et al. [2] concluded that social support, in general, and the availability of family or friends to assist or supervise medications, in particular, are consistently associated with outpatient adherence (p. 642). Taken together these findings suggest the need to not only examine the presence of family members, but also to examine the role of specific family supportive dimensions that may predict psychiatric medication usage. Although family caregivers can sometimes be a valuable source of support, they can also be a source of conflict. The literature on family Expressed Emotion (EE) illustrates the need to take into account the conflicting nature of close relationships rather than the mere presence of supportive relationships. High EE consists of family caregiver negative affect and/or emotional overinvolvement (EOI). Dozens of studies have replicated the finding that individuals with schizophrenia, who live with family caregivers classified as high in EE, have about threefold increased risk of relapse when compared to their low EE counterparts [9, 10]. These findings highlight the need to examine supportive aspects of relationships jointly with possible conflicting aspects. Moreover, because in the typical EE study psychiatric medication usage is a covariate in the relation between EE and relapse, we know little about the relation between EE and medication usage. j Latino ethnicity Ethnocultural context plays an important role in the relationships between family caregivers and their relatives with mental illness [11, 12]. One illustrative finding of this assertion is that Latino family caregivers are more likely to live with their mentally ill relatives [13, 14] and have a higher degree of contact [15] compared to family caregivers with a European American ethnic background. This major cross-ethnic difference can be interpreted in light of the purported notion that Latinos place family relationships at the center of their lives such cultural value is often referred as familismo or familism [16]. However, the purported high levels of familismo among Latinos shed little light to understanding which specific types of family supportive dimensions may be linked to the psychiatric medication usage of their relatives with serious mental illness. Moreover, Latino familism per se does not address how medication usage might be linked to conflicting family support, such as caregiver negative affect toward and overinvolvement with their ill relatives. The poorly developed theoretical background regarding the relation between family caregiving and medication usage among mentally ill Latinos is coupled with substantial paucity in the empirical work on this topic. A couple of studies indicate that medication usage also predicts psychiatric relapse and rehospitalization among Mexican origin Latinos [14, 15]. However, we found only one published study that examined family predictors of medication usage among Latinos with schizophrenia. Hosch et al. [17] found that approximately 50% of the patients in their outpatient sample did not regularly take medications during a continuous 12-month period. These figures are fairly close to those found in two major reviews [2, 5] suggesting that Latino ethnicity is not tied to either lower or higher usage of psychiatric medications. Moreover, Hosch et al. found that those patients who were receiving financial support from relatives were more likely to take their medications as prescribed compared to those who were self-supporting or received federal assistance exclusively. This latter finding also points toward the importance of family support in terms of financial assistance in medication usage among Latinos. j Study overview 625 The major focus of this paper was to test whether family support predicts medication usage among Mexican American individuals with schizophrenia. We tested two specific types of support that have been found in factor-analytic studies [18] instrumental and emotional. Second, we also tested whether family support would predict medication usage over and beyond two key predictor domains. The first domain was patient s psychiatric status such as levels of positive symptoms. Psychiatric status has been found to predict medication usage [5] and thus may confound the link between family support and medication usage, especially if psychiatric status is associated with family support. Family EE was the second predictor domain. This was crucial to ascertain whether supportive family behaviors might impact medication usage beyond the plausible impact of conflicting family relationship elements that have been consistently found to predict relapse (i.e., EE). Moreover, because positive affect (warmth) was found to predict lower likelihood of relapse in this Mexican American sample [15], we also explored whether warmth would mediate the plausible impact of support on medication usage. The study design was prospective family and psychiatric status predictors were assessed near psychiatric hospital discharge and medication usage was judged for a 9-month time frame following discharge.

3 626 Method j Research participants Mexican American dyads (N = 30) of individuals with schizophrenia and a primary family caregiver were included in this study. They were drawn from the Course of Schizophrenia among Mexican Americans (COSAMA) project (see [19] for sample details). They were recruited from three psychiatric facilities in Southern California shortly following admission. Patients met diagnostic criteria for schizophrenia based on the Present State Examination (PSE [20]), were of Mexican descent, were living with family within the 3 months prior to admission, and were years old. The family caregiver was the person with the most frequent face-to-face contact with the patient. In a few instances, two family members (e.g., two parents) were interviewed separately. One relative was chosen randomly, except when both parents were high in EE and one of them was designated as high in EOI. The latter procedure was conducted in order to include as many high EOI parents as possible in the sample given that the base rate for high EOI is low. The final group of 30 dyads in this study met three additional criteria: (a) patients did not show sustained elevated levels of psychotic symptoms during 9 months following discharge, (b) data on medication usage during the 9 months following discharge were available, and (c) an existing audible Camberwell Family Interview (CFI) audiotape with the family caregiver was available. Individuals with schizophrenia were 58% male and predominantly single (77%). Their mean age was 25.7 (SD = 7.6). Patients had an average of 3.6 (SD = 3.0) prior hospitalizations. They were predominantly poor 91% were classified in the 4th and 5th categories of the Hollingshead index. Most patients and their primary caregivers were born in México (52% and 68%, respectively). Over two-thirds (68%) of the family caregivers and over one-third of (39%) of the patients were monolingual Spanish speakers. The majority of the caregivers were mothers (63%), four were fathers, four were spouses, and three were other relatives. j Measures Family caregiver measures Following patients psychiatric hospital admission and diagnosis confirmation, key relatives who had frequent face-to-face contact with patients were interviewed by trained bilingual researchers (see [19] for detailed procedures). The CFI [21], a personal and semistructured interview that yields rich family caregiver narratives of their relationship with their ill relative, was used to assess family caregivers supportive behaviors and EE. The interview inquires about attitudes toward and experiences with the patient s illness and the influence that it has on the life of the family during the 3 months prior to patient hospitalization. The CFI was translated from English to Spanish for those Mexican Americans who spoke primarily Spanish. Expressed emotion. The audio taped material of the min CFI was used by bilingual coders to code EE scales. The CFI Spanish version was the product of back-translation techniques that included both professional Spanish translators and Spanish-speaking clinicians experienced in working with Mexican American families, as well as pilot testing with 22 Spanish-speaking family caregivers. High EE is determined by high levels of negative affect (criticism) and/or EOI. Criticism is the number of critical comments made by the family caregiver during the entire interview. Critical comments are statements that communicate dislike, disapproval, and/or resentment of the ill relative s behavior or characteristic. EOI is a Likert scale ranging from 0 to 5 that primarily assesses family caregivers : (a) overprotective and or self-sacrificing behavior, and (b) exaggerated and or emotional narratives of their ill relatives behavior. Coders of the EE subscales were trained by two highly experienced raters who were the main authors of the first EE study in the US [22]. EE raters of the present study became reliable compared to the trainers on the primary scales used to classify high versus low EE (i.e., criticisms and EOI, r = 0.85) and were periodically compared to them throughout the study in order to minimize rater drift (See [19] for more details regarding CFI translation procedures and EE scoring). Warmth. Levels of warmth were coded from the CFI interviews. It is a Likert scale ranging from 0 to 5 that assesses family caregivers empathy, concern, and interest in their relatives who are mentally ill. Although interrater reliability estimates for warmth ratings were not obtained in the first EE replication study with Mexican Americans because the primary focus at that time was on high versus low EE based on criticism, hostility and EOI [19], warmth raters were trained by the same trainers and with the same procedures used to train for rating criticisms and EOI. Moreover, the CFI manual includes specific definitions and examples to anchor each point in the Likert warmth scale which have been successfully used by raters of a different Mexican American sample who have achieved acceptable to high levels of interrater reliability with the CFI [23]. Family support. A separate team of bilingual coders who were blind to patients EE ratings were trained to code the family support variables based on the entire audio taped CFI material. Bilingual coders also underwent extensive training prior to coding the family support scales (see [24] for more details regarding support coding). There are two family support variables in the present study. Instrumental Support was operationalized as the total number of statements that illustrated family caregiver task-oriented assistance such as completion of errands (e.g., I helped him fill out an employment application ). Emotional Support was operationalized as the total number of statements that illustrated family caregiver demonstrations of reassurance, concern, and affection (e.g., I told her that I love her. ). Prior to coding the support variables, all CFI interviews were transcribed. Coders listened to the interviews while simultaneously observing the transcripts on a computer screen. They underwent training prior to the initiation of coding for the study. During the formal stage of coding, the trainer (CL Chang) randomly chose 16 tapes that were coded independently by her and each of the other three coders. Interrater reliability was measured using the intraclass correlation coefficient (ICC), treating raters as random effects and the individual rater as the unit of reliability [25]. The ICC for instrumental support was 0.95 and for emotional support it was Patient measures Medication usage. The 9-month period following psychiatric hospital discharge was the time frame to assess medication usage. Its operational definition followed the two landmark EE studies [21, 22]. Namely, those patients who took their medications at least 75% of the time with no 4-week or longer interval of discontinued use of medications were classified as Regular medication users, while the rest were considered to be Irregular medication users. This judgment was made by trained clinical psychologists based on an examination of patients charts at the outpatient clinics where they were served, monthly telephone contacts with their family caregivers, and psychiatric assessment conducted in-person at the 9-month mark following hospital discharge. In the larger study that is the source of our current sample [19], only one patient could not be confidently rated on medication usage using this multiple-source method. Psychiatric status predictors. Levels of Positive Symptoms were assessed within 2 weeks of hospital admission with the Psychiatric Assessment Scale (PAS; [26]). The PAS assesses levels of symptoms with a Likert scale ranging from 0 = Absent to 4 = Severe. It

4 627 Table 1 Bivariate correlations and descriptive statistics Med. use Pos. symp. Yrs. Ill Drug abs. Crit. EOI Warmth Emot. support Inst. support Medication usage Positive symptoms 0.28 Years Ill Drug abstinence )0.09 ) Criticisms ) )0.01 EOI )0.22 ) ) Warmth 0.29 ) )0.43** )0.28 Emotional sup. ) ) Instrumental sup. 0.38* ** 0.12 ) M (SD) or % 43% 2.02 (.94) 4.61 (3.61) 43% 3.94 (3.28) 2.26 (1.06) 3.16 (1.37) 0.74 (1.55) 8.19 (5.39) Observed range Note: Correlations involving dichotomous variables (i.e., medication use and drug abstinence) are Spearman coefficients; all others are Pearson coefficients; N = 30 was administered in person by trained bilingual clinical psychologists. Additionally, substance use and years since illness onset were included in the analyses because they have been found to be empirically linked to medication usage in prior studies [2, 5]. Substance use was measured as either absent versus present during the same 9-month follow-up period and with the same sources of data that were used to assess psychiatric medication usage. Barbiturates, cocaine, and PCP (angel dust) were used to define street drugs. In order to be consistent with prior EE studies, marijuana was not taken into account to classify a patient as a street drug user. Psychiatric assessments conducted close to hospital admission provided the source of data used to assess illness onset and years since illness onset. Results j Descriptive statistics and bivariate associations Of the 30 Mexican American individuals with schizophrenia, 13 (43%) were classified as regular medication users and 17 (57%) as irregular. Table 1 shows a correlation matrix between medication usage, psychiatric status, EE, and family support variables. Spearman r was used to estimate correlations involving dichotomous variables (medication usage and substance use) and Pearson r was used to calculate all other associations that involved continuous variables. Descriptive statistics for each variable are shown in the bottom row of the table including means, standard deviations, and ranges for continuous variables and frequencies (percentages) for dichotomous variables. Higher levels of instrumental family support were associated with greater likelihood of medication usage. Although emotional support and instrumental support were moderately related to each other, emotional support was not significantly associated with medication usage. Neither criticisms nor EOI were significantly related to medication usage. None of the psychiatric status variables were significantly related to medication usage; the relation between drug use (abstinence) and medication usage was particularly low (Spearman r = )0.09). Although the relation between positive symptoms and medication usage did not reach statistical significance, higher symptom levels were associated with a higher likelihood of medication usage. Notably, this trend was in the opposite direction compared to other medication usage studies which suggest that the higher the symptoms, the lower likelihood of medication usage [5]. j Multivariate prediction of medication usage by family factors A logistic regression analysis was conducted to assess if family factors, including the independent dimensions of EE and family support variables, would predict usage of psychiatric medications (regular vs. irregular). A hierarchical analytic framework was used to assess if all family variables would predict over and beyond psychiatric status; thus the latter was entered as a first step with family variables entered in the following steps. We also tested whether family support variables would predict medication usage over and beyond criticisms and EOI (EE variables); thus the latter were entered on the second step and the former in the final and third step. Note that in the forthcoming results street drug use was excluded from the models for two reasons. First, its relation with medication usage was small thus it is an unlikely mediator of the relation between family variables and medication usage [27]. Second, the introduction of substance use into the models produced multicollinearity and unstable results, that is, overly inflated regression coefficients and confidence intervals. Table 2 Hierarchical multiple regression model results of medication usage predictors STEP v 2 (df) MODEL v 2 (df) Psychiatric status (Positive 6.6 (2)* 6.6 (2)* symptoms and years ill) EE (Criticisms and EOI) 1.3 (2) 8.0 (4) Support (Instrumental and emotional) 7.1 (2)* 15.0 (6)* Note: N =30 *p < 0.05

5 628 Table 3 Effect sizes (Odds Ratios) of final logistic regression model Predictors Odds ratios 95% CI Positive symptoms , 13.0 Years Ill , 8.4 Criticisms , 7.1 Emotional overinvolvement (EOI) 0.1 a 0.1, 1.1 Instrumental support 4.8* 1.1, 21.7 Emotional support 0.1 a 0.0, 1.5 Note: In order to facilitate interpretation, predictors were standardized via Z score conversion; thus Odds Ratios reflect a one-standard deviation unit change in each predictor; N = 30 a These variables predicted lower medication use *p = 0.05 A summary of the results of the hierarchical analysis is shown in Table 2. To assess the statistical significance of each step and model, v 2 results are shown. The psychiatric status (positive symptoms and years ill) step significantly predicted medication usage. The step with criticisms and EOI (EE indices) did not predict medication usage over and beyond psychiatric status; the model with psychiatric status and EE indices did not predict medication usage significantly. However, family support variables predicted medication usage significantly over and beyond the impact of psychiatric status and EE indices as shown in the results of the final step in the hierarchical analysis. The model with psychiatric status, EE indices and family support predicted medication usage significantly. Predictor results in the final model are shown in Table 3. Because all predictors had different scaling and observed ranges, we standardized the variables via Z score conversion to facilitate interpretation of comparisons across predictors [28, 29]. Therefore, the Odds Ratios reported reflect the change in the odds of belonging to the regular medication user group (as opposed to belonging to the irregular medication user group) associated with a one-standard deviation unit change in each continuous predictor variable. Only family instrumental support predicted medication usage significantly (Odds Ratio = 4.8). Emotional support, criticisms and EOI were not significant predictors of medication usage. Note that as an added precaution, the same analyses reported above were conducted with the inclusion of substance abuse in all steps of the hierarchical analyses. The results were equivalent to those reported above and in Tables 2 and 3. That is, EE indices did not predict medication usage significantly as a block or independently, family support variables predicted above and beyond the impact of psychiatric status and EE indices: v 2 (df = 2) = 8.9, p=0.01, and instrumental support remained a significant predictor of medication usage (p = 0.03). These results including substance abuse in the models are available from the first author upon request. We also explored whether warmth might mediate the relation between instrumental support and medication usage. The Pearson correlation between warmth and medication usage was (r = 0.29, p = 0.12) and for warmth and instrumental support (r = 0.32, p = 0.07). Because these correlations are not trivial there is a reasonable basis to test for mediation [27]. Thus, following Cohen et al. [28] we compared the relation between instrumental support and medication usage without and with warmth partialled out. Regarding the former, instrumental support accounted for 17% of the variance of medication usage (i.e., r 2 = 0.17, p < 0.05). Regarding the latter, instrumental support accounted for 15% of the variance of medication usage while partialling out warmth (i.e., sr 2 = 0.15, p < 0.05). Thus, warmth only accounted for 2% of 17% of the prediction variance of instrumental support. These data suggest that warmth is an unlikely mediator. Discussion The major finding of this study was that higher family instrumental support predicted regular medication usage among Mexican Americans with schizophrenia. Despite the fact that our sample was not large, this finding was robust. We found that a one-standard deviation increase in instrumental support assessed near hospital admission was linked with an increase of odds nearly by five in belonging to the regular medication user group (rather than the irregular medication group) during the 9 months following discharge. Moreover, the statistical effect of instrumental support was significant beyond the impact of patients psychiatric status and family EE on medication usage. In interpreting the relation between social support and medication usage it is crucial to consider the role of patient psychiatric status. For example, it is plausible that those persons with schizophrenia who have a more stable clinical picture (with absent or mild symptoms) are more likely to take both their medications and effectively elicit social support. This plausible process is an unlikely interpretation to our findings. Our multivariate results showed that support variables yielded additional statistically significant predictive variance beyond that accounted for by psychiatric status (positive symptoms, substance use, and years of illness). Another key issue to consider is that higher medication usage may elicit supportive behaviors from caregivers. In turn, family caregiver supportive behaviors may reinforce medication usage, setting a therapeutic circular chain of events involving medication usage and supportive transactions with loved ones [2]. Notably, in the present prospective study instrumental support was assessed with a time frame of 3 months prior to the psychiatric hospitalization while medication usage was assessed with the 9- month hospital discharge time frame. Hence, it is not plausible that medication usage led to higher supportive behaviors. Future studies that include several assessments of both supportive behaviors and medi-

6 629 cation usage spread across time will shed further light into the possibility of bidirectional influence between these variables. In this study instrumental support was operationalized as task-oriented (hands on) assistance by family caregivers such as help in completing errands. Thus, the study findings suggest that support of a pragmatic nature (i.e., hands on ) from family caregivers is helpful in facilitating patients medication usage. The link between pragmatic or tangible assistance and medication usage was also observed by Hosch et al. [17] who found that Mexican Americans with schizophrenia who received financial support from family caregivers were more likely to take their medication than those who were self-supporting or received federal assistance but not family financial assistance. Taking together the present findings with those of Hosch et al., we hypothesize that family caregivers who are high on this type of task-oriented support are more likely to lend direct assistance in those practical and necessary intermediary steps that lead to medication usage such as obtaining, storing, retrieving, and tracking medication usage. One important finding of this study was that the impact of family support behaviors was beyond the possible conflicting emotional climate of the relationship between family caregivers and their ill relatives (i.e., EE). This finding is suggestive of family caregiving processes that may facilitate medication usage despite the emotional caregiving climate. If this finding were sustained in additional studies, it would imply that it is critical to target family supportive behaviors that increase medication usage rather than exclusively target family EE. However, because of our limited sample size and paucity of research on this topic further systematic analysis of the relation between medication usage and family support vis-à-vis EE is vital. We only found one published study that examined the role of EE on medication usage [30], but it did not examine family support. Furthermore, it is crucial that research on family support, EE and medication be conducted with ethnoculturally diverse samples given key cross-ethnic findings regarding family factors and mental illness. Although high versus low EE has also been found to predict relapse among Mexican Americans [19, 23], in one study with a sample that overlaps with our present sample, warmth was found to be a significant predictor of a lower likelihood of relapse among Mexican Americans but not among European Americans [15]. Thus, warmth as measured with the CFI may play an important role in the outcomes of Mexican Americans with schizophrenia. Given the link between warmth and lower relapse among Mexican Americans one might posit that warmth might be related to instrumental support as well and predict medication usage; and hence mediate the relation between instrumental support and medication usage. Indeed, the relation between warmth and instrumental support was not trivial in our study (i.e., r = 0.32). However, warmth only partialled 2% out of 17% of the medication usage variance predicted by instrumental support suggesting that warmth is an unlikely mediator. This finding warrants replication, in light of our sample size and the scarcity of research on this topic. In our study the relation between instrumental support and EOI was noticeably small (r = )0.02) suggesting almost no relation between the hands-on supportive behaviors tapped by the former and EOI. One might expect that EOI would tap into investment by caregivers in the lives of their relatives with schizophrenia and thus be related to caregiver instrumental support. However, when rating EOI, emphasis is placed on extreme forms of caregiver involvement; hence over-involvement as well as overprotective caregiving behaviors are major EOI criteria. Thus, it is plausible that caregivers can provide hands-on assistance without displaying the set of more extreme behaviors associated with high EOI. Although it was not the focus of the present study, we found that a trend for higher levels of positive symptoms being linked with an increased likelihood of taking medications regularly which is in the opposite direction of what is most commonly reported in the literature [5]. It could be that once the most acutely ill patients are removed from a sample, relatively higher levels of symptoms led to greater motivation and/or monitoring of medication usage by caregivers. Future studies should examine this possibility. Another noteworthy finding that was not the focus of the present study was the high relation between instrumental support and drug abstinence. It is plausible that the repertoire of hands-on assistance tapped by our measure of instrumental support is tied to the ability of caregivers to restrict their ill relative s access to street drugs and/or that it is more likely to be elicited by street drug-free persons with schizophrenia. These possibilities should be examined in future studies. Although medication usage was carefully assessed by trained psychologists using multiple sources of information, it was operationalized with a dichotomous procedure. Continuous variables that gauge the range of degree of medication usage might do more justice to represent the breadth of this construct as well as increase statistical power to detect notable predictors. Nonetheless, the rate of irregular medication usage was 57% which is reasonably close to the weighted mean non-adherence rate of 50% that was found by Lacro et al. [5] based on several studies that also used the cut off criterion that patients take medication regularly at least 75% of the time. Thus, our medication usage variable was not out of range compared to prior findings in the literature. Moreover, the 57% versus 43% dichotomous split in our study does not deviate largely from an even split which is desirable to maximize statistical power when using dichotomous variables as outcome criteria [28].

7 630 Additionally, because lower power is expected in dichotomous variables compared to continuous variables, our major findings (e.g., instrumental support) are more likely to be underestimates rather than overestimates. Nonetheless, replication of the present findings with continuous variables of medication usage would strengthen the present findings. The small sample size of this study warrants cautious interpretation of the findings. Null findings should be interpreted in light of the limited statistical power, especially those with notable effect sizes. On the other hand, statistically significant findings are particularly noteworthy because of the limited statistical power due to the small sample and dichotomous rating of medication usage. However, they also warrant empirical replication especially given the paucity of research on the role of family factors on medication usage. Given the high likelihood that Latino family caregivers live with their relatives with schizophrenia [13, 14], family support networks are a fertile ground for psychosocial interventions aimed at stabilizing the course of illness in this ethnic group. Furthermore, albeit limited data in the literature, Latinos with schizophrenia do not appear to be more or less likely to take their medications [17] when compared to estimates based on several studies [5]. Moreover, the extant data suggest that medication usage also predicts psychiatric relapse [15] and psychiatric rehospitalizations [14] among Latinos. In closing, there was substantial variability in the degree of instrumental support reported in our sample of Mexican American family caregivers. Although instrumental support was assessed via ratings of personal interviews with caregivers and future research should also include other indices of support (e.g., observational measures), the variability in family caregiver support found in this study highlights the need to assess family supportive functions even within sociocultural groups that are known to be highly involved in caregiving for individuals with schizophrenia. j Acknowledgements This research was supported by a National Institute of Mental Health NRSA Training Grant MH14584 (Psychological Research on Schizophrenic Conditions) to the first author. An earlier version of this paper was presented at the Family Research Consortium IV, San Juan, Puerto Rico, in July 2004 with support from NIMH to the first author. The paper is partially based on the Honors Thesis of the third author. The first author would like to thank Nicole Allen, Howard Berenbaum, Adriana Umaña- Taylor, and Edelyn Verona for their input on previous drafts. References 1. Ellison JM, Blum N, Barsky AJ (1986) Repeat visitors in the psychiatric emergency service: a critical review. Hosp Community Psych 37: Fenton WS, Byler C, Heinsesen R (1997) Determinants of medication compliance in schizophrenia: empirical and clinical findings. Schizophr Bull 23: Young JL, Zonana HV, Shelper L (1986) Medicatin noncompliance in schizophrenia: codification and update. Bull Am Acad Psychiatry Law 14: Marder SR (1998) Facilitating compliance with antipsychotic medication. J Clin Psychiatry 59: Lacro JP, Dunn LB, Dolder CR, Leckband SG, Jeste DV (2002) Prevalence of and risk for medication nonadherence in patients with schizophrenia: a comprehensive review of recent literature. J Clin Psychiatry 63: Singer JD, Willet JB (1991) Modeling the days of our lives: using survival analysis when designing and analyzing longitudinal studies of duration and the timing of events. Psychol Bull 110: Mari JJ, Streiner DL (1994) An overview of family interventions and relapse on schizophrenia: meta-analysis of research findings. Psychol Med 24: McFarlane WR, Link B, Dushay R, Marchal J, Crilly J (1995) Psychoeducational multiple family groups: four-year relapse outcome in schizophrenia. Fam Process 34: Bebbington P, Kuipers L (1994) The predictive utility of expressed emotion in schizophrenia: an aggregate analysis. Psychol Med 24: Butzlaff RL, Hooley JM (1998) Expressed emotion and psychiatric relapse. Arch Gen Psych 55: Jenkins J, Karno M (1992) The meaning of expressed emotion: theoretical issues raised by cross-cultural research. Am J Psychiatry 149: Lefley HP (1990) Culture and chronic mental illness. Hosp Community Psychiatry 41(3): Guarnaccia PJ, Parra P (1996) Ethnicity, social status, and families experiences of caring for a mentally ill family member. Community Ment Health J 32: Ramírez García JI, Wood JM, Hosch HM, Meyer LD (2004) Predicting psychiatric rehospitalizations: examining the role of Latino versus European American ethnicity. Psychol Serv 1: López SR, Hipke KN, Polo AJ, Jenkins J, Karno M, Vaughn C, Zinder KS (2004) Ethnicity, expressed emotion, attributions, and course of schizophrenia: family warmth matters. J Abnorm Psychol 16. Steidel AG, Contreras JM (2003) A new familism scale for use with Latino populations. Hisp J Behav Sci 25: Hosch HM, Meyer LD, Barrientos GA, Fierro C, Ramírez JI, Moreno A, Pelaez MP, Pérez Y (1995) Predicting adherence to medication by Hispanics with schizophrenia. Hisp J Behav Sci 17: Barrera M Jr (2000) Social support research in community psychology. In: Rappaport J, Seidman E (eds) Handbook of community psychology. Kluwer Academic/Plenum Publishers, New York, NY, pp Karno M, Jenkins JH, de la Selva A, Santana F, Telles C, López SR, Mintz J (1987) Expressed emotion and schizophrenic outcome among Mexican-American families. J Nerv Ment Dis 175: Wing JK, Cooper JE, Sartorious N (1974) The measurement and classification of psychiatric symptoms. Cambridge University Press, London 21. Vaughn CE, Leff J (1976) The measurement of expressed emotion in the families of psychiatric patients. Br J Psychiatry 129: Vaughn CE, Snyder KS, Jones S, et al. (1984) Family factors in schizophrenic relapse: Replication in California of British research on expressed emotion. Arch Gen Psychiatry 41: Kopelowicz A, López SR, Zarate R, O Brien M, Gordon J, Chang C, Gonzalez-Smith V (2006) Expressed emotion and family interactions in Mexican Americans with schizophrenia. J Nerv Ment Dis 194:

8 Chang CL, López SR, Murray AM, Jenkins JH, Karno M (2006) Prosocial processes in caring for relatives with schizophrenia: families affective reactions, social support and expressed emotion. (Manuscript submitted for publication) 25. Shrout P, Fleiss JL (1979) Intraclass correlations: uses in assessing rater reliability. Psychol Bull 83: Krawiecka M, Goldberg D, Vaughan M (1977) A standardized psychiatric assessment scale for rating chronic psychotic patients. Acta Psych Scand 55: Baron RB, Kenny DA (1986) The moderator mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Pers Soc Psychol 51: Cohen J, Cohen P, West SG, Aiken LS (2003) Applied multiple regression/correlation analyses for the behavioral sciences, (3rd edn). Lawrence Erlbaum Associates, New Jersey 29. Pampel FC (2000) Logistic regression: a primer. Sage, Thousand Oaks, CA 30. Sellwood W, Tarrier N, Quinn J, Barrowclough C (2003) The family and compliance in schizophrenia: the influence of clinical variables, relative s knowledge, and expressed emotion. Psychol Med 33:91 96