Linking Spinal Cord Injury Data Sets to Describe the Patient Journey Following Injury: A Protocol

Abstract

Background:

To optimize traumatic spinal cord injury (tSCI) care, administrative and clinical linked data are required to describe the patient’s journey.

Objectives:

To describe the methods and progress to deterministically link SCI data from multiple databases across the SCI continuum in British Columbia (BC) and Ontario (ON) to answer epidemiological and health service research questions.

Methods:

Patients with tSCI will be identified from the administrative Hospital Discharge Abstract Database using International Classification of Diseases (ICD) codes from Population Data BC and ICES data repositories in BC and ON, respectively. Admissions for tSCI will range between 1995–2017 for BC and 2009-2017 for ON. Linkage will occur with multiple administrative data holdings from Population Data BC and ICES to create the “Admin SCI Cohorts.” Clinical data from the Rick Hansen SCI Registry (and VerteBase in BC) will be transferred to Population Data BC and ICES. Linkage of the clinical data with the incident cases and administrative data at Population Data BC and ICES will create subsets of patients referred to as the “Clinical SCI Cohorts” for BC and ON. Deidentified patient-level linked data sets will be uploaded to a secure research environment for analysis. Data validation will include several steps, and data analysis plans will be created for each research question.

Discussion:

The creation of provincially linked tSCI data sets is unique; both clinical and administrative data are included to inform the optimization of care across the SCI continuum. Methods and lessons learned will inform future data-linking projects and care initiatives.

Introduction

Traumatic spinal cord injury (tSCI) is characterized by sudden onset and, in many cases, significant disability and lifetime costs.^1,2 A patient’s journey through the health care system varies depending on the severity of injury, triage protocols, access to rehabilitation and community services, and ultimately their personal resources and health care needs in the community.³ To better understand and optimize the patient’s journey following tSCI, there is a need to leverage existing population-level administrative data sets such as acute care discharge abstracts and emergency department records. This type of information, for example, would allow SCI programs to see whether patients are triaged according to clinical guidelines and provincial protocols, determine whether transitions into the community are successful by monitoring re-admissions within 30 days of discharge, and identify where patients return to live following injury to ensure community services are available. Administrative data are collected for administrative or billing purposes by medical coding personnel and can be leveraged to study care delivery.⁴ These data sets can provide information on the epidemiology by reporting number of incident cases as well as health care utilization in the community; however, they do not capture the particularities of tSCI, and so it is not possible to understand how the severity of injury affects longer term health outcomes and trajectories through the health care system. Clinical data sets specific to tSCI combined with existing population-level administrative data sets have the potential to examine the patient’s journey in the context of their injury and care provided. The availability of both types of data sets is required to create a learning health system, which is a vision for an integrated health system where “… progress in science, informatics, and care culture align to generate new knowledge as an ongoing, natural by-product of the care experience, and seamlessly refine and deliver best practices for continuous improvement in health and healthcare.”⁵(pp43–44),⁶ The underlying premise of a learning health system is to leverage the power of health data and analytics so the system can learn from each patient and report back to patients, clinicians, policy makers, and other stakeholders on what works within a continuous quality improvement framework.⁵

To inform tSCI care within a province or state, there is a need to first identify the population to ensure data is representative (e.g., individuals residing in rural and urban settings). This requires having clinically valid case definitions for identifying tSCI in administrative data. The gold standard for identifying and defining a tSCI is the International Standards for the Neurological Classification of Spinal Cord Injury (ISNCSCI), which is an assessment performed by a trained clinician.⁷ Identifying relevant codes to describe types of SCI (severity based on the American Spinal Injury Association Impairment Scale [AIS] and the neurological level) and mapping these to the International Classification of Diseases (ICD) codes is required to identify patients in administrative databases.⁸ Several groups have proposed and validated the ICD codes for tSCI, which now makes it possible to approximate the patient’s injury using administrative data.^9–11 Once patients have been identified in administrative data, it is then possible to link to other administrative databases to describe aspects of care delivery (e.g., medical procedures); however, these data sets lack detailed clinical information on SCI care, and the ICD codes on the neurological injury are not equivalent to the full ISNCSCI data.

Clinical spinal cord injury (SCI) registries such as the Rick Hansen SCI Registry (RHSCIR) in Canada and the SCI Model Systems in the United States have generated a tremendous amount of knowledge about tSCI acute and rehabilitation care.^12,13 Clinical data are information provided from health care providers and are a record of a patient’s clinical care. Eligibility for these clinical registries is based on a diagnosis of tSCI using the ISNCSCI examination. The clinical registry data sets contain detailed clinical and outcomes data; however, many of these registries need to be linked to community data sets to describe community health surveillance across the lifespan such as patterns of re-admissions to hospital and services such as home care.¹⁴ Furthermore, they are not population-based and therefore cannot be generalized to the tSCI population for a province or state. There is a benefit, therefore, of using both administrative and clinical registry data and ideally linking these data to inform SCI care.

To address the need for comprehensive data along the patient’s journey, the objective of this article is to describe the methods to deterministically link SCI data from multiple databases across the SCI continuum in British Columbia (BC) and Ontario (ON) to answer epidemiological and health services research questions.

Methods

Research questions

In each province, two SCI linked data sets will be created. One is the “Admin SCI Cohort,” which includes all patients identified within a province as having a tSCI using ICD codes (see Figure 1). The second is the “Clinical SCI Cohort,” the subset of patients for which the clinical data is linked to the administrative data within each province (see Figure 1). These two data sets in BC and ON will be used to answer research questions in the areas of (1) epidemiology (e.g., incidence, prevalence, trends), (2) life expectancy, (3) patient trajectories, (4) community health care utilization, and (5) health care costs. See the box “Examples of Research Questions” for research questions using these two types of data sets. Ethical approval for the linkage was provided by the university and hospital institutional review boards (IRB) in BC and ON. For each of the RHSCIR sites, the original ethics permission and consent language was reviewed to ensure it aligned with the linkage project objectives. Ethics permission was obtained in each province to deterministically link patients within BC (using date of birth and personal health number) and within ON (using personal health number) to the various administrative SCI databases and to link the administrative incident cases to the clinical data for a subset of the patients in each province. See Figure 2 for an overview of the clinical and the administrative data available in BC and ON for the Admin SCI Cohort and the Clinical SCI Cohort.

Figure 1.

Overview of the relationship between the Admin SCI Cohort and Clinical SCI Cohort data sets created for British Columbia (BC) and Ontario (ON).

Examples of Research Questions

Epidemiology

What are the trends in the incidence and prevalence of traumatic spinal cord injury (tSCI) in British Columbia (BC) and Ontario (ON)?

How many patients who sustain a tSCI have a work-related injury, and has the incidence changed over time?

Life Expectancy

What is the life expectancy for patients with varying levels of injury (e.g., cervical, thoracic, lumbar) and severity (American Spinal Cord Injury Association Impairment Scale [AIS] A, B, C, D)?

Has life expectancy following tSCI changed over time?

Patient Trajectories

Do patients who receive care in a trauma or rehabilitation center with specialized SCI care have different long-term outcomes (e.g., reduced hospital re-admissions in the community phase)?

What percentage of patients receive care at a Rick Hansen SCI Registry (RHSCIR) Acute (level 1 or 2 trauma hospitals) site in the provinces of BC and ON? What percentage of patients receive care at a RHSCIR Rehab site in the provinces of BC and ON?

Community Health Care Utilization

How many individuals with tSCI are re-admitted to hospital with secondary complications such as urinary tract infections, pressure injuries?

What factors (e.g., patient, injury, location of residence) predict health care utilization following discharge into the community?

Health Care Costs

What is the cost of re-admissions to hospital for secondary complications to the Canadian health care system?

What are the lifetime costs for individuals with varying levels of injury (e.g., cervical, thoracic, lumbar) and severity (AIS A, B, C, D) in Canada?

Figure 2.

(A) British Columbia (BC) Spinal Cord Injury (SCI) data sets. (B) Ontario (ON) SCI data sets. The administrative data include the Population Data BC Data Sets and the ICES Data Sets. The “Admin SCI Cohort” will include linked administrative data sets (plus BC Trauma Registry data in BC). VerteBase (BC only) and RHSCIR (BC and ON) include clinical SCI data (e.g., type of SCI) along the SCI continuum. The “Clinical SCI Cohort” will be created by linking the clinical SCI data in RHSCIR and VerteBase to the traumatic SCI (tSCI) incident cases (identified in DAD) and administrative data sets; the Clinical SCI Cohort is a subset of the larger Admin SCI Cohort. An Admin SCI Cohort and a Clinical SCI Cohort will be created for both BC and ON. CCRS = Continuing Care Reporting System; DAD = Discharge Abstract Database; HCD = Home Care Database; IPDB = ICES Physician Database; MSP = Medical Services Plan; NACRS = National Ambulatory Care Reporting System; NRS = National Rehabilitation Reporting System; OHIP = Ontario Health Insurance Plan Claims Database; RHSCIR = Rick Hansen SCI Registry; RPDB = Registered Persons Database.

Data sources

In BC and ON, provincial identifiable data holdings are housed through Population Data BC¹⁵ and ICES,¹⁶ respectively. Population Data BC facilitates linkage among data sets to provide patient-level, deidentified longitudinal health and education data for the province of BC to facilitate interdisciplinary research on the determinants of human health, well-being, and development.¹⁵ Similarly, ICES is a repository for publicly funded ON health data.¹⁶ Databases that can be accessed through these entities include but are not limited to data sets created by the Canadian Institute of Health Information (CIHI)⁸; details of the data available are provided on the Population Data BC and ICES websites.^15,16

RHSCIR is a prospective pan-Canadian clinical registry of individuals with tSCI. The 30 acute and rehabilitation sites are located in 15 cities in 9 of the 10 Canadian provinces.¹² Data collected follow the patient’s journey and cover the injury, sociodemographics, pre-hospital/acute/rehabilitation care, and outcomes such as the Functional Independence Measure (FIM).¹⁷ For participants who provide consent, a community follow-up is conducted at 1, 2, and then every 5 years post injury.¹² The Praxis Spinal Cord Institute (formerly the Rick Hansen Institute) sponsors the study and manages the national data. In BC, RHSCIR data are collected from Vancouver General Hospital and GF Strong Rehabilitation Centre (see Table 1). Prior to RHSCIR, a clinical quality improvement database called VerteBase collected data on patients admitted to the Vancouver General Hospital between 1994 and 2004. In ON, RHSCIR was implemented between 2006 and 2010 in four cities (see Table 1). There is also the BC Trauma Registry that captures data on patients admitted to 11 designated level 1, 2, and 3 trauma hospitals in the province.¹⁸ Figure 2 displays the data sets along the SCI continuum of care for BC and ON and highlights both the clinical data (VerteBase, RHSCIR) and the administrative data (Population Data BC Data Sets, BC Trauma Registry, ICES Data Sets).

Table 1.

Rick Hansen SCI Registry (RHSCIR) data

Province	City	Hospital	Year initiated RHSCIR	Included in BC/ON RHSCIR cohort (Yes/No)
British Columbia (BC)	Vancouver	Vancouver General Hospital	2004	Yes
		GF Strong Rehabilitation Centre	2004	Yes
Ontario (ON)	Toronto	St. Michael’s Hospital	2010	No
		Sunnybrook Hospital	2009	Yes
		Toronto Western Hospital	2010	Yes
		Toronto Rehabilitation Institute-Lyndhurst	2010	Yes
	Hamilton	Hamilton General Hospital	2006	Yes
		Hamilton Health Sciences Regional Rehabilitation Centre	2006	Yes
	London	University Hospital	2009	No
		Victoria Hospital	2009	No
		Parkwood Hospital	2009	No
	Ottawa	Ottawa General Hospital	2010	Yes
		Ottawa Hospital Rehab Centre	2010	Yes

Cohort creation

Patients with tSCI will be identified from Population Data BC data repository using ICD-9 (1995–2002) and ICD-10 CA codes (2002–2017) (see Table 2) from the Discharge Abstract Database (DAD) between April 1, 1995, and December 31, 2017.^19,20 In ON, ICD-10 CA codes from DAD housed in the ICES data repository between April 1, 2009, and December 31, 2017, will be used.²⁰ The tSCI case definition was validated by mapping ICD-10 CA codes to ISNCSCI descriptions of tSCI by level and severity: cervical complete, thoracic complete, lumbar complete, cervical incomplete, thoracic incomplete, lumbar incomplete, and cauda equina.¹⁰

Table 2.

International Classification of Diseases (ICD)-9 and ICD-10 CA diagnostic codes to identify incident traumatic spinal cord injury (tSCI) cases

ICD-9 codes for tSCI (until 2001/2002)
806	Fracture of vertebral column with spinal cord injury
806.0	Fracture of vertebral column with spinal cord injury; cervical, closed
806.1	Fracture of vertebral column with spinal cord injury; cervical, open
806.2	Fracture of vertebral column with spinal cord injury; thoracic, closed
806.3	Fracture of vertebral column with spinal cord injury; thoracic, open
806.4	Fracture of vertebral column with spinal cord injury; lumbar, closed
806.5	Fracture of vertebral column with spinal cord injury; lumbar, open
806.6	Fracture of vertebral column with spinal cord injury; sacrum and coccyx, closed
806.7	Fracture of vertebral column with spinal cord injury; sacrum and coccyx, open
806.8	Fracture of vertebral column with spinal cord injury; unspecified level, closed
806.9	Fracture of vertebral column with spinal cord injury; unspecified level, open
952	Spinal cord injury without evidence of spinal bone injury
952.0	Spinal cord injury without evidence of spinal bone injury; cervical
952.1	Spinal cord injury without evidence of spinal bone injury; thoracic
952.2	Spinal cord injury without evidence of spinal bone injury; lumbar
952.3	Spinal cord injury without evidence of spinal bone injury; sacral
952.4	Spinal cord injury without evidence of spinal bone injury; cauda equina
952.8	Spinal cord injury without evidence of spinal bone injury; multiple sites of spinal cord
952.9	Spinal cord injury without evidence of spinal bone injury; unspecified site of spinal cord

Table 2.

Continued

ICD-10 codes for tSCI (from 2001/2002 onward)
S14.0	Concussion and oedema of cervical spinal cord
S14.10	Complete lesion of cervical spinal cord
S14.11	Central cord lesion of cervical spinal cord
S14.12	Anterior cord syndrome of cervical spinal cord
S14.13	Posterior cord syndrome of cervical spinal cord
S14.18	Other injuries of cervical spinal cord
S14.19	Unspecified lesion of cervical spinal cord
S24.0	Concussion and oedema of thoracic spinal cord
S24.10	Complete lesion of thoracic spinal cord
S24.11	Central cord lesion of thoracic spinal cord
S24.12	Anterior cord syndrome of thoracic spinal cord
S24.13	Posterior cord syndrome of thoracic spinal cord
S24.18	Other injuries of thoracic spinal cord
S24.19	Unspecified lesion of thoracic spinal cord
S34.0	Concussion and oedema of lumbar spinal cord
S34.10	Complete lesion of lumbar spinal cord
S34.11	Central cord lesion of lumbar spinal cord
S34.12	Anterior cord syndrome of lumbar spinal cord
S34.13	Posterior cord syndrome of lumbar spinal cord
S34.18	Other injuries of lumbar spinal cord
S34.19	Unspecified lesion of lumbar spinal cord
S34.30	Laceration of cauda equina
S34.38	Other and unspecified injury of cauda equina
T06.0	Injuries of brain and cranial nerves with injuries of nerves and spinal cord at neck level
T06.1	Injuries of nerves and spinal cord involving other multiple body regions

At the designated RHSCIR sites in BC and ON, eligible participants are those admitted to acute care or in-patient rehabilitation with evidence of a new impairment of the spinal cord or cauda equina (i.e., motor/sensory deficit, AIS A-D) resulting from the application of an external force of any nature and magnitude (e.g., blunt, penetrating, etc.), including injuries secondary to surgical procedures. In the BC Trauma Registry, patients with tSCI are identified for this project using the Abbreviated Injury Scale (AbIS) seven-digit codes.²¹ Given the differences in how tSCI is coded among the databases, a cross-walk will be created between the ICD-9 and -10 codes and the AbIS 7-digit codes with ISNCSCI as the gold standard.

In BC, data for patients admitted for tSCI will be extracted from the following Population Data BC data sets: DAD, National Ambulatory Care Reporting System (NACRS), Vital Statistics, PharmaNet, PharmaCare, Medical Services Plan (MSP), Mental Health Services, Home and Community Care, and WorkSafeBC. In the BC Trauma Registry, patients with a tSCI will be identified using AbIS seven-digit codes for the same date range and transferred to Population Data BC. Similarly, eligible tSCI participants within VerteBase (1995–2004) and Vancouver RHSCIR (2004–2017) will be identified and transferred by the Praxis Spinal Cord Institute. All data will be managed in accordance with Population Data BC’s privacy and security procedures. Linkage will be done using date of birth and the personal health number. A deidentified patient-level linked data set will be uploaded to Population Data BC’s Secure Research Environment (SRE). An Information Sharing Agreement between Vancouver Coastal Health (on behalf of the Vancouver Spine Database that includes VerteBase and RHSCIR) and the University of British Columbia (on behalf of Population Data BC) was completed. Permission from the respective data stewards for RHSCIR, VerteBase, and BC Trauma Registry was also obtained. A Researcher Services Agreement was also required between Population Data BC and the principal investigator prior to the linkage. Research agreements were required for the Ministry of Health, Vital Statistics, and WorkSafeBC with the principal investigator. Patients will be included in the SCI linked data sets if there is a tSCI diagnosis corresponding to a diagnosis type M (main diagnosis); 1 (pre-admit comorbidity); 3 (secondary diagnosis); W, X, or Y (types of service transfer diagnosis codes) and if they receive an acute level of care for the index tSCI admission. Patients under the age of 18 years will be excluded. The BC Admin SCI Cohort will include data on patients with tSCI from Population Data BC and the BC Trauma Registry. The BC Clinical SCI Cohort will be created by linking the tSCI incident cases from Population Data BC along with the additional data from the BC Trauma Registry to the clinical SCI data from VerteBase and Vancouver RHSCIR (see Table 1); the BC Clinical SCI Cohort is a subset of the larger BC Admin SCI Cohort.

In ON, the administrative data cohort will consist of patients who have a tSCI diagnosis corresponding to a diagnosis type M (main diagnosis); 1 (pre-admit comorbidity); 3 (secondary diagnosis); and W, X, or Y (service transfer diagnosis) in the DAD. Patients under the age of 18 years will be excluded. Incident cases will be linked to the following databases to create the ON Admin SCI Cohort: NACRS, Ontario Health Insurance Plan Claims Database (OHIP), National Rehabilitation Reporting System (NRS) data from adult inpatient rehabilitation facilities, Home Care Database (HCD), Continuing Care Reporting System (CCRS), ICES Physician Database (IPDB), and Registered Persons Database (RPDB). RHSCIR collects data from acute care hospitals and in-patient rehabilitation hospitals, and so a cohort of tSCI patients admitted to in-patient rehabilitation using Rehabilitation Client Group (RCG) codes in the NRS will be created. This cohort with NRS data will be examined with respect to the acute care cohort. The ON Clinical SCI Cohort will be created by linking tSCI incident admissions to RHSCIR data sets from seven hospitals (see Table 1); the ON Clinical SCI Cohort is a subset of the larger ON Admin SCI Cohort. RHSCIR data will be provided to ICES by both the contributing RHSCIR sites and the Praxis Spinal Cord Institute. The contributing RHSCIR sites will provide the RHSCIR ID along with the personal health number, and Praxis will provide the required data along with the RHSCIR ID. Data will be transferred according to ICES’s privacy and security procedures, and an encrypted, deidentified patient-level data set will be available for an ICES analyst to access on servers housed within a closed computing system at ICES-Central on the campus of Sunnybrook Health Sciences Centre in Toronto. Linkage will be done using an encrypted personal health number at ICES. Data-sharing agreements were required between Praxis Spinal Cord Institute and ICES as well as with ICES and each RHSCIR ON site contributing data to the study. Ethical approval and permissions to transfer the data in BC and ON have been obtained and linkage will now proceed.

Variables

Patient and injury variables that will be reported for the Admin SCI Cohorts include age, sex, external causes of injury (traffic collision, fall, struck by/against object or other, etc.), other concomitant injuries, brain injury concurrent with SCI, Charlson Comorbidity Index Score,^22,23 spine surgery (Y/N), SCI level of injury (cervical, thoracic, lumbar, cauda equina), spinal cord severity (compete, incomplete), seasonality of injury (i.e., time of year), rurality based on first 3 digits of the postal code (where a middle digit of 0 depicts rural), health region, and income quintile. Patients included in the Clinical SCI Cohort will also have patient and injury data collected as part of RHSCIR (and VerteBase in BC) such as time of injury, ISNCSCI data, and living setting and employment status in those patients who consent to be part of the RHSCIR community follow-up.

Outcome variables will include number of incident tSCI admissions per year; times and dates to describe transitions from site of injury, acute care, rehabilitation hospitals, and community; acute and rehabilitation length of stay; number and type of in-hospital and postdischarge complications (using diagnosis type 2 codes and subsequent M and 1 diagnosis type code); mortality; number of emergency and hospital re-admissions in 1, 5, and 10 years post injury; number of specialist visits during 1 year after acute care discharge; and mortality (in-hospital, 30 day, 90 day, 1 year, 5 year, life expectancy). Patients included in the Clinical SCI Cohort will also have other outcome data collected as part of RHSCIR (and VerteBase in BC) such as the FIM in rehabilitation.

Data validation

Data will be validated using a number of different steps. A central derived data set will be created for both the BC and ON linked data sets containing commonly used derived variables (e.g., index SCI visit, first ISNCSCI exam, etc.). Venn diagrams will be created to check the reliability of the linked cohort; patients who cannot be linked among the various data sets will be explored. Cohort sizes will be checked against available literature and knowledge from other projects. Derivations of key variables using administrative data such as index SCI visit and indirect versus direct admission to a specialized center will be validated against RHSCIR data. Cross-walks between ICD codes and AbIS 7-digit codes with ISNCSCI will further define SCI severity for cases with only administrative data.

Analyses

Descriptive statistics will be conducted for the patient and injury variables such as age, sex, mechanism of injury, and SCI severity (complete, incomplete; AIS). Study-specific data sets and analysis plans will be created for each research question (see the box “Examples of Research Questions”). Missing data will be managed by substitution from other data sources within the linked cohort where possible or statistical imputation where appropriate using multiple imputation by chained equations (MICE).²⁴

Lessons learned

There are several lessons learned from the progress to date. Many of the databases that will be used in this study are national (i.e., CIHI-DAD); however, provincial differences have presented a challenge in terms of aligning the variables and analyses between ON and BC. Specifically, the NRS is not available in Population Data BC, thus only FIM data from Vancouver RHSCIR is available; whereas in ON, FIM data are collected in the NRS. The BC and ON teams agreed on case definitions for identifying SCI (see Table 2) and measures that can be derived from common data sources (e.g., deriving Charlson Comorbidity Index using DAD). Other challenges to date included the need to align the data extracts to the format of the data variables lists provided to Population Data BC and ICES, which requires an additional derivation step. Finally, each RHSCIR study site has a slightly different approved IRB consent form, and as a result IRB approval was not obtained for all RHSCIR ON sites (see Table 1).

Discussion

This article describes the methodology that will be used to create Canadian provincially linked SCI data sets in BC and ON, including administrative linked data (Admin SCI Cohorts) as well as the clinical and administrative linked data (Clinical SCI Cohorts). Data in BC will be linked over 22 years, and in ON it will cover 8 years. The ethical approval and permissions to transfer the data in BC and ON have been obtained, and linkage will now proceed. To our knowledge, this is the first project that will deterministically link data between SCI clinical registries and provincial data holdings in Canada. The creation of these linked data sets will help to fill gaps in knowledge pertaining to epidemiology, health outcomes, health service utilization, and the economic impact along the SCI continuum.

Previous studies have been conducted in Canada on the tSCI population. Couris et al.⁹ described the incidence, patient variables (e.g., age, gender), and injury factors (e.g., mechanism of injury, level of injury) for patients with a tSCI in ON between 2003 and 2004 and between 2006 and 2007. A similar study was conducted by Dryden et al.²⁵ where administrative data were used to report on the incidence and prevalence of tSCI in Alberta between 1997 to 2000. The estimated incidence of tSCI in ON was reported to be much lower, ranging between 24.2 cases per million in 2003 to 23.1 cases per million in 2006,⁹ compared 44.3 cases per million in Alberta between 1997 to 2000.²⁵ Given the discrepancies between the incidence of tSCI in ON and Alberta, it will be helpful to compare the epidemiology of tSCI in BC and ON using similar methodology over 8 years to identify trends. Furthermore, these linked SCI data sets will enable the life expectancy to be calculated for tSCI in Canada as current models estimating prevalence of tSCI utilize data from the United States.^26,27

Another area of important research that these two SCI linked data sets will support is understanding patient trajectories. The RHSCIR sites are level 1 or 2 trauma hospitals in the provinces, and so this will enable the specialization of tSCI care to be examined in both BC and ON. A systematic review by Parent et al.²⁸ recommended that patients with tSCI should be transferred as soon as possible to a specialized SCI center with an integrated multidisciplinary team, but more research is needed to better understand the impact on outcomes. Access to specialized rehabilitation in Canadian RHSCIR Rehab sites was reported to increase the probability of returning home compared to long-term care, with an adjusted odds ratio (OR) = 3.1 (95% CI, 1.6–5.9).²⁹ However, these studies were limited as they only included patients admitted to RHSCIR sites (i.e., sites with specialized SCI care). By analyzing the SCI linked data sets, it will be possible to compare patient trajectories and outcomes, which can inform how patients are triaged and treated. It will be possible to explore how patient and injury factors as well as access to specialized SCI care impact long-term health care utilization. Using administrative data, Guilcher et al.³⁰ reported that 45.3% of patients visited the emergency department in the first year and the percentage remained high (34.5%) up to 6 years post injury. It was not possible, however, to determine the effect of the injury severity or access to specialized care on emergency department usage; information about both of these factors is needed to create provincial patient care pathways to optimize SCI care.

Having access to these SCI linked data sets will provide an opportunity to work toward developing learning health systems for tSCI in BC and ON. Having data on the patient journey that links initial care with utilization patterns in the community will be a valuable resource for SCI organizations, clinicians, researchers, program managers, and policy makers. Given the universal health coverage in Canada, the findings from this research in BC and ON can be generalized across Canada. This type of population-based research may be difficult to conduct in the United States due to the presence of both private and public health insurance in a hybrid system. Limitations using the linked data sets include being restricted to a specified date range and only including patient-reported outcomes and clinical data on participants in RHSCIR. Because these data sets and research questions are approved under a research model, it will also be difficult to answer emerging questions from the SCI community.

A number of obstacles have been encountered in preparing the data. The process was complex and required coordination from the principal investigator’s institution and Praxis Spinal Cord Institute, as well as from each of the participating RHSCIR sites and with Population Data BC and ICES. Contract experts were consulted at participating institutions to clarify which agreements were necessary. Considering all the steps and time needed to create these linked SCI data sets, a mandated data collection for tSCI in Canada must be considered as part of a broader data strategy to advance SCI care.¹⁴ This top-down approach would facilitate the collection of clinical data, health services administrative data, and patient-reported outcomes to assist policy makers in ensuring that the health care system produces optimal patient outcomes. This would increase efficiencies in data collection and interpretation and expand the use of artificial intelligence methods to leverage existing SCI data. Furthermore, there would be the opportunity to bring together rich data sources such as biomarkers (e.g., imaging, genomic data) and clinical, patient-reported, and administrative data that would facilitate personalized care in tSCI.

In conclusion, this article describes the creation of linked data sets for tSCI in BC and ON. The methodology outlined may inform other groups doing similar research. The project is unique because it includes SCI clinical registry data with provincial administrative data to describe the epidemiology and care across the SCI continuum. Findings from this study will determine the value of existing SCI data as well as future data requirements to support a learning health system designed to optimize care.