Data Resource Profile: The Manitoba Multigenerational Cohort

Key Features.

• The Manitoba Multigenerational Cohort (MMC) has been developed using routinely collected and de-identified administrative healthcare data to facilitate health and social research using family-based designs.

• The MMC is a population-based cohort of individuals born in the province of Manitoba, Canada, from 1974 to 2019 (n = 744 265) with linkages across four generations and multiple family relationships: parents, siblings, grandparents, aunts, uncles, cousins and great-grandparents. Overall, 98% of individuals are linked to mothers; 60% are linked to fathers.

• Individuals in the MMC are followed from birth until death or out-of-province migration. The median follow-up ranges from 1 year (for those born in 2019) to 27.7 years (for those born in 1974).

• The MMC can be linked to other health, social, education and justice data sets contained in the Manitoba Population Research Data Repository.

• Access to the MMC requires review by the Health Information Privacy Committee of Manitoba Health and Seniors Care and ethics approval through the University of Manitoba Health Research Ethics Board. Researchers interested in using the MMC for their own studies are encouraged to visit http://umanitoba.ca/faculties/health_sciences/medicine/units/chs/departmental_units/mchp/resources/access.html and/or e-mail mchp_access@cpe.umanitoba.ca.

Data resource basics

Families are important components of society; given their shared genetic and social environments, studying families can provide critical insight into health and social outcomes within family members and across generations. Few places in the world have high-quality linkages within families at a population level; one such location is Manitoba, Canada. The Manitoba Multigenerational Cohort (MMC) has been developed to facilitate health and social research using family-based designs.

The MMC is derived from the Manitoba Health Insurance Registry (‘the Registry’)—a population-based registry of all individuals registered with Manitoba Health and Seniors Care (MHSC).¹ The Registry is updated at the Manitoba Centre for Health Policy (MCHP) twice a year and is integrated with historical registry data to create a longitudinal population-based registry. Over 80 administrative and survey-based data sets are linkable at the individual level; these data sets make up the Manitoba Population Research Data Repository (‘ the Repository’) and are housed in a secure environment at the MCHP.

Manitoba is a Canadian province that has >1.2 million residents; >700 000 reside in Winnipeg, the capital and largest city in Manitoba.² The Registry captures the entire Manitoba population through a universal and publicly funded healthcare system, which covers hospitalizations and outpatient physician visits among other services. Within the Registry, individuals are identified based on a unique personal health identification number and families are identified through a unique family registration number (FRN); both numbers are assigned at birth. Partners who are married or in a common-law relationship can be listed under the same FRN (although this is not required)—known as the ‘head of household’ FRN. Offspring are listed under the head of household FRN. When offspring turn 18 years old (or 19 years old prior to 1992), they receive their own FRN .

The MMC, which was created in 2020, includes individuals born in Manitoba from 1974 to 2019 who are identified in the Registry. Individuals born in Manitoba remain in the Registry until they die or move out of the province; these longitudinal data allow for linkages within families and across generations. For example, an individual's FRN can be used to identify parents and siblings; their parents' FRN can be used to identify grandparents, aunts and uncles; and their aunts’ and uncles' FRN can be used to identify cousins. The use of FRNs in Manitoba has changed over time. Up to about 2000, it was typical for both mothers and fathers to share a FRN, but this is not mandatory. As nontraditional family structures became more prevalent, it has become more common for children to share an FRN with only one parent, which is usually their mother. For more information on the specification of family relationships identified in the MMC, visit http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?conceptID=1515.

In total, 744 265 individuals were born in Manitoba from 1 April 1974 to 31 March 2019 (Table 1). Individuals in the MMC can be followed from birth to end of health-insurance coverage, resulting in ≤46.5 years of follow-up (for those born in 1974). Many individuals have linkages to parents, siblings, grandparents, aunts, uncles, cousins and great-grandparents. The percent of individuals with linkage to family members for each fiscal year (April–March) are presented in Supplementary Tables S1–S3 (available as Supplementary data at IJE online).

Table 1.

Familial linkages available in Manitoba, overall and in 1 year per decade, 1974–2019

	Total	Fiscal (March–April) birth year
		1974	1980	1990	2000	2010	2019
Number of births	744 265	18 557	16 729	17 424	14 154	15 768	16 600
Years of follow-up (median, range)	16.8 (0.0, 46.5)	27.7 (0.0, 46.5)	26.7 (0.0, 40.5)	24.1 (0.0, 30.5)	17.6 (0.0, 20.5)	9.3 (0.0, 10.5)	1.0 (0.0, 1.5)
Linkage to family members (%)
Mother	98.1	93.3	95.3	98.8	99.3	98.9	98.8
Father	59.9	85.5	83.4	68.1	52.2	39.7	36.0
Siblings
Multiple birth, e.g. twin	1.5	3.1	2.0	1.4	1.2	1.0	1.1
Full (non-twin) sibling	51.9	75.8	73.0	59.6	44.8	34.5	23.7
Half sibling	1.7	2.0	2.9	2.6	1.5	0.6	0.2
Sibling; share at least one parenta	39.1	11.7	16.5	34.5	51.0	58.3	42.2
Grandparents
Maternal grandmother	65.7	18.9	43.4	71.4	82.0	77.4	70.0
Maternal grandfather	55.0	17.4	40.2	64.4	70.1	59.1	47.2
Paternal grandmother	36.2	32.2	44.6	46.7	39.1	24.9	16.2
Paternal grandfather	34.3	30.3	42.3	44.2	37.2	23.5	14.8
At least one biological aunt	60.2	29.4	49.2	69.4	70.1	62.3	53.9
At least one biological uncle	62.4	31.3	51.6	71.1	72.8	64.5	56.1
At least one cousin	57.0	28.5	48.8	69.0	69.3	56.3	38.9
At least one great-grandparent	24.9	0.1	0.2	6.2	31.8	54.5	59.9

^aOnly one common parent is known due to missing linkages to either mother or father for one or both siblings.

Over time, linkages to biological mothers have improved and linkages to biological fathers have worsened (Figure 1). By 1998, 82% of cohort members had linkages to maternal grandmothers (decreasing to 70% by 2019) and 71% had linkages to maternal grandfathers (decreasing to 47% in 2019). Fewer individuals had linkages to paternal grandparents. Before 1986, very few individuals had linkages to any great-grandparents; this increased to just under 57% by 2019.

Figure 1.

Percentage of births that have linkages to parents, grandparents and great-grandparents over time

For 30 years, the Manitoba government has funded the MCHP to maintain the Repository. Research projects using data from the Repository do not require individual consent based on legislation allowing the use of these routinely collected and de-identified data under conditions that maintain privacy and confidentiality.

Data collected

Measures

The MMC contains information on each individual's personal health identification number and those of family members identified through the FRN: the mother and father, the maternal and paternal grandparents, and the maternal and paternal great-grandparents. The MMC also has information on each individual's birth date, sex, end of coverage date and reason for end of coverage (e.g. death, migration out of Manitoba).¹

Data sets

Linkability to health data

Among the administrative health data sets in the Repository linkable to the MMC, Hospital Abstracts, Medical Claims and the Drug Program Information Network records have been widely used in research (Table 2); these databases contain information for all Manitoba residents eligible to receive health services through the MHSC. The Hospital Abstracts database includes ∼300 000 records per year. From 1970 to 1979, diagnoses and procedures were recorded using the International Classification of Diseases (ICD), Adapted, 8^th Revision (ICDA-8) coding system.³ The 9th ICD edition with clinical modifications (ICD-9-CM) was used to report diagnoses and procedures from 1979 to 2004.⁴ Since 2004, the 10th ICD edition with Canadian enhancements (ICD-10-CA) and the Canadian Classification of Health Interventions (CCI) have been used to report diagnoses and procedures, respectively.⁵

Table 2.

Examples of available data sets in the Manitoba Population Research Data Repository that can be merged with the Manitoba Multigenerational Cohort

Data set/data provider	Data years	Data level	Main measures
Health data
Hospital Abstracts/MHSC	1970–Present	Individual	Demographic information and clinical data with up to 25 diagnosis codes and 20 procedure codes
Medical Claims database/MHSC	1970–Present	Individual	Diagnoses, procedures and physician specialty information
Drug Program Information Network/MHSC	1995–Present	Individual	Prescription information such as drug name, dosage, quantity and prescription date
Emergency—Admission, Discharge, and Transfer and E-Triage/MHSC	1999–2011	Individual	Emergency, admission, discharge, transfer and triage information such as method of arrival to ER , disposition status, time spent in ER and proportion who left against medical advice
Long Term Care Assessment/MHSC	2000–Present	Individual	Functional and clinical assessments and screening
The Manitoba Cancer Registry and Treatment/MHSC	1984–Present	Individual	Patient characteristics, cancer diagnosis and treatment
The Manitoba Immunization Monitoring System/MHSC	1986–Present	Individual	Information on immunization history (since 2000 for adult Manitobans)
Social data
Canada Census/Statistics Canada	1971–Present	Aggregate	Area-level census information on demographics, including age, sex, marital status, education, employment and income
Babies First and Families First Screen/Manitoba Families	2000–Present	Individual	Parental and child biological, demographic and social risk factors such as alcohol and drug use, smoking and criminal involvement during pregnancy
The Social Allowances Management Information Network (SAMIN)/Manitoba Families	1995–Present	Individual	Monthly records of all individuals receiving Employment and Income Assistance, type of assistance and relationship to primary applicant
Child and Family Services Information System (CFSIS)/Manitoba Families	1992–Present	Individual	Indicators on health, maltreatment, disabilities, injuries, and school and social outcomes for children and families receiving support and protection services from Child and Family Services
Other data
Enrollment, Marks, and Assessments/Manitoba Education	1995–Present	Individual	Demographics and data on school enrolment, courses, marks, tests and graduation status
Vital Statistics Mortality/Vital Statistics	1970–Present	Individual	Demographic information such as birth date, sex and marital status, and death information, including the date and place of death and the primary and underlying causes of death
Prosecution Information and Scheduling Management (PRISM)/Manitoba Justice	2002–Present	Individual	Involvement with the criminal justice system; incident type, type of involvement (accused, victim or witness) and the initial charges

ER, emergency room; MHSC, Manitoba Health and Seniors Care.

The Medical Claims database includes claims for physician visits in offices, hospitals and outpatient departments, which are collected for reimbursement purposes. The database contains ≤17 million records per year. Diagnoses are identified using ICDA-8 from 1970 to 1979 and ICD-9-CM onward, and procedures are identified using tariff codes, which are numerical codes used to identify medical services delivered by physicians and nurse practitioners. Tariff codes are relatively stable over time; however, codes are added or removed based on changes in the medical knowledge, terminology and practices. For example, new tariff codes were created in response to the COVID-19 pandemic and the resulting need for virtual patient care.⁶ The Drug Program Information Network is an electronic prescription-drug database, which includes complete profiles of drugs dispensed from community pharmacies for all the Manitoba population since 1995 and captures close to 12 million transactions per year.

Other health data sets in the Repository include the Emergency—Admission, Discharge, and Transfer and E-Triage data set, which describes Emergency Department visits in the city of Winnipeg, the largest city in the province, where ∼60% of the provincial population reside; Long Term Care Assessment data set, which includes functional and clinical assessments and screening of residents of Winnipeg personal care homes; the Manitoba Cancer Registry and Treatment, which describes patient characteristics, cancer diagnosis and treatment; and the Manitoba Immunization Monitoring System, which contains information on immunization history (Table 2).

Linkability to social data

Canada Census, Babies First and Families First Screen, the Social Allowances Management Information Network (SAMIN) and the Child and Family Services Information System (CFSIS) are a few of the social data sets contained in the Repository (Table 2). Canada Census is a nationwide population survey conducted by Statistics Canada every 5 years. This data set has been used to create area-level socio-economic status measures such as income quintiles and the Socioeconomic Factor Index (SEFI).^7–10 The SEFI uses social determinants of health, including employment, education, income and proportion of single parents, to estimate a proxy of socio-economic status.¹¹ The first Canada Census was conducted in 1971 and reported a Manitoba population of 988 247; the number increased to 1 278 365 individuals in the most recent census conducted in 2016.^2,12

Babies First and Families First Screen contains information on parental and child biological, demographic and social risk factors of ∼14 000 births per year (∼83% of all live births).¹³ Families with newborns agreeing to participate in the screen are interviewed by a Public Health Nurse within a week of the newborn's hospital discharge to collect information on the risk factors. The SAMIN data set provides information on all Manitoba residents who received assistance from the Employment and Income Assistance Program; this data set contains ∼50 000 individual records per year. Receiving income assistance has been used in previous studies as a proxy for individual-level socio-economic status.^7,9,14 The CFSIS includes information on children in care and families receiving protection and support services from the Manitoba Child and Family Services (CSF) (>103 000 cases). The data set was used in previous studies to identify children who were taken into the care of child-protection services.^15,16

Linkability to other data

The Enrollment, Marks, and Assessments data set is an example of the educational data contained in the Repository (Table 2). It provides information on Manitoba students from Kindergarten to Grade 12. Among the recorded measures, high-school-graduation status has been used in previous research to measure educational level; test data have been used to measure educational attainment.^9,17–21 The enrolment file of the data set includes records of ∼180 000 students annually.

The Vital Statistics Mortality data set is a population-based registry of deaths in Manitoba (Table 2). It is a legal requirement for all deaths to be reported to both the federal and provincial governments; therefore, the Registry includes records of virtually all deaths occurring in Manitoba. This death registry includes ∼10 000 records per year and provides demographic and death information, such as the underlying cause of death. The Prosecution Information and Scheduling Management (PRISM) data set contains records of all contacts with the justice system (Table 2). The data set currently holds 2 million incident records of >500 000 individuals.

A complete list of the data sets contained in the Repository and are linkable to the MMC can be found at http://umanitoba.ca/faculties/health_sciences/medicine/units/chs/departmental_units/mchp/resources/repository/descriptions.html. The MCC can additionally be linked to Electronic Medical Records in Manitoba available through the Manitoba Primary Care Research Network (MaPCReN).²²

Data anonymization

The linkage of family members in the Registry and across different data sets is based on de-identified records to maintain privacy and confidentiality. The MHSC assigns encrypted personal health identification numbers to each individual, which are common across all health data sets and the MMC. FRNs are available in the Registry as a restricted data element. Access to FRNs is limited to MCHP data analysts; external researchers require special permissions to access these identifiers. Merging with non-health data sets is performed at the MHSC using identifying information; this information is removed before the files are transferred to the Repository.

Data quality

A semi-automated process is used to assess the quality of the acquired data before including them in the Repository; this process uses both re-abstraction and data-linkage approaches. The evaluation is based on a data-quality framework with five core dimensions: accuracy, internal validity, external validity, timeliness and interpretability.²³

Data resource use

Whereas the formalization of the MMC data is recent, multigenerational data have been used to conduct research for more than a decade. These studies fall into three categories: sibling and/or cousin comparisons, intergenerational transmission of diseases and effects of exposures across the generations. Examples of each of these applications are provided below. A full citation list of previous publications from 2008 to 2020 is available in Supplementary File S1 (available as Supplementary data at IJE online).

Sibling and/or cousin comparisons

Using family members as a comparison group provides a tool to account for unmeasured confounding by genetics and shared environmental factors such as socio-economic status. For example, a previous study examined the risk of mortality among Manitoban mothers whose children were taken into care by child-protection services.¹⁵ Since mortality tends to cluster in families due to environmental and genetic factors, the study used a discordant-sibling analysis to account for these factors that present a potential confounding source. The study included 1974 biological sisters who were discordant in exposure status, i.e. having a child taken into care. Mothers who had a child taken into care were three times more likely to die during the study period than their sisters who did not have a child taken into care. This finding highlights the need for specialized support programs and public health interventions for mothers involved with child-protection services.

Another study examined school readiness in children who were placed in the care of child-protection services before the age of 5 years.¹⁶ The study created a cohort of 53 477 children and subcohorts of discordant siblings (n = 809) and discordant cousins (n = 517). Children placed in care were significantly less likely to be ready for school in the analysis based on the entire cohort but not in the sibling or cousin subcohorts. This indicates that the observed association in the entire cohort is likely confounded by shared familial factors, accounted for in the sibling and cousin subcohorts.

Intergenerational transmission of diseases

Complex diseases often cluster within families and are passed across the generations.^24–28 These trends can be studied at the population level using multigenerational record linkage such as the MMC linkages. For example, a study examining the intergenerational association of preterm birth²⁹ found that women born at term with a sister born preterm had a similarly elevated risk of delivering a preterm infant as their preterm sisters.

Another research investigated the intergenerational transmission of osteoporotic fractures in a cohort of >200 000 offspring.^30,31 Both parental hip and non-hip osteoporotic fractures were independently associated with increased risk of offspring osteoporotic fractures. Sibling osteoporotic fractures were also associated with an increased risk of osteoporotic fractures.³² These findings enable improving fracture risk-assessment tools using family fracture history.

An ongoing project is examining the familial relationships in substance use disorder (SUD) in MCC individuals born from 1984 to 2000. It aims to test whether having a mother and/or a father with SUD diagnosis increases the offspring risk of adolescent SUD and whether having an older sibling with adolescent SUD increases the risk of the same disorder in a younger sibling.

Effects of exposures across generations

Multigenerational record linkage enables studying the effects of exposures across the generations. For example, a study investigated whether the grandchildren of adolescent mothers have lower school-readiness scores than their peers in a cohort of 11 326 children.³³ Children with grandmothers who were adolescent mothers were less likely to be ready for school, regardless of whether the children's mothers were also adolescent mothers, which supports the long-term intergenerational effects of adversity in families.

Another study examined the social, mental and physical outcomes of 1415 parents bereaved by the suicide death of an offspring.³⁴ Parents were more likely to have depression and anxiety disorders and have a marital breakup in the 2 years after offspring death compared with the period before offspring death. This highlights the need for support initiatives and clinical assessment of grieving parents for the early management of arising mental health issues.

Strengths and weaknesses

There are several strengths to the MMC. First, it is a population-based registry, which facilitates creating large cohorts representative of the entire population and enables examining rare outcomes. Second, the data span decades and allow examining long-term outcomes and undertaking longitudinal investigations. Third, the MMC can be merged with a rich collection of health, social, education and justice data. This merging of the MMC with other health and non-health data sets broadens the scope of research questions that can be addressed and allows capturing a large number of important health and non-health measures and their proxies. For example, COPD diagnosis identified in hospital and physician-visit records has been used as a population-based proxy of smoking; substance- or alcohol-use diagnosis has been used as a proxy of high alcohol intake.^30,32,35 Fourth, numerous disease case definitions have been validated to accurately and objectively ascertain disease measures in the Repository, such as diabetes, asthma and hypertension.³⁶ Finally, with appropriate approvals, the MMC can be updated to include children born in years later than 2019, as the data for these children become available.

There are a few limitations to note. First, children adopted at birth will have the same FRN as the adoptive parents, resulting in misclassification of biological relationships in the MMC. The amount of misclassification is not expected to be significant at the population level. Birth records, available in the Repository, can be used to supplement the MMC and ensure only biological relationships are captured. Second, a challenge arises from the increase in the prevalence of non-traditional family structures, including unmarried couples with children.³⁷ Whereas mothers can be identified for virtually all offspring, identifying fathers in the MMC is conditional on marriage or common-law status being reported for the father to be assigned the same FRN. Therefore, the ability to establish paternal-offspring linkage has been declining over time.

Data resource access

The MCHP welcomes researchers who wish to conduct multigenerational studies to apply for access to the MMC in these four steps.³⁸ First, all researchers using the MMC must complete the MCHP Accreditation, which provides an overview of the MCHP, data access and available data. Second, to ensure that the data elements required to answer specific research questions are available, a feasibility assessment is to be conducted by the MCHP. Third, the investigators must have their project approved by (i) the university research ethics board, (ii) the government-mandated Health Information Privacy Committee and, if requesting linkages to non-health data sets, (iii) approvals from non-health data providers. Specific data sets in which Indigenous individuals are overrepresented, such as the CFSIS data set, also require approvals from the Health Information Research Governance Committee within the First Nations Health and Social Secretariat of Manitoba and the Manitoba Metis Community Research and Ethics Protocol within the Manitoba Metis Federation. Fourth, after all approvals have been obtained, the investigators will complete a researcher agreement between the University of Manitoba and the MHSC. On a strictly cost-recovery basis, there are some charges related to the administration and housing of the data at the MCHP.

Investigators can either work with the data directly or with an MCHP-based analyst to carry out the proposed research. To access the data directly, investigators must log into the MCHP's virtual private network (VPN); access to this VPN is available only in Canada. If working with an MCHP-based analyst, that individual will carry out all analyses at the MCHP and share procedures and aggregate results according to the analysis plan and researcher agreement. Whereas most studies are conducted by MCHP-affiliated researchers, hundreds of projects using the data housed at the MCHP have been conducted by external and international investigators, as long as they collaborate with a Manitoba-based investigator. A local collaborator is a requirement to obtain ethics approvals. Researchers interested in using the MMC for their own research are encouraged to visit http://umanitoba.ca/faculties/health_sciences/medicine/units/chs/departmental_units/mchp/resources/access.html and/or e-mail mchp_access@cpe.umanitoba.ca to learn how to apply for access.

Ethics approval

The study was approved by the University of Manitoba Health Research Ethics Board and the Health Information Privacy Committee of Manitoba Health and Seniors Care.

Author contributions

Concept and design: A.F.H., E.W.W., L.L.R.; acquisition and analysis: E.W.W., R.W.; interpretation of data: all authors; drafting of the manuscript: A.F.H., E.W.W.; critical revision of the manuscript: all authors.

Supplementary data

Supplementary data are available at IJE online.

Funding

This work was funded through E.W.W.’s Canada Research Chair in Population Data Analytics and Data Curation. L.M.L. is supported by a Canada Research Chair in Methods for Electronic Health Data Quality. M.L.U. is supported by a Canada Research Chair in Applied Population Health. L.L.R. acknowledges financial support from the Canadian Institutes of Health Research [PJT-162111].