Article Text

Original research
How representative is the Victorian Emergency Minimum Dataset (VEMD) for population-based injury surveillance in Victoria? A retrospective observational study of administrative healthcare data
  1. Ehsan Rezaei-Darzi,
  2. Janneke Berecki-Gisolf,
  3. Dasamal Tharanga Fernando
  1. Victorian Injury Surveillance Unit (VISU), Monash University Accident Research Centre, Clayton, Victoria, Australia
  1. Correspondence to Ehsan Rezaei-Darzi; ehsan.rezaeidarzi{at}monash.edu

Abstract

Objective The Victorian Emergency Minimum Dataset (VEMD) is a key data resource for injury surveillance. The VEMD collects emergency department data from 39 public hospitals across Victoria; however, rural emergency care services are not well captured. The aim of this study is to determine the representativeness of the VEMD for injury surveillance.

Design A retrospective observational study of administrative healthcare data.

Setting and participants Injury admissions in 2014/2015–2018/2019 were extracted from the Victorian Admitted Episodes Dataset (VAED) which captures all Victorian hospital admissions; only cases that arrived through a hospital’s emergency department (ED) were included. Each admission was categorised as taking place in a VEMD-contributing versus a non-VEMD hospital.

Results There were 535 477 incident injury admissions in the study period, of which 517 207 (96.6%) were admitted to a VEMD contributing hospital. Male gender (OR 1.13 (95% CI 1.10 to 1.17)) and young age (age 0–14 vs 45–54 years, OR 4.68 (95% CI 3.52 to 6.21)) were associated with VEMD participating (vs non-VEMD-participating) hospitals. Residing in regional/rural areas was negatively associated with VEMD participating (vs non-VEMD participating) hospitals (OR=0.11 (95% CI 0.10 to 0.11)). Intentional injury (assault and self-harm) was also associated with VEMD participation.

Conclusions VEMD representativeness is largely consistent across the whole of Victoria, but varies vastly by region, with substantial under-representation of some areas of Victoria. By comparison, for injury surveillance, regional rates are more reliable when based on the VAED. For local ED-presentation rates, the bias analysis results can be used to create weights, as a temporary solution until rural emergency services injury data is systematically collected and included in state-wide injury surveillance databases.

  • public health
  • epidemiology
  • health policy

Data availability statement

Data may be obtained from a third party and are not publicly available. Data are available at Victorian Injury Surveillance Unit (VISU). Access fees for data and/or analyses may apply.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • The strength of this study is the use of whole of population Victorian injury admissions data from patients who arrived through the emergency department.

  • The Victorian Admitted Episodes Dataset is well suited for this analysis because it includes data from all Victorian hospitals, including small, regional hospitals.

  • Only emergency department (ED) presentations that were subsequently admitted were included in the analysis: this study is therefore limited to ED presentations for relatively severe injuries only.

Introduction

Background

Injuries are a major cause of morbidity and mortality: in Australia 2019/2020 over 527 000 hospitalisations were due to injury (2100 per 100 000 population).1 Injuries are preventable2; the first step towards injury control is injury surveillance. The purpose of injury surveillance is to measure the size of the injury problem in a timely manner, determine at-risk populations, identify emerging injury causes as well as recurrent injury problems, and help generate hypotheses regarding injury aetiology.3 Furthermore, surveillance is about translating information into action, as phrased by Horan and Mallonee4; injury surveillance serves to set injury prevention programme priorities, inform prevention and evaluate the effect of interventions.

Injury surveillance relies on high-quality, population-based data. The full potential of an injury surveillance system is only reached when it captures a range of injury data sources, and when quality of the data in terms of scope and accuracy are high. The surveillance system, if not capturing all injuries, should at least capture a representative sample: the representativeness of the system can be investigated and quantified.5

In the state of Victoria, Australia, the Victorian Injury Surveillance Unit (VISU) is a dedicated academic research group responsible for state-wide injury surveillance. VISU holds several key deidentified injury surveillance datasets detailing injury-related morbidity and mortality. For short-term time-dependent trends in injury, as well as for product-related injury surveillance, VISU is largely reliant on the Victorian Emergency Minimum Dataset (VEMD). The VEMD is a deidentified database of emergency department (ED) presentations to public hospitals with a designated ED, throughout Victoria (currently 39 hospitals). The VEMD data collection commenced in 1995 at the Department of Health and was progressively rolled out across EDs around the state, with data being collected by triage nurses at the point of presentation to the ED. It is intended to inform epidemiological research, health service planning and resource allocation, policy assessment and healthcare quality improvements.6 The VEMD contains a suite of dedicated injury surveillance data items: injury cause, injury type, bodily region, place of injury, activity when injured, human intent and a narrative (free text) regarding the circumstances of the injury occurrence.

The VEMD is a valuable injury surveillance data source that contains public hospitals ED data only; however, it has limitations. In this study, we address the potential limitation in representativeness of the VEMD data. A recent study reporting on Rural Acute Hospital Data Registry (RAHDaR), a rural emergency dataset based on data collected from ten rural hospitals in South West Victoria,7 highlights the gaps in the VEMD: only two of the participating centres also contributed to the VEMD.8 There are many small rural emergency services centres in Australia; they manage a total of 16% of emergency presentations in Australia but mostly do not meet the resource requirements to be termed an ED.9 10 The VEMD, which collects emergency data from 39 public hospitals in Victoria, does not therefore capture all emergency presentations in Victoria. Particularly, regional/rural presentations are likely to be under-represented. The aim of this study is to determine the representativeness of the VEMD for injury surveillance in Victoria. The data source for this study is the Victorian Admitted Episodes Dataset (VAED), which contains deidentified admission records from approximately 315 public and private hospitals in Victoria. In other words, the under-representation of certain care settings that limits the completeness of the VEMD is not present in the VAED. A dataset of injury-related admissions that arrived through the hospitals’ ED is extracted from the VAED; within this dataset, cases from VEMD-participating hospitals are identified and compared with cases from non-participating hospitals. Specifically, the following research questions are addressed: (1) What proportion of injury admissions that arrived through the ED (as reported in the VAED) are captured in the VEMD; (2) Is VEMD representation associated with sociodemographic variables such as age, sex, region of residence and socioeconomic status and (3) Is VEMD representation associated with injury type and severity? A secondary aim is to enable to results of this study to be used to address potential data biases in the VEMD, by suggesting weights to use for reporting population-based injury statistics for Victoria. Furthermore, this study is also relevant more broadly: the observed patterns of bias in data from hospitals with a 24-hour ED vs all hospitals with emergency care are likely to occur in other Australian jurisdictions, and in the national non-admitted patient ED care data (National Minimum Data Set11).

Method

This study is set in the state of Victoria, which is the second-most populous state in Australia with a population of 6.7 million (in 2020). Australia has a publicly funded universal healthcare scheme, Medicare, which provides Australian citizens and permanent residents with access to most healthcare services. Public hospital services are covered by Medicare. Many Australians have private health insurance. With private health insurance, patients can choose between treatment as a public patient in a public hospital or as a private patient in a private or public hospital. Victoria has over 200 public and private hospitals. In Victoria, major trauma patient care is managed through the Victorian State Trauma System (VSTS). The VSTS encompasses an integrated system featuring triage and transfer protocols, retrieval services, three designated major trauma services, research, education and training as well as patient care quality management.

This is a retrospective observational study of injury patients (not limited to major trauma) using existing hospital admissions data sourced from the VAED. The VAED dataset provides unit records of all private and public hospital admissions in Victoria. A retrospective analysis of data of patients admitted to hospital as a result of injury, between July 2014 and June 2019, was conducted. Injury presentations to EDs that were subsequently admitted to ward, and therefore recorded in the VAED, were selected. In other words, hospital admissions data from the VAED were used to analyse VEMD (in-)completeness, as the VAED contains a more comprehensive set of hospitals, and VEMD participation/nonparticipation could be flagged within this dataset.

In the VAED, hospital admissions that arrived through ED were categorised as (1) one of the 39 Victorian public hospitals contributing to the VEMD or (2) a hospital not contributing to the VEMD.

Data extraction

Hospital admissions between 1 July 2014 and 30 June 2019 were filtered for cases that their admission type states emergency admissions. Those admissions arrived through ED of the same campus. Persons of all ages were included. The International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Australian Modification (ICD-10-AM) was used to select injury-related hospital admissions. Hospital admission records in the VAED contain up to 40 ICD-10-AM diagnosis codes. Hospital admissions with any ICD-10-AM diagnosis codes related to injury, poisoning and certain other consequences of external cause codes (S00-T98) or external cause codes related to morbidity or mortality (U50-U73, U90-Y98) were selected. Cases with a gender coded as intersex or region coded as ‘unincorporated Victoria’ were excluded (both due to small cell counts). Cases that the cause code was medical injuries and late effects, the ICD-10-AM diagnosis code between Y40 and Y89, were also excluded. In order to select incident injury admissions, admissions that were statistical separations from the same hospital or inward transfers from another hospital were excluded. Following these exclusions, data included in this study originated from 39 public hospitals contributing to the VEMD and 43 public hospitals not contributing to the VEMD; the remainder stemmed from private hospitals not contributing to the VEMD. In other words, the VEMD data originates from 39 public hospitals only, whereas there are a further 43 Victorian public hospitals that do not contribute to the VEMD and a large number of private hospitals, none of which contribute to the VEMD. The data selection process is shown in Consolidated Standards of Reporting Trials flow diagram (online supplemental figure S1).

Figure 1

Interaction plot for VEMD-contribution outcome between region and age group in Victoria for the baseline model. VEMD, Victorian Emergency Minimum Dataset.

Study variables and outcome

The baseline explanatory factors were financial year, age group, sex, region (Department of Health Region)12 and Socio-Economic Indexes for Areas (SEIFA). The ‘year’ variable includes five financial years from 2014–2015 to 2018–2019. Patients were categorised into nine age groups: 0–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84 and 85+ years. The ‘region’ variable was extracted from the 2015 Local Government Area Statistical Profile, developed by the Victorian Department of Health.13 This is based on the location of the patient’s residence, not the location of the hospital. Each region is mapped to either regional/rural or metropolitan Victoria: this dichotomy is indicated in the results. To address the effect of regionality, the dichotomous variable regional/rural versus metropolitan is also used in modelling. SEIFA deciles (the index rankings and quantiles) were classified using the Index of Relative Socio-Economic Advantage and Disadvantage.14 15 Higher SEIFA deciles correspond with relatively lower measures of disadvantage; Decile 1 corresponds with the most disadvantaged areas.

Injury variables of interest were injury severity, cause, intent group (unintentional/assault/self-harm) and activity. Injury severity was calculated based on the ICISS (ICD-9 Injury Severity Score).16 Assuming the worst injury method,17 a serious injury was determined as ICD-based Injury severity score less than or equal to 0.941.18

The outcome was inclusion in the VEMD. Hospital-admitted injury patients arriving through the ED in VEMD-contributing hospitals were coded as having a positive outcome. Negative outcomes were assigned to hospital admitted injury patients arriving through the ED in a non-VEMD-contributing hospital.

Statistical methods

Logistic regression modelling was used to assess the association between sociodemographic and injury profile as independent variables with VEMD-contribution as an outcome variable. The baseline model was defined as one that contains only the socio-demographic variables. Relevant injury variables were then individually added stepwise to the baseline model to determine the impact of each. Significance of interaction effects were evaluated based on improvements in Hosmer and Lemeshow Goodness of Fit and Area Under the Curve (AUC) statistics; each model was checked with the Akaike information criterion (AIC) and AUC performance measurements. The model that has the lower AIC score indicates a more parsimonious model compared with a model with higher AIC. A model with the additional injury factor to the baseline that has AIC difference less than 10 and larger AUC provides further improvement to the model fit. Data and analysis steps were conducted in SAS V.9.4.

Patient and public involvement

Patients and/or public were not involved in the design, conduct and reporting of this study.

Result

In total, 535 477 hospital records of incident injury admissions that arrived to the hospital through the ED were included in the analyses. Of these, 517 207 (96.6%) were admitted to a VEMD contributing hospital, and the remaining 18 270 (3.4%) were admitted to a non-VEMD hospital. Of the non-VEMD hospital admissions, 69.5% were in a private hospital. The annual number of cases increased slightly over the years, but the proportions for each group (VEMD vs non-VEMD) was relatively stable, with the exception of a slight increase in the VEMD-hospital proportion observed in 2018/2019 (table 1). There was a steady age gradient in the VEMD-hospital proportion (χ2 test for linear trend: slope differs significantly from zero): patients less than 15 years old had the highest (99.05%) and patients aged 85 years and above had the lowest proportions (92.9%). Non-VEMD hospital (vs VEMD hospital) presentation was relatively more common among females (4.0%) than males (2.9%).

Table 1

Characteristics of hospital-admitted injury patients arriving through the emergency department by VEMD contributing and non-VEMD contributing hospitals

Overall, the study sample’s socioeconomic index level was above state average, as shown in the overrepresentation of patients in the SEIFA deciles 7–8 and 9–10 (31% and 27%, respectively, vs the expected 20% for each based on state averages). Although VEMD versus non-VEMD hospital attendance differed by SEIFA decile (table 1), there was no incremental, linear association between SEIFA and the hospital’s VEMD participation. The proportion of cases that were in VEMD-contributing hospitals were lowest in the Grampians region (80%) and highest in the Southern metropolitan region (99%). Injury severity groups (serious vs other injuries) were equally distributed across VEMD and non-VEMD contributing hospitals. Regarding injury cause, the lowest VEMD participation proportion was observed for unspecified unintentional injuries (93.7%) and falls (95.5%); the highest VEMD participation proportion was observed for intentional self-inflicted (99.0%) and other or undetermined intent injuries (98.7%). In terms of injury intent, assault injuries had the highest representation of VEMD hospitals (99.3%) and unintentional injuries had the lowest (96.2%). In terms of activity, the greatest representation of non-VEMD hospitals was observed for injuries that took place while doing vital activities, resting, eating, sleeping (5.6%).

Model results

Five models were run, each determining VEMD-participation status (ie, the outcome), starting with the baseline model (sociodemographic variables), followed by models that include injury-related variables, namely: injury severity, cause, intent group and activity. Model fit and predictive power statistics for each model are shown in table 2. The baseline model had very good predictive power (more than 83%), indicating that sociodemographic factors alone are good predictors of VEMD-participation in injury admissions. All other models were in this range, and the CIs of the AUC overlapped with the baseline model, except model 3, indicating improved model performance when adding the ‘cause’ variable. AIC measurement also confirmed this: model 3 had the lowest value. Compared with the baseline model, model 2 did not have better model performance, but had a lower AIC.

Table 2

Performance of important injury variables in assessing the association between injury variables and VEMD injury presentation in hospital admissions (Victoria)

We fit an additional model containing the rural variable instead of the region variable without considering any interaction. The results (not shown in tables) indicate that individuals attending VEMD participating hospitals were less likely to reside in rural areas (OR=0.107 (95% CI 0.102 to 0.112)), rural Victoria vs Melbourne metropolitan).

Further model results are shown in table 3. Based on the baseline model, in comparison to 2018/2019 (reference group), VEMD versus non-VEMD hospital presentation was less common in the years leading up to 2018/2019. Females had less representation in VEMD hospitals compared with males. The odds of being in a VEMD contributing hospital was higher for patients in SEIFA deciles 1–2 to 7–8 compared with those in SEIFA deciles 9–10 (the reference group), which indicate higher socioeconomic status.

Table 3

OR estimates and Wald CIs for logistic regression models for the outcome of presentation to VEMD contributing hospitals versus presentation to non-VEMD contributing hospitals

We expected to find that the association between age group and VEMD-hospital presentation varied by region. Therefore, interaction effects between age group and region on VEMD hospital presentation were added to the model. This interaction effect improved the model performance. In table 3, the coefficients of the age group and region represented a comparison with the reference groups of the variables that were used for the interaction term. In the following section, the main effects for age group and region and their interaction effects are explained.

The youngest group versus patients aged 45–54 years esiding in the North and West metropolitan were more likely to contribute in VEMD hospitals (OR=4.68 (95% CI 3.52 to 6.21)). People aged 45–54 years and residing in the Grampians region had a 7% higher tendency to contribute in VEMD hospitals than those in the North and West metropolitan region. However, the odds of contributing to VEMD hospitals for patients aged 45–54 years and residing in the Southern metropolitan was at least 1.54 times higher compared with those in the North and West metropolitan region.

Figure 1 shows the interaction plot by region and by age group; it shows that older age has an approximately linear association with decreased VEMD representation in Hume and Loddon regions and a curve relationship in other regions. Details of the OR estimations for the interaction effects between age group and regions by each model are shown in online supplemental table S1. It shows that the largest changes in the association of the age group main effect happened for those aged 0–14 years, when those in the Hume region were considered, compared with those in North and West metropolitan (OR=3.53 (95%CI 1.36 to 9.16)).

This graph shows the interaction effects between region and age group for patients contributing to the VEMD hospitals in contrast to the 45–54 years age group and North and West metropolitan region. The youngest group has a wide CI (95% CI 1.36 to 9.16) for Hume region; the error bars have been omitted to make the graph display clearly.

Based on the second model, patients with a more serious injury, had a higher likelihood of VEMD-contribution (OR 1.35 (95% CI 1.30 to 1.41)) than patients with other (non-serious) injuries. The third model, which adds injury cause to the baseline model, indicates that patients with self-inflicted or transport related injuries have higher odds (2.58 (95% CI 2.29 to 2.91) and 1.97 (95% CI 1.83 to 2.12), respectively) of presenting to VEMD (vs non-VEMD) hospitals compared with patients presenting due to a fall. Based on the baseline model with intent group added, the odds of presenting to a VEMD hospital for patients with assault injuries are around 2.88 (95% CI 2.40 to 3.46) and for self-harm injuries around 2.47 (95% CI 2.20 to 2.78), compared with unintentional injuries (reference group). Based on the fifth model, which adds activity to the baseline model, patients who were injured while engaging in leisure activities more likely to be treated in VEMD (vs non-VEMD) hospitals (OR 2.13 (95% CI 1.86 to 2.43)) compared with patients who were injured during sporting activities (reference group).

Discussion

In this study, we aimed to explore the representativeness of the VEMD for hospital-treated injury in Victoria. Although the vast majority of cases contributed to the VEMD, it was found that non-VEMD cases differed significantly in terms of residential region, SEIFA, age, sex and injury profile. Over two-thirds of non-VEMD admissions were in a private hospital. Males and children aged 0–14 years were more likely to be in VEMD-participating hospitals. People residing in the regions Barwon South Western, and Eastern metropolitan regions were relatively likely to be in non-VEMD hospitals. Intentional self-inflicted injury (compared with falls injury) was associated with VEMD-participating hospitals, as was intentional injury overall (assault and also self-harm, compared with unintentional injury).

The strengths of this study lie in the use of large, whole of population administrative data from all hospitals in Victoria, to determine potential shortfalls of injury surveillance captured in the VEMD. The VAED is well suited for this analysis as it includes data from smaller, regional hospitals; for example, in their study of falls presentations in EDs in the western region of Victoria, Holloway-Kew et al listed the smaller local hospitals that are not required to report to the VEMD. These hospitals are all captured in the VAED.19 Furthermore, the results of this study do not only provide insight into the gaps in the VEMD in terms of missing injury cases, but also the resulting selection bias.

This study is limited to the analysis of admitted injury cases in Victoria. Emergency presentations that are subsequently admitted are only a subset of all emergency presentations: therefore, the current findings are not necessarily generalisable to the all injury presentations. In the VEMD, 2014/2015–2018/2019, 20.8% of incident injury presentations were subsequently admitted to ward and 1.3% were transferred to another hospital (VISU in-house data). In a study by Baker and Dawson of five rural emergency services in South West Victoria, it was found that 16.4% of patients were admitted to ward and 4.4% were transferred.20 As the proportions of presentations that are subsequently admitted did not differ considerably between injury cases in the VEMD and the rural emergency services in the study by Baker and Dawson, this suggests that the patient casemix (in terms of urgency/severity) is similar across various Emergency Care settings in Victoria. If admission rates were to differ vastly between VEMD and non-VEMD contributing EDs, this would introduce a new study bias; based on the above comparison, this appear to be unlikely. However, it is recommended that future studies confirm the current findings, in all presentations’ data, not limited to cases that were subsequently admitted.

In this study, 3.4% of injury presentations that were captured in the VAED were in hospitals that do not report to the VEMD. Others have found much higher proportions: for hospital-treated child injuries in South West Victoria, a 35% deficit in VEMD capture was reported by Peck et al, based on 10 participating hospitals in the RAHDaR study21; specifically the shortfall in the VEMD and its implications for VISU are listed. The VEMD deficit is likely to be much larger in regional areas than in metropolitan areas, as also shown in the ranges reported in the various regions (table 1: 0.89% in Southern Metropolitan to 20.1% in the Grampians) and therefore the overall, state-wide VEMD deficit is much lower than 35%. However, the RAHDaR study highlights the limitations of the VEMD in regional Victoria. Rural emergency services have the capacity to collect accurate, episode-level electronic data9 ; over the course of a year, these services have been demonstrated to encounter the full range of clinical problems seen in full EDs.20 Rural emergency services are an integral part of VSTS; major trauma patients can be resuscitated and stabilised before being transported to a major trauma service. It is highly recommended that injury surveillance data is collected electronically in rural emergency services and reported to the VEMD. For feasibility, this could be done intermittently or only in representative service centres, allowing for extrapolation of these additional, rural emergency services data to complete the currently collected VEMD injury surveillance data.

The findings of this study have several implications for the practice of injury surveillance. First, the ‘VEMD deficit’ must be acknowledged. This deficit is not sizeable when considered across the whole of Victoria, but it varies vastly by region, with substantial ‘VEMD deficit’ in regions such as the Grampians and Barwon South Western. For injury surveillance, injury rates by region are therefore much more reliable when based on admissions data from the VAED, which captures all hospitals in Victoria. If the VEMD is used to supply injury rates by region, the proportions reported in this study can be used create weights to correct for VEMD deficits; this can be a temporary solution until rural emergency services injury surveillance data is systematically collected and included in state-wide injury surveillance databases. Similarly, for state-wide injury surveillance statistics that based on the VEMD, the selection bias that is quantified in this study can be used to adjust for the under-representation of older people and females in the VEMD. As there are region-dependent differences in selection bias with regard to age, demonstrated in the age–region interaction effects, this may be overly complex, and supplementation of the VEMD with regional emergency care services data is the preferred solution.

Data availability statement

Data may be obtained from a third party and are not publicly available. Data are available at Victorian Injury Surveillance Unit (VISU). Access fees for data and/or analyses may apply.

Ethics statements

Patient consent for publication

Ethics approval

Use and dissemination of Victorian Injury Surveillance Unit hospital admissions data was reviewed by the Monash University Human Research Ethics Committee. The Committee has granted approval of the National Statement on Ethical Conduct in Human Research, the project id is 21427.

Acknowledgments

The authors would like to acknowledge the Victorian Department of Health as the source of VAED data for this study.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors Conceptualisation: JB-G, DTF and ER-D, Methodology: JB-G and ER-D, Formal analysis: ER-D, Writing: ER-D, JB-G and DTF. All authors read and approved the final manuscript.

  • Funding The Victorian Injury Surveillance Unit is supported by the Victorian Government.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.