Organization quality systems and department-level strategies: refinement of the Deepening our Understanding in Quality in Australia (DUQuA) organization and department-level scales

Abstract

Objective

The aim of this study was to develop and refine indices to measure organization and care pathway-level quality management systems in Australian hospitals.

Design

A questionnaire survey and audit tools were derived from instruments validated as part of the Deepening Our Understanding of Quality improvement in Europe (DUQuE) study, adapted for Australian hospitals through expert opinion. Statistical processes were used to explore the factor structure, reliability and non-redundancy and descriptive statistics of the scales.

Setting

Thirty-two large Australian public hospitals.

Participants

Audit of quality management processes at organization-level and care pathway processes at department level for three patient conditions (acute myocardial infarction (AMI), hip fracture and stroke) and senior quality manager, at each of the 32 participating hospitals.

Main Outcome Measure(s)

The degree of quality management evident at organization and care pathway levels.

Results

Analysis yielded seven quality systems and strategies scales. The three hospital-level measures were: the Quality Management Systems Index (QMSI), the Quality Management Compliance Index (QMCI) and the Clinical Quality Implementation Index (CQII). The four department-level measures were: Specialised Expertise and Responsibility (SER), Evidence-Based Organisation of Pathways (EBOP), Patient Safety Strategies (PSS) and Clinical Review (CR). For QMCI, and for seven out of eight subscales in QMSI, adequate internal consistency (Cronbach’s |$\alpha$| >0.8) was achieved. For CQII, lack of variation and ceiling effects in the data resulted in very low internal consistency scores, but items were retained for theoretical reasons. Internal consistency was high for CR (Cronbach’s |$\alpha$| 0.74–0.88 across the three conditions), and this was supported by all item-total correlations exceeding the desired threshold. For EBOP, Cronbach’s |$\alpha$| was acceptable for hip fracture (0.80) and stroke (0.76), but only moderate for AMI (0.52). PSS and SER scales were retained for theoretical reasons, although internal consistencies were only moderate (SER) to poor (PSS).

Conclusions

The Deepening our Understanding of Quality in Australia (DUQuA) organization and department scales can be used by Australian hospital managers to assess and measure improvement in quality management at organization and department levels within their hospitals and are readily modifiable for other health systems depending on their needs.

Introduction

The ‘Deepening our Understanding of Quality in Australia’ (DUQuA) study [1] was a 5-year Australia-wide, multi-level, multi-million dollar cross-sectional study aiming to identify how quality management systems (QMS), leadership and culture in Australian hospitals are related to care delivery and patient outcomes for acute myocardial infarction (AMI), stroke and hip fracture. Based on the original, landmark European Union-funded ‘Deepening our Understanding of Quality improvement in Europe’ (DUQuE) study [2], the rationale for undertaking DUQuA included potential for comparison with the large European sample of 188 hospitals across seven countries, in addition to in-depth understanding of quality management in Australian acute settings. Evidence- or consensus-based measurement tools were designed or modified and then utilized to collect quantitative data on QMS at hospital and care pathway levels, department-level safety culture and leadership amongst clinicians, clinical treatment processes, patient outcomes and patient perceptions of safety. Collection methods included paper-based and electronic surveys, medical record reviews, external audits and accessing national datasets from the National Stroke Foundation registry and the Australian Institute of Health and Welfare. Linear and multi-level modelling was applied to the datasets to identify relationships between quality management, care delivery and patient outcomes. Findings have the potential to influence decision-making and improvements in quality and safety in Australian and international hospitals.

The aim of this study was to develop and validate indices to measure organization and care pathway-level QMS in Australian hospitals. The DUQuA measurement tools were derived from the DUQuE tools and, where possible, the DUQuE terminology retained. For example, the definition of QMS mirrors that of DUQuE: ‘a set of interacting activities, methods and procedures used to monitor, control and improve the quality of care’ [3]. The organization and department scales described are the same measures used by DUQuE for organization-level QMS, modified for the Australian context as described later in our ‘Method’ section. The three organization-level measures are the Quality Management Systems Index (QMSI), the Quality Management Compliance Index (QMCI) and the Clinical Quality Implementation Index (CQII). The four measures of quality management activities at department level focus on Specialised Expertise and Responsibility (SER), Evidence-Based Organisation of Pathways (EBOP), Patient Safety Strategies (PSS) and Clinical Review (CR).

This article presents information about the structure, reliability and non-redundancy and descriptive statistics of the QMSI, QMCI and CQII organization scales and the SER, EBOP, PSS and CR department scales. Relations between the organization and department scales and other outcomes are reported elsewhere in this Supplement [4].

Method

Development of the organization scales

The three organization-level measures from DUQuE (QMSI, QMCI and CQII) were modified by the DUQuA team to reflect the evolution of our condition-specific indicators and to ensure relevance to the Australian context, in consultation with national quality assessment experts, such as the Australian Council on Healthcare Standards. QMSI quantifies the managerial aspects of quality management that might influence the implementation of quality systems in hospitals. QMSI consists of 10 subscales: quality policy documents, hospital governance board activities, quality resources, quality management, evidence-based medicine protocols, preventive protocols, internal quality methods for general activities, personnel, clinical practice and patients. Each item is rated on a four-point Likert scale (range 1–4).

QMCI measures the managerial aspects of quality improvement in hospitals on three subscales: quality planning, monitoring patient and professional opinions and quality control and monitoring. CQII quantifies implementation of quality systems at hospital level including whether systems exist, to what extent implementation has been monitored and whether implementation is sustainable. CQII consisted of seven subscales: preventing and controlling healthcare-associated infections, medical safety, preventing patient falls, preventing pressure injuries, routine assessment and diagnostic testing of patients in elective surgery, safe surgery that includes an approved checklist and recognizing and responding to clinical deterioration in acute healthcare. For QMCI and CQII, items are rated on a five-point Likert scale (range 0–4).

The final DUQuA instruments are similar, but not identical, to the tools used in the DUQuE study. For example: some questions in the QMSI survey are worded differently to align with Australian terminology and processes, and a small number of additional questions have been included.

Development of the department scales

Following a similar process, the DUQuA team modified the DUQuE measures [5] (SER, EBOP, PSS and CR) for the Australian context. SER explores how clinical responsibilities were assigned for a particular condition; EBOP measures whether department processes, such as admission, acute care, rehabilitation and discharge, were organized to facilitate evidence-based care recommendations; PSS measures the use of clinical practice guidelines, and CR assesses the integration of audit and systematic monitoring with department quality management mechanisms. SER, CR and PSS used identical measures for the three conditions, while EBOP has the same structure for each condition-specific department (AMI, stroke and hip fracture) but the content varies to reflect the condition-specific evidence recommendation.

There are differences between the DUQuE and DUQuA department scales. As in the QMSI survey, some audit items are worded differently to align with Australian terminology and processes. Unlike DUQuE, the DUQuA project recognizes the Emergency Department (ED) as a common entry point to the hospital for the majority of AMI, hip fracture and stroke patients. ED condition-specific measures were developed for SER, EBOP and CR, to ensure relevance to the ED context; thus, for example, the ED assessment of SER comprised questions for each of AMI, hip fracture and stroke. For ED, the PSS items are all generic (i.e. not condition-specific).

Setting, participant recruitment and data collection

The DUQuA project commenced in February 2014 in 32 large Australian public hospitals, and data collection was completed by November 2017. In each state and territory, general public hospitals that met the following criteria were recruited: (i) with approximately 200 beds or more, (ii) with an ED and (iii) that regularly admit more than 30 each of stroke, AMI and hip fracture patients over a period of 3 months. A group of 78 hospitals were invited to participate, 70 that met all inclusion criteria and eight that were close to the specified thresholds or lacked at least one of the chosen departments; 62 hospitals initially agreed (89%). Due to factors associated with healthcare leadership changes, hospital relocations and obtaining ethics approval, a number of hospitals withdrew; however, 32 hospitals participated until completion of the study. The QMSI questionnaire was assessed by the senior person responsible for the coordination of quality improvement activities in each of the 32 participating hospitals.

Hospitals responding to an initial research invitation were provided with a formal letter detailing the study’s background, data collection procedure and timeline. A Local Principal Investigator (LPI) was nominated by each hospital as the study point of contact. The role of the LPI was to maintain collaborative relationships between the hospital staff and the research team, to contact the quality manager to complete the QMSI questionnaire and to coordinate a site visit for accredited surveyors to conduct an external quality assessment (EQA) visit, during which the two organization-level measures (QMCI and CQII) and the four department-level measures (SER, EBOP, PSS and CR) were assessed. Details about surveyors’ visits are reported elsewhere in this Supplement [4, Appendix A]. Questionnaires were completed anonymously, and EQA was conducted as a two-day site visit in the hospital, including the three condition-specific wards (AMI, hip fracture and stroke) and the ED. Data collection commenced in February 2016 and concluded in November 2017.

Statistical analysis

DUQuE applied the following rules in assessing reliability: Cronbach’s |$\alpha$| >0.7 [6] and item-total correlation coefficient >0.4 [6] were considered acceptable evidence of internal consistency and consistency with the subscale construct, respectively, and Pearson’s correlation coefficient <0.7 between subscale scores was accepted as demonstrating non-redundancy of a subscale [5, 6]. DUQuA adopted the same rules, except that Cronbach’s |$\alpha$| >0.8 was deemed preferable, based on the work of Nunnally [7]. These thresholds were used to indicate a need for consideration of model re-specification for the three organization-level measures (QMSI, QMCI and CQII). The four department-level measures are single factor scales, so model re-specification was not considered, but we nevertheless report the test-values for internal reliability to inform interpretation of the measures. Further details on statistical analysis methods used for DUQuA have been published elsewhere in this Supplement [8]. Confirmatory Factor Analysis was not feasible because of the small sample size.

Data were analysed in SAS/STAT 9.4 (Cary, North Carolina, USA). Characteristics of the hospitals were summarized. We used a combination of theoretical grounds, item-total correlation and Cronbach’s |$\alpha$|⁠, to refine the theoretical model.

After reaching the final model structure, each organization-level subscale was calculated as the mean of all non-missing items. If more than 50% of the items within the subscale were missing, the subscale was set to missing. We performed 100 multiple imputations, using Markov Chain Monte Carlo (MCMC) method [9], to impute missing organization subscales. After imputation, we calculated the organization scales by summing all subscales (for QMSI, eight was subtracted from the sum, in line with DUQuE procedures). To assess the degree of redundancy, Pearson’s correlation coefficient was calculated between pairs of subscales within each scale; the analysis was repeated 100 times for each imputation and the SAS/STAT MIANALYZE procedure [10] was used to obtain the pooled parameter estimates and standard errors, if any imputed values were involved in the calculations. For department-level and ED-level scales, each department-level scale was the mean of all applicable non-missing items. If more than 50% of the applicable items within the scale were missing, the scale was set to missing.

Results

Characteristics of participants

Although 32 hospitals participated in DUQuA, two hospitals shared a quality manager and QMS, and therefore, 31 QMS questionnaire surveys were included for analysis. All 32 hospital provided audit data. The characteristics of participating hospitals are summarized in Supplementary eTable A1, Appendix A.

Organization scales—structure, reliability and non-redundancy

Missing data

For QMSI, there were no missing data at subscale level. For QMCI, the percentage of missing subscales was 1.6% and ranged from 0 to 3.1% per subscale, affecting one hospital only. For CQII, the percentage of missing subscales was 0.4% and ranged from 0 to 3.1% per subscale; again, only one hospital was affected.

Descriptive statistics for QMSI, QMCI and CQII subscales

The descriptive statistics for each subscale of each of the three organization-level measures are summarized in Table 1. Subscales for all three measures were left-skewed with all but one scale having a mean value above 3.0. For QMSI, the subscale mean ranged from 3.2 to 3.7; for QMCI, both subscales had a mean of 3.5, and for CQII ‘Routine assessment and diagnostic testing of patients in elective surgery’ had a mean of 2.3 with the remaining subscale means ranging from 3.3 to 3.9. Item-level summary statistics for each measure can be found at Supplementary eTables A2–4, Appendix A.

QMSI—structure, reliability and non-redundancy

From the original 10-factor theoretical model for QMSI, we merged item Q7.1 to 7.6 from the ‘evidence-based medicine protocols’ construct with item Q7.7 to 7.13 from the ‘preventive protocols’ construct to form a 13-item ‘preventive protocols’ construct. Items Q8.1 to 8.12 originally forming three ‘internal quality methods’ constructs (general activities, personnel and clinical practice) were restructured to ‘internal quality methods—general’ and ‘performance monitoring’. The final eight-factor model is presented in Table 2, along with the Cronbach’s |$\alpha$| scores and item-total correlations. All subscales except ‘internal quality methods—patients’ achieved or nearly achieved internal consistency reliability (Cronbach’s |$\alpha$| >0.8). Twelve out of 51 items did not meet the desired threshold (>0.4) for item-total correlation, to demonstrate internal consistency within the subscale construct. Appendix A, Supplementary eTable A5, shows between subscale correlations for QMSI. Twenty-five pairs of correlations were within (0.16–0.69) the acceptable range of <0.7; the three scales with potential redundancy were ‘Quality resources’ (0.75), ‘Quality management’ (0.74) and ‘Internal quality methods—general activities’ (0.77), each of which correlated with ‘Hospital governance—board activities’.

QMCI—structure, reliability and non-redundancy

From the original three-factor theoretical model for QMCI, we moved two items into the ‘quality control and monitoring’ subscale: item XQ01 from the single item ‘quality planning’ subscale and XQ08 from ‘monitoring patient and professional opinions’. The final two-factor model can be found in Table 3. Both subscales showed adequate internal consistency through Cronbach’s |$\alpha$| without reaching our preferred threshold of 0.8. One item in each subscale had an item-total correlation <0.4, indicating less than ideal fit with the subscale construct. The between subscale correlation coefficient for QMCI (0.73) indicated non-redundancy.

CQII—structure, reliability and non-redundancy

We retained the original seven-factor theoretical model for CQII. The model can be found in Table 4 along with the Cronbach’s |$\alpha$| and item-total correlations. Only three factors (‘preventing pressure injuries’, ‘routine assessment and diagnostic testing of patients in elective surgery’, and ‘Safe surgery that includes an approved checklist’) achieved internal consistency through Cronbach’s |$\alpha$| and item-total correlations. The other four subscales and their constituent items failed to meet preferred thresholds. We explored removing items to improve Cronbach’s |$\alpha$|⁠, but were unable to significantly improve results. This was not surprising, as the majority of hospitals scored the maximum plausible value of four for these items (the mean of each of these four subscales was 3.9). We decided to retain these items for theoretical reasons (presented in the Discussion); however, the limited range in scale scores meant that numerous item-total correlations were not calculable. Supplementary eTable A6 provides the intra-scale correlation coefficients for CQII. The correlation coefficient between the ‘Medication safety’ and ‘Recognizing and responding to clinical deterioration in acute healthcare’ subscales was 0.88, the only subscale pairing indicating potential redundancy.

Department scales—structure, reliability and non-redundancy

Missing data

There were 119 participating departments (27 AMI, 29 hip fracture, 32 stroke and 32 ED). For EBOP and CR, there were no missing data; for SER, missing items were restricted to hip fracture departments, with 1.1% of hip fracture items missing (ranging from 0 to 3.5% per item); and for PSS, the percentage of missing items was 0.4% (ranging from 0 to 3.8%) per item in AMI and 0.4% (ranging from 0 to 3.5%) per item in hip fracture. There were no departments with a scale missing.

Descriptive statistics for SER, EBOP, PSS and CR

Descriptive statistics for SER, EBOP, PSS and CR are summarized in Table 5 and for their items in Supplementary eTable A7. Items were left-skewed with all means calculated at two or above (in a 0–4 plausible range). Consistently across the three departments, PSS had the highest (3.4–3.6) and CR had the lowest (2.2–2.9) mean scores.

Descriptive statistics for ED-level scales

Descriptive statistics for ED-level SER, EBOP, PSS and CR are summarized in Table 6 and for their items in Supplementary eTable A8. Items were left-skewed with ceiling effect. The PSS, which was assessed using generic rather than condition-specific questions, had a mean score of 3.4. Across all three conditions, EBOP consistently had the highest mean scores (2.5–3.6) and CR had the lowest mean score (1.3–2.0); mean scores for hip fracture were lower than for either AMI or stroke.

SER, EBOP, PSS, CR—structure, reliability and non-redundancy

Department-level SER, EBOP, PSS and CR. The models are shown in Table 7 along with the Cronbach’s |$\alpha$| and item-total correlations. Internal consistency reliability through Cronbach’s |$\alpha$| was high for CR (0.74–0.88 across the three conditions), and this was supported by all item-total correlations exceeding the desired threshold. For EBOP, Cronbach’s |$\alpha$| was acceptable for hip fracture (0.80) and stroke (0.76), but only moderate for AMI (0.52); as suggested by these results, item-total correlations were poor for one item in stroke (−0.16) and two items in AMI (−0.06 and 0.19). For PSS, Cronbach’s |$\alpha$| suggested its constituents were effectively unrelated (−0.12 to 0.17), with virtually all item-total correlations poor. For SER, Cronbach’s |$\alpha$| showed a moderate association amongst its constituent items (0.32–0.52), again with relatively few item-total correlations above 0.40.

While exploratory factor analysis was used to reduce and determine which items would be aggregated to build a scale for SER and CR, the items comprising EBOP and PSS were determined based on theoretical importance and background knowledge. It was not possible to build one generic scale for the EBOP, because of the different items across pathways. The other scales developed in this analysis used the same items to compute scores for each pathway. Despite the same items being used across pathways for the quality measure PSS, no generic scale for the four pathways was revealed after factor analysis.

The final models for ED measures are shown in Table 8 along with the Cronbach’s |$\alpha$| scores and item-total correlations. PSS in the ED, assessed through a single set of generic questions, had only a moderate Cronbach’s |$\alpha$| (0.44), but was clearly stronger than in the condition-specific departments. For all three conditions, CR in the ED was again the most reliable scale in terms of both Cronbach’s |$\alpha$| (0.80–0.93) and item-total correlation (all items). EBOP in the ED showed a similar pattern to the condition-specific results, with an acceptable Cronbach’s |$\alpha$| for hip fracture (0.80) and stroke (0.78), but only a moderate score for AMI (0.46). The two SER items relating to AMI in the ED were essentially unrelated (Cronbach’s |$\alpha$| of 0.02) and moderate (0.40–0.41) for the other conditions.

Discussion

Interpretation of results

This study aimed to refine and validate, in the context of the Australian healthcare system, three scales for measuring quality improvement at organization level (QMSI, QMCI and CQII), and four scales for measuring quality improvement at hospital care pathway-level for AMI, hip fracture and stroke conditions (SER, EBOP, PSS and CR).

The final QMSI scale consists of eight subscales, measuring: quality policy, hospital governance board activities, quality resources, quality management, preventive protocols, internal quality methods, general activities and performance monitoring. The final QMCI consists of two subscales, measuring: monitoring patient and professional opinions and quality control and monitoring. The final CQII consists of seven subscales, measuring: preventing and controlling healthcare-associated infections, medical safety, preventing patient falls, preventing pressure injuries, routine assessment and diagnostic testing of patients in elective surgery, safe surgery that includes an approved checklist and recognizing and responding to clinical deterioration in acute healthcare. In addition to the differences with the DUQuE scales noted in the introduction, some items are grouped within different subscales following subscale analysis.

The final CQII scale could not be validated due to the scores clustering at the high end of the scale. This is likely a consequence of mandatory accreditation requirements for Australian hospitals, as there is a strong alignment between the topics addressed by the CQII subscales and accreditation measures. Nevertheless, we believe it is important to retain the CQII, as the performance assessed has been shown via research to be important for patient safety [9, 10], and the scale needs to allow for the prospect that not all future hospitals will be as high performers as those in DUQuA.

Within each scale, we retained some subscales that had Cronbach’s |$\alpha$| lower than the desired 0.8, where those subscales measured aspects of quality that were considered to be a critical component of hospital quality management, or where the lower correlation could be explained by understanding how hospital care was structured. For example, in the QMSI scale, all subscales but the ‘internal quality methods—patients’ achieved (or nearly achieved) internal consistency reliability using the cut-off value. This subscale consisted of ratings of whether patient satisfaction or experience is measured and monitored at least annually, and whether periodical evaluation of patient complaints is used to drive improvements. The DUQuE QMSI scale found a similar result for this subscale [11], and was retained for both studies, recognizing the importance of the patient voice in hospital care. There are good reasons to keep items and scales within the measurement instruments because they are theoretically important, even if they did not fulfil the high statistical standards set before the analysis. While this may seem to undermine the importance of these criteria and the methods used, it is of practical relevance for the final scales to be able to measure an adequate range of elements that make up quality management and quality improvement activities in hospitals.

In many scales and subscales we also had some items below our acceptable correlation cut-off of 0.4. Similarly, we retained items that had lower correlation with other items within the same factor where those items were considered to be an important part of hospital activity. For example, within the QMSI subscale of ‘preventative protocols’, removing the item rating prevention of ventilator-associated pneumonia would have improved Cronbach’s |$\alpha$| from 0.797 to >0.8, but we resisted this because patients in Australian hospital are normally only ventilated in the Intensive Care Unit, during surgery, and—occasionally—in the ED, potentially explaining why it did not correlate well with other activities that are performed more broadly across the hospital. In another example within the same subscale, we retained the item ‘an up-to-date hospital protocol for recognizing and responding to clinical deterioration in acute healthcare’ due to its importance in patient care, despite an item-total correlation of only 0.137. This is an element of care that has received strong endorsement from State health departments over the last decade [12, 13], and the data were strongly left skewed due to high performance in this aspect of care in most participating hospitals.

For department-level care pathway scales, the final scales were not modified as a result of our analysis. We developed a generic scale for PSS, but it was not possible to build generic scales for SER, EBOP and CR, as the processes are different depending on the specific condition and the department; for this reason, we also developed condition-specific versions of each of these scales for use in the ED.

Limitations

While most of the scales are based on independently audited data, the QMSI scale is based on perceptions of the quality manager. In mitigation, we worded the questions in the quality manager survey to include only questions on facts, to minimize bias associated with self-report. Additionally, while we have shown how the organization and department scales have been modified from their European antecedents, the DUQuA scales have only been validated for the Australian context, and their generalizability to other settings remains to be demonstrated. Differences between the DUQuA and DUQuE scales must be taken into account when comparing findings from Australian and European hospitals.

Implications for research, policy and practice

It can be difficult for hospitals to access validated measurement tools for assessing quality management that are powerful yet easy to use. The DUQuA scales are now publicly available to fill this niche. The scales could also be used to collect longitudinal data, for example before and after an intervention, used to collect information to assist in designing a quality improvement intervention, or as part of a comparison of hospitals or hospital departments.

At present, the CQII scale is likely to just confirm what is known from accreditation results. More work is required to develop or refine a tool that is more able to discriminate between hospitals with high implementation indices due to participation in external accreditation processes. Collection of data from international hospitals that are not subject to mandatory accreditation may assist this process.

Conclusions

The DUQuA organization and department scales can be used by Australian hospital managers to assess and measure improvement in quality management at organization and department levels within their hospitals and are readily modifiable for other health systems depending on their needs.

Acknowledgements

Members of the original DUQuE team (Oliver Gröne and Rosa Suñol) provided extensive input and advice in the initial stages of the project, supporting design modifications from DUQuE and the planned DUQuA study approach. Dr Annette Pantle provided expert advice on revision and development of DUQuA measures, and Professor Sandy Middleton, Associate Professor Dominique Cadillac, Kelvin Hill, Dr Carmel Crock and Professor Jacqueline Close provided input into the development of our stroke, AMI and hip fracture clinical process indicator lists. Nicole Mealing and Victoria Pye provided statistical advice in the initial phases of the project. We greatly appreciate their efforts.

Funding

This research was funded by the National Health and Medical Research Council (NHMRC) Program Grant APP1054146 (CI Braithwaite).

Authors’ contributions

The research team consists of experienced researchers, clinicians, biostatisticians and project managers with expertise in health services research, survey design and validation, large-scale research and project management, sophisticated statistical analysis, quality improvement and assessment, accreditation, clinical indicators, policy and patient experience. JB conceived the idea, led the research grant to fund the project and chairs the steering committee. RCW and NT co-led the detailed study design, managed the project across time and contributed to the development of the manuscript. HPT and GA provided statistical expertise for the study design and developed the analysis plan for the manuscript. TW contributed to the logistics of project management, the refinement of measures and the development of the manuscript.

Ethics approval

Ethical approvals were secured from State and Territory human research ethics committees in New South Wales (#14/206), Victoria (#15/36), the Australian Capital Territory (#15/131), South Australia (#15/260), the National Territory (#15-2509), Tasmania (#H0015383) and Queensland (#15/361). Site-specific authorizations, including permission for external researchers to collect data in hospitals, were granted. We complied with confidentiality requirements of national legislation or standards of practice of each jurisdiction.

Data sharing statement

Data will be made publicly available to the extent that individual participants or participating hospitals cannot be identified, in accordance with requirements of the approving Human Research Ethics Committees.

References