Moving for Care: Findings from the HIV Cost and Services Utilization Study

Andrew S. London, Syracuse University
Janet Wilmoth, Syracuse University

Introduction In this paper, we use baseline data from the HIV Cost and Services Utilization Study (HCSUS), a nationally representative sample of HIV-infected adults receiving regular medical care in early 1996, to examine three key research questions: (1) What factors are associated with internal migration prior to and since knowing about HIV infection (relative to never having moved)?; (2) Among those who moved since knowing about their HIV infection, what factors are associated with moving for care (either informal or formal) as opposed to moving for other reasons?; and (3) Among those who moved for care, what factors are associated with moving for formal or informal care respectively? Background There is an extensive literature describing the migration patterns of the elderly, and various models and approaches have been articulated to explain why the elderly move (Bradsher et al. 1992; De Jong and Fawcett 1981; De Jong et al. 1995; Findley 1988; Litwak and Longino 1987; Longino et al. 1991; Meyer and Speare 1985; Silverstein 1995; Silverstein and Angelelli 1998; Speare, Avery, and Lawton 1991). Longino (1990) argues that three general perspectives on elderly migration prevail in the literature: (1) taxonomies of moves; (2) delineation of person-environment adjustment factors that affect migration decision-making; and (3) identification of triggering mechanisms associated with migration-related life course events. Elaborating typologies of movers has been a major focus of the literature on elderly migration. Perhaps the most widely cited model used to describe elderly migration is the developmental model proposed by Litwak and Longino (1987). This model posits three kinds of moves that correspond to different life course stages: a first move for lifestyle and leisure-related reasons at retirement; a second move that occurs with increasing age, chronic illness, or disability, often after widow(er)hood; and a third move that occurs with increasing frailty and the transition from independent or assisted community-based living to institutional care. A second widely used approach, which emerges from the work of De Jong and Fawcett (1981) and Meyer and Speare (1985), seeks to identify unique types of moves as indicated by reason-for-move data. These two approaches are not mutually exclusive, and can fruitfully be combined, as indicated by the recent analysis by De Jong et al. (1995). Following De Jong et al. (1995), in this paper, we draw on both of these approaches to examine moving for care among persons living with HIV/AIDS. In broad terms, our analysis focuses on the moves that correspond to the second move in the Litwak and Longino typology (Litwak and Longino 1987; Longino et al. 1991) and assistance-seeking moves in Meyer and Speare’s framework (Meyer and Speare 1985). On theoretical grounds, we might expect that individuals with HIV are likely to be the most mobile prior to their HIV infection, experience more pushes and greater pulls, and have the greatest resources for and the fewest barriers to migration. One might expect, therefore, that white, well-educated, relatively more affluent gay men would be the most mobile; they would face great push and pull factors for initial migration away from families and communities of origin, while they would have the most resources to do so. They might also have greater opportunities to move for college or work and fewer traditional ties that would anchor them in a single place. The lure of urban gay communities might motivate their initial move. While these factors might also play a role post-HIV infection (possibly by creating the structural conditions that obviate the need to move for care), illness and more advanced disease stage may also be important over and above or in place of these factors. This is the so-called second move for assistance; a move for care, either informal or formal. Additionally, we might also suppose that individuals with fewer economic and social resources (e.g., possibly those infected through intravenous drug use) might be increasingly likely to move post-HIV infection as their need for care might exceed their ability to provide for themselves or the resources available in their communities. Finally, some sub-populations may also be distinctively motivated to move for access to specific treatment protocols or greater access to medical care. Data and Methods The HIV Cost and Services Utilization Study (HCSUS) contains a nationally representative probability sample of HIV-infected adults receiving medical care within the contiguous United States. The reference population was limited to persons 18+ years old with known HIV infection who made at least one visit to a non-military, non-prison medical provider (other than an emergency department) in the context of regular or ongoing care during a predefined population definition period (PDP) in early 1996. The PDP extended from January 5, 1996 to February 29 1996 in all but one metropolitan area, where the start was delayed until March. We describe key elements of the study design below (for more complete descriptions of the study designsee Shapiro et al., 1999;and Frankel et al., 1999). The HCSUS used a complex, multi-stage probability sampling design in which geographic areas, medical providers, and patients were sampled in successive stages. Of the 4,042 eligible participants sampled, we interviewed 76%; 71% completed long-form interviews (2,864 interviews) and 5% completed short-form or proxy interviews. For a further 16%, we obtained some basic non-response data from providers. The overall coverage rate is 68% for the long-form interviews, most of which were conducted in person using CAPI instruments that were designed for this study (Berry et al. 1999). Interviewers approached anonymously selected subjects for interview only after providers or their agents obtained permission. Baseline interviews began in January 1996 and ended 15 months later. Only data from the long-form interviews are included in our analyses. Analytic weights were developed to adjust for differential sampling probabilities, non-response, and multiplicity. The product of these three weights forms the analytic weight for each respondent, which is equivalent to an estimate of the number of persons represented by that respondent (Duan et al. 1999). All analyses incorporate these analytic weights. To adjust the standard errors and statistical tests for the differential weighting and complex sample design, we used the linearization methods (Kish and Frankel 1974) available in the STATA software package (Duan et al. 1999). Measures Dependent Variables The HCSUS questionnaire included two questions that make it possible to determine whether a respondent: (1) ever moved (Have you always lived in (name of the city where the respondent currently lives); and (2) if they ever moved (meaning changed city or state) since you’ve known you were HIV positive. Using responses to these two questions, it is possible to create a 3-category outcome: never moved, moved prior to knowing about HIV infection, and moved since knowing about HIV infection. The HCSUS questionnaire also asked persons who had moved since learning about their HIV infections to indicate the reasons for their most recent move (more than one reason could be cited). They were asked to respond yes or no regarding whether each of the following was a factor in their move: (1) to be near people who could help take care of you when you are sick (37.4%); (2) to get away from negative attitudes or discrimination where you previously lived (30.7%); (3) to be near a group or community that shared your needs or interests (30.2%); (4) reasons related to work or school (23.5%); (5) wanting a physician more accepting or knowledgeable about treating people with HIV (22.4%); (6) the availability of regular medical care or the distance to travel to reach it (19.5%); (7) getting Medicaid benefits or other services, such as case management, substance abuse treatment, housing services, food, or financial assistance (16.8%); (8) the cost of regular medical care (9.1%); and (9) the ability to enroll in a clinical trial or drug program (8.5%). As indicated in parentheses above, the most frequently endorsed reason was to be near an informal caregiver. The next three most frequently endorsed items do not appear to be related to issues of care per se; they focus more on concerns about community, school, and work. The final set of items relate to issues of formal care (medical or social service system-related care or support). Using these items, we created variables that measure moving for any care (informal or formal together) versus moving for factors un-related to care, and variables measuring moving for informal and formal care respectively. Independent Variables Our preliminary models include gender, race/ethnicity (white, African American, Hispanic, and other race/ethnicity), age, educational attainment, and income. We also include the individual’s exposure/transmission category (men who have sex with men [MSM] only, intravenous drug use [IDU] only, MSM+IDU, heterosexual transmission, hemophilia/transfusion, unknown), and two stage of disease variables: whether they have been diagnosed with AIDS and their lowest CD4 Count ever. Additionally, because those who have known about their HIV longest have a greater risk of moving since knowing about their HIV infection, while the converse is true for those who learned of their infection more recently (i.e., they have a greater risk of having moved prior to knowing about their infection), we include in the model a variable that measures the year the persons first tested positive to adjust for this selection. Preliminary Results Note: We refer to tables in the text, but have not provided the tables with this submission because we are concerned about formatting. These preliminary tables are available upon request (aslondon@maxwell.syr.edu). As seen in Table 1, the HCSUS sample represents a population of a little less than 250,000 adults receiving medical care for HIV in the contiguous United States in early 1996. The population is about three-quarters male, 50% white, one-third African American, and mostly between the ages of 30 and 49. More than half the population has high school education or less, while almost three-quarters have total family incomes less than $25,000 in 1995. Approximately 75% reported that they were exposed to HIV through sex with other men or through intravenous drug use; 20% reported exposure through heterosexual sex. A little over a third of the population had been diagnosed with AIDS, while over half had lowest ever CD4 cell counts below 200. About 60% found out about their HIV infection within the prior five years. It is noteworthy that a relatively small number of people gave a pre-1985 response. This variable was measured with a question about when they first tested positive. The HIV-antibody test was not available until 1985. We did not adjust these answers or exclude these cases because we assumed they indicated (in retrospect) when they could first date their illness. Table 2 presents the bivariate associations between each of the demographic and stage of disease variables selected for this analysis and the 3-category migration outcome (never moved, moved prior to HIV infection, moved after HIV infection). Overall, 27% of the population had never moved, 41% moved prior to knowing about their HIV infection, and 32% moved since knowing about their HIV infection. Each of the demographic and stage of disease variable are associated significantly with this outcome at the bivariate level, except CD4 count. As seen in Table 3, many more variables are associated with moving prior to knowing about HIV infection than are associated with moving since knowing about it. The odds of having moved prior to knowing about HIV infection (relative to never having moved) are significantly reduced for African Americans and Hispanics relative to whites, and persons exposed to HIV through intravenous drug use, heterosexual transmission, and hemophilia/blood transfusion relative to exposure through sex between men. The odds of having moved prior to knowing about HIV infection are significantly higher among older men (those aged 40-49 and 50+ relative to men aged 18-29), those with college educations or more (relative to those with less than high school education), and all those with family incomes over $5,000. As expected, there is also a significant association with years since testing positive; as the year since testing positive gets larger — more recent — the odds of having moved before testing positive increase significantly. Taken together, these results suggest that white gay older well-educated and relatively more affluent men are the most likely to have moved prior to knowing about their HIV infection. This is consistent with the migration of men to urban gay enclaves at earlier periods in their life course, as well as out-of-state migrations among more affluent youth related to education and work. Relative to never having moved at all, the odds of moving since finding out about being HIV positive are significantly lower for African Americans, intravenous drug users, persons exposed through heterosexual transmission (at least marginally so), persons exposed through hemophilia/blood transfusion, and persons with unknown transmission route (relative to males exposed through same sex sexual behavior). They are also lower among persons who found out about their infections more recently (as would be expected). The odds of having moved since becoming infected are higher among those with college educations or more. Taken together, these results suggest that well-educated white gay men are also relatively mobile post-HIV infection. Overall, about half of those who moved post-HIV infection (51.3%) moved for care since finding out about their HIV infection (and recall that 60% had heard about their infection between 1991 and 1996). As seen in Table 4, at the bivariate level, moving for care is associated with gender, with females being more likely to move for care than males. It is also associated with education and income, with persons with less education and lower income being more likely to move for care. Persons exposed through hemophilia/transfusion are very likely to move for care, and persons with more advanced disease (i.e., persons with AIDS and persons with lower lowest CD4 counts) are respectively more likely to move for care. Table 4 also presents descriptive data on those who moved for informal only, formal only, and both types of care respectively among those who moved for care. Roughly a third of the sample falls into each category and there is very little variation across most of the sociodemographic and disease stage variables. We estimated a two-part logistic regression model. The first part predicts moving for care among those who ever moved post-HIV infection. The second part predicts moving for informal and formal care among those who moved for care. These latter two equations are not mirror images of one another because of the group who moved for both informal and formal care; they appear in the “1" category in both models. As seen in Table 5, the odds of moving for care are significantly higher among women, persons exposed through intravenous drug use and same-sex sexual behavior, and persons with AIDS, while they are significantly lower among those with the highest educational attainments and incomes. This is consistent with the notion that moves for care are associated with greater need and fewer resources. Moves for informal care (as opposed to formal care only) among those who moved for care are marginally higher among women, and there is some evidence that such moves are lower among those with higher incomes and persons exposed to HIV through hemophilia/blood transfusion. Among those who moved for care, this latter group is very likely to have moved for formal care. Discussion Our preliminary results suggest that the models and conceptual frameworks developed for the purpose of analyzing the migration patterns of the elderly may be generalized with modifications for use in analyses of the migration patterns of persons with HIV/AIDS and potentially other non-elderly populations. We believe that further investigation of moving for care is can contribute to our understanding of internal migration. While there is discussion of these issues in the gerontology literature, demographers have not adequately addressed questions about moving for care generally. Our analyses of moving for care among persons with HIV/AIDS suggest that moving for care is a phenomenon that might be worth examining in more detail. In an aging population and a society that privatizes a great deal of health care within the family system, it would seem that a significant but at present unknown proportion of internal migration is care-related. The evidence presented here can complement models developed for the elderly and other populations with specific needs and concerns. It appears that the second move model applies to some extent in this population, as it does for the elderly, and that resources also shape care-related moves. There might be other care-related moves that would be interesting to investigate. For example, other life cycle migrations may be care-related. Newlyweds moving closer to parents prior to or just after starting childbearing to be closer to (grand)parents for help with caring for children. This might be conceptualized as another form of second move. In general, we argue that understanding motives for moves generally and care-related migration specifically better would provide new insights into the dynamics of internal migration and may yield information that would be useful to policy planners and program managers.