Introduction

Different incidences of traumatic spine fractures (TSF) with or without spinal cord injury (SCI) have been reported in different geographic regions [1,2,3,4,5]. The incidence of traumatic SCI (TSCI) varies between 3.6 and 195.4 per one million around the world; in developing countries, an incidence of 25.5 SCI per million has been reported [6, 7]. In Iran, Heidari et al. reported TSF in 3.8% (619 out of 16,321) of trauma admissions between 1999 and 2004 using the National Trauma Registry (NTR) data [8]; 5.8% (36 out of 619) of people with TSF admitted to a hospital had an associated SCI. In another study, the incidence of TSF in Tehran was estimated to be about 16.35 per 100,000 and less than half of these patients had a simultaneous SCI [9]. The incidence of TSCI in Iran was estimated as 9 per 100,000 persons by another study [10].

When the spinal cord is damaged, long-term disability often ensues [10]. TSCI is a condition with multiple comorbidities and considerable health, social, and economic impacts. The burden of these impacts can be managed by improving quality of care. The lack of a reliable source of information about spine trauma with or without SCI is a significant barrier in evaluation and planning of systems of prevention, control, and care [11]. Until now, there was no spine-specific trauma database or registry system in Iran. The National SCI Registry of Iran (NSCIR-IR) is a project supported by the Ministry of Health and Medical Education to collect and provide TSF and SCI data to evaluate and improve the quality of care of persons with SCI [12].

The development of a data set is an essential phase of designing a registry system [13]. According to Kowal et al., a data set is a common set of data elements used to collect and report data in the registry [14]. Many studies have emphasized the importance of a data set to create national databases and support information sharing on diseases, injuries, and other health-related problems [14,15,16]. One of the most important and effective tasks for the success or failure of a registry system is the selection of appropriate data items [17]. Data set development is a critical step for nonepidemiologic registries on traumatic conditions that can lead to long-term disabilities such as TSF/SCI. The objectives for a SCI registry are focused on care and outcome measurements, requiring various data inputs from different independent and distinct resources. This study focuses on the development process of the data set for the NSCIR-IR and delineates our experiences, which might be useful for future registries, specifically in developing countries with limited budgets, resources, and information technology infrastructures.

Methods

The development of NSCIR-IR has been one of the main priorities of the National Iranian SCI Research Network since 2012. The data set of NSCIR-IR was developed between March and October 2015. An expert panel with three neurosurgeons, one physiatrist, one general practitioner, one nurse, one community medicine physician, one emergency physician, two epidemiologists, and one health information management specialist was formed. We followed a stepwise approach in the development of the data set. The steps included (i) review of the data sets in similar SCI registries, (ii) selection of a data set as the base framework of the NSCIR-IR data set by the expert panel, and (iii) modification of the basic framework according to data needs, registry objectives, and feasibility of data collection.

First, a simple review was performed on the data sets of three SCI registries of Australia (Australian Spinal Cord Register), Canada (Rick Hansen SCI Registry (RHSCIR)), and Europe (European Multicenter Study about SCI).

Later, the RHSCIR data set was selected as our basic framework to develop an initial draft data set. The selections were made based on the similarity of RHSCIR to our registry in registry type (Spinal column and SCI (TSF/SCI)) inclusion criteria. Sixteen sessions of the expert panel were held between March and October 2015. We translated the RHSCIR data set dictionary into Farsi and then our expert panel independently reviewed the translation. The accessibility of the data elements of the RHSCIR was examined by experts who were aware of the clinical documentation status and the general content of medical records in hospitals of Iran. It was noted that several variables are not recorded in patients’ hospital charts; therefore, patient medical records were considered as the secondary data source of the registry. Comparability of our registry data with other countries was another issue that was discussed by the expert panel. International data standards are recommended by the World Health Organization (WHO), therefore, the International SCI Data Sets (ISCIDS) were also reviewed. At the third step, the expert panel made changes to Iranian data set of NSCIR-IR in order to accommodate our major data elements and concerns of quality of care. For example, our data set registers different time points from injury time to the time of arrival of ambulance to the trauma scene, time of arrival to hospital, or to a referral center, and time of the surgical stabilization with or without spinal cord decompression. With this design, it will be possible to identify major time delay points in our care providing system.

Results

Table 1 compares the RHSCIR data set with the two other registries [18, 19]. The RHSCIR data set includes more than 307 variables that cover quality of care [18, 20]. The RHSCIR data set has two types of minimal and extended data elements in which the minimal data set includes 200 mandatory data elements that must be recorded for all patients including those who do not consent to be part of the registry. These minimal data set includes name, family name, age, time, place and mechanism of injury, and questions which are essential for patients’ treatment by care providers. The extended data elements (107 elements) are recorded only for patients who consent to be part of the registry [21]. Our expert panel reviewed the RHSCIR data elements. Some of the RHSCIR data elements were not accessible through primary data sources of the registry e.g., patient-reported or clinician-reported data, observation and measurement by the registry members. Medical records were considered as the secondary data source of the registry, however, several variables in the RHSCIR data set were also not recorded in the patient medical records. Table 2 shows the data items and variables, which we did not include in our registry (Table 2).

Table 1 Compression of data set in three SCI similar registries [19, 20, 35,36,37].
Full size table
Table 2 The data items and variables which we did not include in National Spinal Cord Injury Registry of Iran (NSCIR-IR).
Full size table

According to the review of WHO and ISCIDS [22,23,24], our strategic committee had considerations regarding comparability of the designed data set with international data. WHO recommends the use of standardized coding systems such as the international classification of diseases (ICD), International Standards for Neurological Classification of SCI and ISCIDS for data collection to facilitate comparison between different patients, centers, and countries [22]. The RHSCIR data set and other data sets (Appendix 1) were different from the ISCIDS. Therefore, the expert panel decided was not convinced to use the RHSCIR data set due to concerns of accessibility and comparability of data in our registry with the RHSCIR data set. The expert panel concluded to use ISCIDS as the basic framework of NSCIR-IR data set. ISCIDS includes a core data set and several specialized data sets. The core data set includes 24 variables that are the minimal data required for collection for all people with SCI in acute inpatient settings [25]. There are 12 more data sets which are focused on different aspects of SCI and SCI consequences [24]. According to the registry objectives and registration process, the expert panel selected 174 data items from ISCIDS. Data selections were based on accessibility of data and usefulness in quality of care assessment in the acute settings. Selected data items were organized into five sections based on the logical sequence in the process of care (Table 3).

Table 3 Initial National Spinal Cord Injury Registry of Iran (NSCIR-IR) data set [25, 41,42,43,44,45,46,47,48,49,50,51,52].
Full size table

After translation of data items into Farsi, case report forms were designed. Panel members evaluated designed forms and applied the required methods. Major and minor modifications were made to overcome the mentioned challenges in Box 1 (Table 4).

Table 4 Lists of major and minor modifications on initial National Spinal Cord Injury Registry of Iran (NSCIR-IR) data set.
Full size table

The final version of the NSCIR-IR data set was developed with 350 data elements. It includes one data set for acute and one data set for chronic phase of injury (rehospitalization data). Our data items are two of types: mandatory and conditional. Conditional data items are data items that are considered as mandatory based on the clinical situation of the patient. If a data item is not applicable for a patient, it is not necessary to be completed for that patient. For example, if the accident was not a traffic accident, then it is not necessary to record the accident data. Table 5 shows the components of the final version of NSCIR-IR data set. The items were organized into seven case report forms for acute phase and four case report forms for the chronic phase. All were used in the pilot phase of registry implementation (Table 5). After 8 months of implementation of the pilot phase in three trauma hospitals, changes were made to the data set that are (i) details on the types of respiratory, cardiovascular, and other comorbidities were added into admission form, (ii) procedures in the first hospital in cases with inter-hospital patient transfer were added into admission form, (iii) injury type of Atlas and Axis cervical vertebra (C1-C2), were separated from the other vertebrae and were added as a question into the injury form, (iv) one item was added into the intervention form for the vertebra or vertebrae on which surgery was performed, and (v) one item was added into the discharge form for the type of external fixation device, which patient uses at the time of hospital discharge.

Table 5 Components of National Spinal Cord Injury Registry of Iran (NSCIR-IR) data set.
Full size table

Discussion

Our study presents an overview of the data set development process for the NSCIR-IR. Our evidence-based approach along with interdisciplinary expert panel review helped us to develop a comprehensive, yet applicable, data set. Some groups have used review of current evidence and formation of working groups to select the registry data elements [26, 27]. Others have conducted either a survey or Delphi method on an initial version of the data set [28,29,30]. Svensson-Ranallo et al. [31] performed a comprehensive review for methodologies of data set development in health care. The results of this review showed that there are different methods such as survey, systematic review, chart review, and Delphi technique. The most common methods are the use of experts and stakeholders, but the authors suggested a combined approach with literature review, chart review, expert committee, and organized data for developing the minimum data set [31]. In our study, we used the literature review and expert committee to form the data sets.

It is important to choose the right data elements for a registry. Tee et al. stated that correct data selection items of a registry lead to a balance between comprehensiveness and practicality of the registry [32]. Although, at first, we wanted to have a comprehensive and detailed data set, which meets all present and future data requirements for research and care evaluation, the unavailability of most data items and the necessity of an international framework for comparability of our registry led us toward using the ISCIDS.

Compliance evaluation of determined data elements with routine clinical practice was effective to reduce registration workload. Imposition of required uncommon practice in current clinical practice increases the cost. Tee et al. emphasized the number of data elements collected that should be practical based on routine process [32]. For example, when we chose the RHSCIR data set as our framework, we found that some of the data were not accessible at all through primary data sources of the registry. For example, we found the hesitancy of our patients in providing data related to their income, work status, drinking, and some other measures as a big barrier for data collection. In other words, assessing income is very difficult in Iran especially in people who are self-employed. Most self-employed Iranians do not express their real income for cultural reasons. Usually, government employees have specific income data.

In addition, assessments and interventions for patients in clinical practice differ between Iran and Canada. For instance, height and weight of patients are not taken on arrival to the hospital. Also, some clinical assessments such as mobility or pain assessment with specialized tools (e.g., Berg Balance Scale, Leeds Assessment of Neuropathic Symptoms and Signs, Douleur Neuropathique 4) are not part of the routine assessments in Iran. Also, rehabilitation and assistive devices are not provided in a specialized manner to the patients in the acute phase. Instead, rehabilitation and support services are available at the level of long-term care. Although, there is connection between the acute care system and long-term care system; there is a lack of communication and information sharing between care providers working in the two levels of Iran health care system. Patients normally work as messengers, carrying information between the two levels. In contrast, Canada has a robust information technology infrastructure. There are multiple database and reporting systems at all of the health care services levels. These include the Hospital Morbidity Database and Discharge Abstract Database for ambulatory and hospital care, National Rehabilitation Reporting System for rehabilitation care, NTR for trauma care, Continuing Care Reporting System, and Home Care Reporting System for continuing and long-term care [33,34,35,36,37,38]. Using primary data sources is not practical in Iran due to limited human resources for registration and funds allocated according to the large scale of data that led to a time-consuming and tedious registration process. Therefore, it was predicted that using the Canadian SCI registry as the backbone of data set in NSCIR-IR is impractical.

The NSCIR-IR data set contains spine trauma data such as the injury morphology and injury type according to the AOSPINE injury classification system. The classification is used only in the RHSCIR and not used by other TSF/SCI registries. Therefore, it is a strong point of the NSCIR-IR data set. In relation to patient outcome, the data set includes the Glasgow Outcome Scale that is a physician-reported outcome; however, it does not include any patient-reported outcome data, specifically about patient functional independence. Other SCI registries use standard measures for patient-reported outcome such as the Short Form 12 or 36 Health Survey, the Spinal Cord Independence Measure, or the Functional Independence Measure [21, 39, 40]. This is a weak point of the NSCIR-IR data set.

Using the ISCIDS as the basic framework of the NSCIR-IR data set facilitates international comparisons on the quality of provided care to TSF/SCI patients. One of the limitations of our study was that we chose variables based on the expert’s opinion on the availability of data in current medical records. However, due to lack of electronic health records and national or regional standards in medical records documentations, expert opinion was the best resource that we could rely on.

In summary, the NSCIR-IR data set was developed to meet the quality objectives of the registry. It focuses on data representing quality of care provided to patients with TSCI in the acute and chronic phases of their injury. The selected basic framework of the data set for the ISCIDS can help to compare national data with data from other countries. Expert panel modifications facilitate the integration of the registration process with the current clinical workflow within the hospitals.