Patterns of geographic mobility of persons with AIDS in Canada from time of AIDS index diagnosis to death
Robert S. Hogg, PhD
Jeff Whitehead, MD
Maura Ricketts, MD
Katherine V. Heath, MSc
Edward Ng, PhD
Pierre Lalonde, BSc
Martin T. Schechter, MD, PhD
Clin Invest Med 1997;20(2):77-83
[résumé]
Dr. Hogg is with the British Columbia Centre for Excellence in HIV/AIDS, St. Paul's Hospital, and the Department of Health Care and Epidemiology, Faculty of Medicine, University of British Columbia, Vancouver, BC; Dr. Whitehead is with the Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Ont.; Dr. Ricketts is with the Laboratory Centre for Disease Control, Health Canada, Ottawa, Ont.; Ms. Heath is with the British Columbia Centre for Excellence in HIV/AIDS, St. Paul's Hospital, Vancouver, BC; Dr. Ng and Mr. Lalonde are with the Health Statistics Division, Statistics Canada, Ottawa, Ont.; and Dr. Schechter is with the British Columbia Centre for Excellence in HIV/AIDS, St. Paul's Hospital, and the Department of Health Care and Epidemiology, Faculty of Medicine, University of British Columbia, Vancouver, BC.
(Original manuscript submitted Oct. 22, 1996; accepted Jan. 22, 1997)
Reprint requests to: Dr. Robert Hogg, British Columbia Centre for Excellence in HIV/AIDS, 6081081 Burrard St., Vancouver BC V6Z 1Y6; fax 604 631-5464; bobhog@hivnet.ubc.ca
Contents
Abstract
Objective: To characterize migration patterns of persons with AIDS in Canada during the period from AIDS diagnosis to death.
Design: Descriptive, population-based study.
Setting: Canada.
Patients: Canada's AIDS Case Reporting Surveillance System (ACRSS) was linked to deaths in the Canadian Mortality Data Base (CMDB). Probabilistic linkage was based on initials, date of birth, date of death, birthplace, and location at diagnosis and at death. Analysis was restricted to AIDS cases reported from Jan. 1, 1982, to Sept. 30, 1994, and to deaths reported from Jan. 1, 1982, to Dec. 31, 1992.
Main outcome measures: Change in usual place of residence; migration rates by region and community size.
Results: A total of 5755 AIDS cases recorded in the ACRSS were linked to deaths in the CMDB. Of these linked cases, 5366 (93%) included information on province or territory of usual residence or community size. A total of 160 (3.0%) persons with AIDS changed their province or territory of residence between the time of their AIDS diagnosis and death. Multivariate analysis indicated that those who changed residences between AIDS index diagnosis and death were more likely than other persons with AIDS to live in provinces other than British Columbia, Ontario and Quebec (p < 0.001), to be diagnosed earlier (p = 0.004), to be younger (p < 0.001) and to be gay or bisexual (p = 0.042).
Conclusions: Our analysis revealed that only a small proportion of persons changed their residence between AIDS diagnosis and death. Geographic mobility was the greatest among persons with AIDS residing outside of the regions where the overwhelming majority of persons with AIDS in this country reside.
Résumé
Objectif : Définir les tendances de la migration chez les personnes atteintes du SIDA au Canada au cours de la période écoulée entre le diagnostic du SIDA et la mort.
Conception : Étude démographique descriptive.
Contexte : Canada.
Patients : On a établi un lien entre le Système de déclaration et de surveillance des cas de SIDA (SDSCS) du Canada et les décès dans la base canadienne de données sur la mortalité (BCDM). Le lien probabiliste a été fondé sur les initiales, la date de naissance, la date du décès, le lieu de naissance et la localité au moment du diagnostic et à celui de la mort. L'analyse a été limitée aux cas de SIDA déclarés entre le 1[er] janvier 1982 et le 30 septembre 1994, et aux morts déclarées entre le 1[er] janvier 1982 et le 31 décembre 1992.
Principales mesures des résultats : Changement du lieu habituel de résidence, taux de migration selon la région et taille de la communauté.
Résultats : Au total, 5755 cas de SIDA consignés dans le SDSCS ont été reliés à des décès consignés dans la BCDM. Sur ces cas où l'on a établi un lien, 5366 (93 %) comportaient des renseignements sur la province ou le territoire habituels de résidence ou sur la taille de la communauté. Au total, 160 (3,0 %) personnes atteintes du SIDA ont changé de province ou de territoire de résidence entre le moment du diagnostic et celui de la mort. Une analyse multivariée a indiqué que les personnes qui ont changé de résidence entre le moment où le SIDA a été diagnostiqué et celui de la mort étaient plus susceptibles que d'autres personnes atteintes du SIDA de vivre dans des provinces autres que la Colombie-Britannique, l'Ontario et le Québec (p < 0,001), d'avoir été diagnostiquées plus tôt (p < 0,004), d'être plus jeunes (p < 0,001) et d'être gaies ou bisexuelles (p < 0,042).
Conclusion : Notre analyse a révélé qu'une faible proportion seulement de personnes ont changé de lieu de résidence entre le moment où le SIDA a été diagnostiqué et celui de la mort. La mobilité géographique a été la plus importante chez les personnes atteintes du SIDA qui résidaient en dehors des régions où habitent la grande majorité des personnes atteintes du SIDA au Canada.
[ Top of document ]
Introduction
Measuring mobility patterns of persons with AIDS within a nation requires that information be gathered on people's movement between residences. An ideal source of data would contain continuing accounts of the moves that take place. However, available data sources do not provide such an accurate history of moves by individuals. For example, existing AIDS registries typically incorporate only information on the place of residence at the time of AIDS diagnosis, whereas vital-statistics data include only information on the person's residence at the time of death.
Deriving components of those histories from a linkage of AIDS-registry and vital-statistics data has become a substitute for geographic-mobility histories. For example, in the US a recent study examined migration patterns of persons in 12 states from AIDS diagnosis to death.[1] This type of analysis allows one to obtain a truncated picture of the person's mobility during a certain period and to use baseline characteristics from the AIDS case record to characterize those who change residences.
This study was undertaken to characterize the migration patterns of persons with AIDS in Canada. In particular, we were interested in describing the patterns of mobility of persons who had crossed provincial and regional boundaries between the time of AIDS index diagnosis and death. We gathered the information necessary to determine the extent of migration of persons with AIDS between regions and the individual determinants associated with geographic mobility in this population.
[ Top of document ]
Methods
The Canadian National AIDS Case Reporting Surveillance System (ACRSS) has been collecting reports of AIDS, defined using WHO criteria, since 1982. These records contain information on persons with AIDS, including their initials, date of birth, vital status, date of death, geographic information, AIDS-defining illnesses and risk factors for HIV infection, as well as containing reporter information. Between Jan. 1, 1982, and Sept. 30, 1994, 10 391 reports were received at the national office.
The Canadian Mortality Data Base (CMDB) is a national database containing the death-certificate information for every Canadian who has died in this country. The electronic database contains the full name of the person, date of birth, date of death, underlying cause of death (but not antecedent or associated causes), location at time of death, usual residence and place of birth. Between Jan. 1, 1982, and Dec. 31, 1992, 2.0 million death registrations, generating 2.8 million death files, were received at Statistics Canada. The larger number of death files than of deaths reflects Statistics Canada's practice of generating 1 file for each surname in order to be able to link records; consequently, married women who have both a maiden and a married name have 2 files.
The reporting delay for AIDS in Canada during this period was calculated to be approximately 9 months, with 59% of cases being reported by 12 months and 79% by 24 months. We decided to use data up to Sept. 30, 1994, (19 months after the last record used in Statistics Canada) to allow time for cases to be reported.
The record linkage was conducted by Statistics Canada using the Generalized Iterative Record Linkage System. The method links information from 2 databases by comparing 7 overlapping primary-variable fields. The first, middle and last initial, date of birth, date of death, sex and province of diagnosis in the ACRSS were compared with the first letters of the first, middle and last names, date of birth, date of death, sex and geographic fields, including the province of residence and death, in the CMDB.
Four runs were performed based on sex and on whether the ACRSS file contained initials (males with initials, males without initials, females with initials and females without initials). These multiple runs were necessary because a significant proportion (16.1%) of the ACRSS records lacked initials. It was recognized that the probability of a high-quality match was severely compromised by the absence of this important personal identifier. In files lacking initials, only the date of birth, date of death and geographic information were used to make the record linkage. Different comparison rules were created for final determination of a match for files with and without initials.
Fields were compared through a complex iterative process involving calculated probabilities of each field-area match, resulting in a score that reflected the probability of the record linkage being true. For example, the probability of occurrence of a first letter of a name was calculated for the larger of the 2 databases (CMDB). If the letter appeared frequently (i.e., was more likely to appear), then the score for matching was lower than if the letter appeared infrequently. In addition, the program allowed near-matches on dates, using a sliding scale for the nearness of the date of the match. Record linkage efficiency was improved by blocking based on 1 or more of the variables considered to be highly predictive of a match (i.e., date of birth, date of death, initials, and portions of these, such as birth month and year). This blocking process saves computer time by limiting matching to pairs of records that are potentially linkable.
In total, as many as 6 successive block passes were conducted. The first block pass was conducted using only the full date of birth. If there was a match, then the program continued to compare all other variable fields, ending in the calculation of a composite weight for the match. If the first block pass failed to find a match for the full date of birth, the record would then be compared in the next block (i.e., based on date of death).
Statistical analysis
Comparing place of AIDS diagnosis with place of death is an important source of data for measuring the mobility of persons with AIDS. From this 2-way classification, estimates of in- and out-migration rates can be made for a specific region or province.
Comparisons of categorical variables were conducted with MantelHaenszel methods. Proportions were compared with the chi2 test. All reported p values are 2-tailed. Stepwise unconditional logistic regression analysis was used to adjust for potential confounding variables. Subjects with missing values were excluded from multivariate analyses. Variables included in the model were those observed to be statistically significant (p ¾ 0.10) in the univariate analysis.
[ Top of document ]
Results
A total of 5755 AIDS cases recorded in the ARCSS were linked to deaths in the CMDB. Of these linked cases, 5366 (93%) included information on province or territory of usual residence or community population size at the time of AIDS diagnosis. Of the cases for which the province of residence at AIDS diagnosis was known, 2218 (41.3%) were from Ontario, 1590 (29.6%) from Quebec, 1018 (19.0%) from British Columbia and 540 (10.1%) from other Canadian provinces and territories (hereafter referred to as "the rest of Canada"). In terms of community size at the time of AIDS diagnosis, a total of 4478 (83.4%) cases were from communities with populations greater than 500 000, 433 (8.1%) were from communities with populations of between 100 000 and 499 000 and the remaining 455 deaths (8.5%) were from communities of less than 100 000.
A total of 160 (3.0%) individuals with AIDS changed province or territory of residence between the time of their AIDS diagnosis and death. Rates of out-migration for the rest of Canada were approximately double those of British Columbia, Ontario and Quebec. The latter 3 provinces had out-migration rates of 33, 20 and 26 persons per 1000 population, respectively, whereas the rest of the country had a rate of 75 persons per 1000 population.
Table 1 compares all AIDS cases reported to the ACRSS with those linked to the CMDB by selected demographic characteristics. As indicated, the persons with AIDS linked to the CMDB were more likely to be gay or bisexual men (78.5% v. 75.9%; p < 0.001), to be diagnosed earlier (p < 0.001) and to be older (p = 0.028) than all of those in the ACRSS. There was no statistically significant difference between the 2 groups in the relative proportion of male and female cases.
Table 2 and Table 3 compare deaths due to AIDS with respect to change in provincial residence between AIDS index diagnosis and death by selected sociodemographic and regional characteristics. In comparison with those who did not move, persons who moved between provinces were more likely to be diagnosed with AIDS earlier (p = 0.002), to be younger (p < 0.001), to be from Ontario (p < 0.001), to be from the rest of Canada (p < 0.001) and to have lived in communities with a population of less than 100 000 (p = 0.025). The median age at diagnosis for the persons with AIDS who moved was 33 years, whereas the median age at diagnosis for the persons who did not move was 37 years. There was no difference between these 2 groups in terms of sex, HIV risk factors and residence in Quebec or British Columbia at the time of AIDS diagnosis.
Ontario was by far the most popular destination for migrants. About one-third (52, 32%) of the 160 interprovincial AIDS migrants settled there before death. Migrants to Ontario came from the rest of Canada (43%), Quebec (30%) and British Columbia (27%). Other popular destinations were British Columbia and Quebec. These 2 provinces were the destination for 43 AIDS migrants (27%) during the period studied. Most migrants to British Columbia (54%) came from the rest of Canada, and most migrants to Quebec (76%) came from Ontario.
Table 4 presents the results of the multivariate stepwise unconditional logistic regression analysis. As indicated, those who changed residence between AIDS index diagnosis and death were more likely than those who did not change residence to live in provinces other than British Columbia, Ontario, or Quebec (p < 0.001), to be diagnosed with AIDS earlier (p = 0.004), to be younger (p < 0.001) and to be gay or bisexual (p = 0.042). No other variable that was significant in the univariate analysis retained statistical significance in the final model.
[ Top of document ]
Discussion
Our analysis revealed that only a small proportion of persons with AIDS changed province of residence between AIDS diagnosis and death. The overall rate of interprovincial migration, which was comparable to that of the Canadian population, was approximately 30 per 1000 population during the period of study.[2] Geographic mobility had the greatest proportional impact on persons with AIDS residing outside of the provinces of British Columbia, Ontario and Quebec, where the overwhelming majority of persons with AIDS in this country reside.[3] Although there is a greater number of persons with AIDS leaving British Columbia, Ontario and Quebec than leaving other provinces, they represent a smaller proportion of the total population living with AIDS in their provinces than do the migrants moving from the rest of Canada.
The increase in the number of persons with AIDS who died in lower-prevalence provinces can be partially attributed to mobility from higher-prevalence provinces. In the US, where HIV/AIDS has spread considerably to and within lower-prevalence areas,[4] this urbanrural pattern of mobility has had a notable effect on service needs in low-prevalence areas.[57] For example, a recent study in the state of Iowa found that 23% of patients with AIDS being treated at tertiary care facilities were not included in state AIDS-case figures.[6,8] Similarly, a study in rural North Carolina found that, among persons receiving HIV-related care, 55% of those with AIDS were diagnosed outside of the study area.[5] Clearly, if this pattern of mobility into low-prevalence areas continues in Canada, the demand for HIV-related services outside of British Columbia, Ontario and Quebec will increase.
The pattern of geographic mobility exhibited in this population is most representative of young gay and bisexual men. The fact that geographic mobility has the smallest impact on the provinces with the overwhelming majority of persons with AIDS is not surprising, given that metropolitan centres such as Vancouver, Toronto and Montreal, have large gay and bisexual communities. Myers and colleagues[9] have shown that gay and bisexual men are attracted to these larger metropolitan centres from smaller cities and rural areas across the country. This pattern is usually in only 1 direction and may be attributable to these men becoming more economically and socially independent.
Motives behind individual migration are less obvious. Persons with AIDS are a heterogeneous population whose members may not be motivated by similar factors. For example, persons with AIDS may change province or region of residence in order to avail themselves of social support systems unique to some areas or to be near family or friends. Movement may also be precipitated by a need for access to health care or tertiary facilities perceived to be of high quality.[10]
Caution is advised in the interpretation of these data because of the nature of the study design.[11] The main disadvantage of an analysis based on data on place of AIDS diagnosis is that one is uncertain about the date of arrival or length of stay at that location or about previous migration. There are also uncertainties about area boundaries at the time of AIDS diagnosis and about the reporting of AIDS diagnosis for men and women who were not diagnosed at their usual place of residence.
[ Top of document ]
Conclusion
Given the limitations of this approach, we have derived considerable information on the geographic mobility of persons with AIDS in Canada. Most notably, we have demonstrated that geographic mobility had the greatest impact on persons with AIDS residing outside the areas where the overwhelming majority of persons with AIDS reside in this country. Future studies need to characterize the impetus for migration and the services used by persons with AIDS. These efforts will result in a more comprehensive understanding of current resource utilization and requirements and assist in projecting future needs for AIDS-related health care services, especially in low-prevalence areas.
[ Top of document ]
Acknowledgements
This work was supported by the Bureau of HIV/AIDS and STD, Laboratory Centre for Disease Control, Health Canada, and by the National Health Research Development Programme of Health Canada through a National Health Research Scholar Award to Dr. Hogg and through a National AIDS Research Scientist Award to Dr. Schechter.
We thank Bonnie Devlin and Thinh Le for their research assistance.
References
- Buehler J, Frey RL, Chu SY. The migration of persons with AIDS: data from 12 states, 1985 to 1992. Am J Public Health 1995;85:1552-5.
- Dumas J, Bélanger A. Report on the demographic situation in Canada 1995. Current demographic analysis. Ottawa: Demographic Division, Statistics Canada; 1996.
- Quarterly surveillance update: AIDS in Canada. Ottawa: Division of HIV/AIDS Surveillance, Bureau of HIV/AIDS and STD, Laboratory Centre for Disease Control, Health Canada; Apr 1996.
- Gardner LI, Brundage JF, Burke DS, McNeil JG, Visintine R, Miller RN. Evidence for spread of the human immunodeficiency virus epidemic into low prevalence areas of the United States. J Acquir Immune Defic Syndr 1989;2:521-32.
- Rumley RL, Shappley NC, Waiver LE, Esinhart JD. AIDS in rural eastern North Carolina -- patient migration: a rural burden. AIDS 1991;5:1373-8.
- Davis K, Stapleton J. Migration to rural areas by HIV patients: impact of HIV-related health care use. Infect Control Hosp Epidemiol 1991;12:540-43.
- Verghese A, Berk SL, Sarubbi F. Urbs in rure: human immunodeficiency virus in rural Tennessee. J Infect Dis 1989;160:1051-5.
- Davis KA, Cameron B, Stapleton JT. The impact of HIV patient migration to rural areas. AIDS Patient Care 1992;6:225-8.
- Meyers T, Godin G, Calzavara L, Lambert J, Locker D. The Canadian survey of gay and bisexual men and HIV infection: men's survey. Ottawa: Canadian AIDS Society; 1993.
- Hogg RS, Schechter MT, Schilder A, Strathdee SA, Goldstone IL, O'Shaughnessy MV. Access to health care and geographic mobility of HIV/AIDS patients. AIDS Patient Care 1995;9:297-302.
- Shryock H, Siegel JS. The methods and materials of demography. San Diego: Academic Press, 1976.
