Beware of multiple names in database linkage research: prevalence of aliases in female prison populationBMJ 2005; 331 doi: https://doi.org/10.1136/bmj.331.7512.335 (Published 04 August 2005) Cite this as: BMJ 2005;331:335
- Ruth Elwood Martin (), clinical associate professor1,
- T Gregory Hislop, senior epidemiologist2,
- Garry D Grams, assistant professor3,
- Veronika Moravan, statistician4,
- Betty Calam5, associate professor
- 1 Department of Family Practice, Mather Building, 5804 Fairview Crescent, University of British Columbia, Vancouver, V6T 1Z3, British Columbia, Canada
- 2 Cancer Control Research, BC Cancer Agency, Vancouver, V5Z 4E6
- 3 Department of Family Practice, University of British Columbia
- 4 Population and Preventive Oncology, BC Cancer Agency
- 5 Department of Family Practice, University of British Columbia
- Correspondence to: R Elwood Martin
- Accepted 2 July 2005
A common strategy for health services research involves linking the records held in administrative databases using personal name data. Studies may also require following individuals over time, again necessitating names.
Multiple names became a dilemma in our research of a screening intervention for female prisoners in British Columbia, Canada, such that we excluded from database linkage those women's records with at least five surnames or at least four first given names.1 We found no reports in the health services research literature regarding multiple names. We describe the distribution of aliases within a female prison population and compare the sociodemographic characteristics of prisoners with and without aliases. We discuss the implications for healthcare studies involving people with multiple names.
Participants, methods, and results
We investigated the multiple names of 847 female prisoners at the Burnaby Correctional Centre for Women in 2000-1. These women were serving federal sentences (at least two years, including life sentences), provincial sentences (less than two years), or remanded to custody (awaiting sentencing).
We took data from the centre's inmate records for the cervical cancer screening linkage project,1 including all known aliases (multiple names) and the following sociodemographic characteristics: age (as a continuous variable), ethnicity (white, Aboriginal, other), education (no high school, high school, post high school), and type of incarceration (remand, sentenced). We grouped the names into the following classes: surnames, first names, and full names (first names and surname). We used a Poisson model to assess the significance of association of the sociodemographic characteristics with numbers of names (starting at zero for one name) for each class. We used backward elimination to check for independent effects of these factors.
The mean number of names was 2.7 for surnames (maximum 21), 2.4 for first names (maximum 23), and 3.5 for full names; 48% of the inmates had three or more names, and 24% five or more (maximum 28). Ethnicity and type of incarceration were highly significant for all three classes of name (table; P < 0.0001). Aboriginal women had more names than any other ethnic groups. Women serving sentences had more names than women in remand. Level of education had an inverse relationship with number of first names (P = 0.003) and full names (P < 0.03). The number of names increased with age for surnames and full names (P < 0.0001) and first names (P = 0.04).
Multivariable Poisson models showed ethnicity, incarceration type, and age to be significant predictors for numbers of surnames and full names (P < 0.0001) and first names (P < 0.0001, P < 0.0001 and P < 0.01 respectively) after adjusting for covariates. Education level was not a significant independent predictor for multiple names.
What is already known on this topic
The prevalence or distribution of multiple names in women is unknown
What this study adds
Multiple names are common in certain subgroups of female prison inmates, which raises important challenges for health services provision and database linkage research
This first study of multiple names in women prisoners raises caution about representativeness and generalisability of research that uses names for linking databases. Women who were older, Aboriginal, and with longer prison sentences were more likely to be excluded from our database linkage because of multiple names.1
Many women seek their healthcare while in prison. Follow-up is often required after release from prison. Healthcare providers should be aware of multiple names to avoid loss to follow-up in the community.
Men also report multiple names. In a study of US male prisoners, 31% had changed their names at some time for intentional deception.2 Aliases were associated with age, ethnicity, education, and level of criminality in inmates in New Zealand.3 Male forensic patients with multiple names had more psychiatric illness.4
Name change is done for various reasons—marriage, alias, pseudonym, manipulation, fraud, psychosis, amnesia, and factitious disorder.5 “Ethnic naming” is sometimes done to resolve competing national and ethnic identities, to honour family and ancestors, or to seek new purpose when not accepted by the prevalent culture. Ethnic naming may be a common reason for multiple names among Aboriginal people, who comprised a sizeable proportion of our study group. Our findings could have implications for provision of health services and database linkage research among ethnic and Aboriginal communities not limited to the prison population.
We thank the Cervical Cancer Screening Program at the British Columbia Cancer Agency, the Corrections Branch of the Ministry of Public Safety, and The Solicitor General for giving approval for this study and for giving information needed for the linkage.
Contributors Dennis Wardman, community medicine specialist, First Nations and Inuit Health Branch, reviewed and commented on early versions of the manuscript and approved the final version for publication. TGH, GDG, BC, and REM conceived the design, implementation, and evaluation of the cervical cancer screening intervention, which led to the writing of this paper. TGH, GDG, and REM wrote the initial versions of this manuscript and all authors were involved in editing subsequent versions. VM linked the databases and did the analysis for this paper; she was also involved in the writing and editing. REM is guarantor.
Funding British Columbia Medical Services Foundation.
Competing interests None declared.
Ethical approval University of British Columbia.