Selection bias found in interpreting analyses with missing data for the prehospital index for trauma

Injury severity bias in missing prehospital vital signs: Prevalence and implications for trauma registries

BACKGROUND

Vital signs are important factors in assessing injury severity and guiding trauma resuscitation, especially among severely injured patients. Despite this, physiological data are frequently missing from trauma registries. This study aimed to evaluate the extent of missing prehospital data in a hospital-based trauma registry and to assess the associations between prehospital physiological data completeness and indicators of injury severity.

METHODS

A retrospective review was conducted on all adult trauma patients brought directly to a level 1 trauma center in Toronto, Ontario by paramedics from January 1, 2015, to December 31, 2019. The proportion of missing data was evaluated for each variable and patterns of missingness were assessed. To investigate the associations between prehospital data completeness and injury severity factors, descriptive and unadjusted logistic regression analyses were performed.

RESULTS

A total of 3,528 patients were included. We considered prehospital data missing if any of heart rate, systolic blood pressure, respiratory rate or oxygen saturation were incomplete. Each individual variable was missing from the registry in approximately 20 % of patients, with oxygen saturation missing most frequently (n = 831; 23.6 %). Over 25 % (n = 909) of patients were missing at least one prehospital vital sign, of which 69.1 % (n = 628) were missing all four of these variables. Patients with incomplete data were more severely injured, had higher mortality, and more frequently received lifesaving interventions such as blood transfusion and intubation. Patients were most likely to have missing prehospital physiological data if they died in the trauma bay (unadjusted OR: 9.79; 95 % CI: 6.35-15.10), did not survive to discharge (unadjusted OR: 3.55; 95 % CI: 2.76-4.55), or had a prehospital GCS less than 9 (OR: 3.24; 95 % CI: 2.59-4.06).

CONCLUSION

In this single center trauma registry, key prehospital variables were frequently missing, particularly among more severely injured patients. Patients with missing data had higher mortality, more severe injury characteristics and received more life-saving interventions in the trauma bay, suggesting an injury severity bias in prehospital vital sign missingness. To ensure the validity of research based on trauma registry data, patterns of missingness must be carefully considered to ensure missing data is appropriately addressed.

Effects of Post-Hospital Arrival Factors on Out-of-Hospital Cardiac Arrest Outcomes During the COVID-19 Pandemic

IMPORTANCE

The relationship between post-hospital arrival factors and out-of-hospital cardiac arrest (OHCA) outcomes remains unclear.

OBJECTIVES

This study assessed the impact of post-hospital arrival factors on OHCA outcomes during the COVID-19 pandemic using a prediction model.

DESIGN, SETTING, AND PARTICIPANTS

In this cohort study, data from the All-Japan Utstein Registry, a nationwide population-based database, between 2015 and 2021 were used. A total of 541,781 patients older than 18 years old who experienced OHCA of cardiac origin were included.

MAIN OUTCOMES AND MEASURES

The primary exposure was trends in COVID-19 cases. The study compared the predicted proportion of favorable neurologic outcomes 1 month after resuscitation with the actual outcomes. Neurologic outcomes were categorized based on the Cerebral Performance Category score (1, good cerebral function; 2, moderate cerebral function).

RESULTS

The prediction model, which had an area under the curve of 0.96, closely matched actual outcomes in 2019. However, a significant discrepancy emerged after the pandemic began in 2020, where outcomes continued to deteriorate as the virus spread, exacerbated by both pre- and post-hospital arrival factors.

CONCLUSIONS AND RELEVANCE

Post-hospital arrival factors were as important as pre-hospital factors in adversely affecting the prognosis of patients following OHCA during the COVID-19 pandemic. The results suggest that the overall response of the healthcare system needs to be improved during infectious disease outbreaks to improve outcomes.

The association between geospatial and temporal factors and pre-hospital response to major trauma: a retrospective cohort study in the North of England

BACKGROUND

Major trauma is a leading cause of premature death and disability worldwide, and many healthcare systems seek to improve outcomes following severe injury with provision of pre-hospital critical care. Much research has focussed on the efficacy of pre-hospital critical care and advanced pre-hospital interventions, but less is known about how the structure of pre-hospital critical care services may influence response to major trauma. This study assessed the association between likelihood of pre-hospital critical care response in major trauma and factors important in the planning and development of those services: geographic isolation, time of day, and tasking mechanism.

METHODS

A local trauma registry, supported with data from the Trauma Audit and Research Network alongside additional information regarding pre-hospital management, identified patients sustaining major trauma admitted to Major Trauma Centres in the North of England. Data was extracted on location and time of incident, mechanism of injury, on-scene times, and presence or absence of pre-hospital critical care team. An isochrone map was constructed for 30-minute intervals to regional Major Trauma Centres, defining geographic isolation. Univariate logistic regression compared likelihood of pre-hospital critical care response to that of conventional ambulance response for varying degrees of geographic isolation, day or night period, and mechanism of injury, and multiple linear regression assessed the association between geographic isolation, service response and on-scene time.

RESULTS

2619 incidents were included, with 23.3% attended by pre-hospital critical care teams. Compared to conventional ambulance services, pre-hospital critical care teams were more likely to respond major trauma in areas of greater geographic isolation (OR 1.42, 95% CI 1.30-1.55, p < 0.005). There were significant differences in the mechanism of injury attended and no significant difference in response by day or night period. Pre-hospital critical care team response and increasing geographic isolation was associated with longer on-scene times (p < 0.005).

CONCLUSION

Pre-hospital critical care teams are more likely to respond to major trauma in areas of greater geographic isolation. Enhanced pre-hospital care may mitigate geographic inequalities when providing advanced interventions and transport of severely injured patients. There may be an unmet need for pre-hospital critical care response in areas close to major hospitals.

Pre-hospital and retrieval medicine in Scotland: a retrospective cohort study of the workload and outcomes of the emergency medical retrieval service in the first decade of national coverage

BACKGROUND

The Emergency Medical Retrieval Service (EMRS) has provided national pre-hospital critical care and aeromedical retrieval in Scotland since 2010. This study investigates trends in the service and patients attended over the last decade; and factors associated with clinical deterioration and pre-hospital death.

METHODS

A retrospective cohort study was conducted of all service taskings over ten years (2011-2020 inclusive). The EMRS electronic database provided data on location, sociodemographic factors, diagnoses, physiological measurements, clinical management, and pre-hospital deaths. Binary logistic regression models were used to determine change in physiology in pre-hospital care, and factors associated with pre-hospital death. Geospatial modelling, using road and air travel time models, was used to explore transfer times.

RESULTS

EMRS received 8,069 taskings over the study period, of which 2,748 retrieval and 3,633 pre-hospital critical care missions resulted in patient contact. EMRS was more commonly dispatched to socioeconomically deprived areas for pre-hospital critical care incidents (Spearman's rank correlation, r(8)=-0.75, p = 0.01). In multivariate analysis, systolic blood pressure < 90mmHg, respiratory rate < 6/min or > 30/min, and Glasgow Coma Score ≤ 14 were associated with pre-hospital mortality independent of demographic factors. Geospatial modelling suggested that aeromedical retrieval reduced the mean time to a critical care unit by 1 h 46 min compared with road/ferry transportation.

CONCLUSION

EMRS continues to develop, delivering Pre-Hospital and Retrieval Medicine across Scotland and may have a role in addressing health inequalities, including socioeconomic deprivation and geographic isolation. Age, specific distances from care, and abnormal physiology are associated with death in pre-hospital critical care.

Evaluation of Missing Prehospital Physiological Values in Injured Children and Adolescents

INTRODUCTION

Prehospital vital signs and the Glasgow Coma Scale score are often missing in clinical practice and not recorded in trauma databases. Our study aimed to identify factors associated with missing prehospital physiological values, including systolic blood pressure, heart rate, respiratory rate, peripheral oxygen saturation, and Glasgow Coma Scale.

METHODS

We used our hospital trauma registry to obtain patient, injury, resuscitation, and transportation characteristics for injured children and adolescents (age <15 y). We evaluated the association of missing documentation of prehospital values with other patient, injury, transportation, and resuscitation characteristics using multivariable regression. We standardized vital sign values using age-adjusted z-scores.

RESULTS

The odds of a missing physiological value decreased with age (odds ratio [OR] = 0.9, 95% confidence interval [CI] = 0.9, 0.9) and were higher when prehospital cardiopulmonary resuscitation was required (OR = 3.3, 95% CI = 1.9, 5.7). Among the physiological values considered, we observed the highest odds of missingness of systolic blood pressure, respiratory rate, and oxygen saturation. The odds of observing normal emergency department physiological values were lower when prehospital physiological values were missing (OR = 0.9, 95% CI = 0.9, 1.0; P = 0.04).

CONCLUSIONS

Missing prehospital physiological values were associated with younger age and cardiopulmonary resuscitation among the injured children treated at our hospital. Measurement and documentation of physiological variables of patients with these characteristics should be targeted.

Practical guidance to handle missing values in the 25-question Geriatric Locomotive Function Scale (GLFS-25): a simulation study

OBJECTIVES

Despite the possible large number of missing values on the 25-question Geriatric Locomotive Function Scale (GLFS-25), how we should treat them is unknown. In a simulation study, we investigated how to handle missing values in the GLFS-25.

DESIGN, SETTING AND PARTICIPANTS

We used three datasets with different participant characteristics: community dwellers who could walk by themselves, outpatients of orthopaedics owing to pain, and patients who required surgery for total knee replacement or lumbar spinal canal stenosis.

OUTCOME MEASURES

The missing items of the datasets were artificially created, and four statistical methods, complete case analysis, multiple imputation, single imputation using individual mean, and single imputation using individual domain average, were compared in terms of bias and mean squared error. Simulation studies were conducted to compare them under varying numbers of participants with missing values (5%-40%) and under varying numbers of missing items of GLFS-25 (4-16).

RESULTS

Multiple imputation had the lowest root mean squared error. Complete case analysis showed the largest bias, and the performances of the single imputation were between those methods. The relative performances were similar across the three datasets. The absolute bias of the single imputation was<0.1. The bias and mean squared error of multiple imputation and single imputation were comparable when the number of missing items was less than or equal to eight.

CONCLUSIONS

Multiple imputation is preferable, although single imputation using subject average/subject domain average can be used with practically negligible bias as long as the number of missing items is up to 8 out of 25 items in each individual of the population.

Managing Missing Data in the Hospital Survey on Patient Safety Culture: A Simulation Study

BACKGROUND

Case-wise analysis is advocated for the Hospital Survey on Patient Safety culture (HSOPS).

OBJECTIVES

Through a computer-intensive simulation study, we aimed to evaluate the accuracy of various imputation methods in managing missing data in the HSOPS.

METHODS

Using the original data from a cross-sectional survey of 5064 employees at a single university hospital in France, we produced simulation data on two levels. First, we resampled 1000 completed data based on the original 3045 complete responses using a bootstrap procedure. Second, missing values were simulated in these 1000 completed case data for comparison purposes, using eight different missing data scenarios. Third, missing values were imputed using five different imputation methods (1, random imputation; 2, item mean; 3, individual mean; 4, multiple imputation, and 5, sparse nonnegative matrix factorization. The performance for each imputation method was assessed using the root mean square error and dimension score bias.

RESULTS

The five imputation methods yielded close root mean square errors, with an advantage for the multiple imputation. The bias differences were greater regarding the dimension scores, with a clear advantage for multiple imputation. The worst performance was achieved by the mean imputation methods.

DISCUSSION AND CONCLUSIONS

We recommend the use of multiple imputation to handle missing data in HSOPS-based surveys, whereas mean imputation methods should be avoided. Overall, these results suggest the possibility of optimizing the HSOPS instrument, which should be reduced without loss of overall information.

Completeness of Medical Records of Trauma Patients Admitted to the Emergency Unit of a University Hospital, Upper Egypt

Trauma records in Egyptian hospitals are widely suspected to be inadequate for developing a practical and useful trauma registry, which is critical for informing both primary and secondary prevention. We reviewed archived paper records of trauma patients admitted to the Beni-Suef University Hospital in Upper Egypt for completeness in four domains: demographic data including contact information, administrative data tracking patients from admission to discharge, clinical data including vital signs and Glasgow Coma Scale scores, and data describing the causal traumatic event (mechanism of injury, activity at the time of injury, and location/setting). The majority of the 539 medical records included in the study had significant deficiencies in the four reviewed domains. Overall, 74.3% of demographic fields, 66.5% of administrative fields, 55.0% of clinical fields, and just 19.9% of fields detailing the causal event were found to be completed. Critically, oxygen saturation, arrival time, and contact information were reported in only 7.6%, 25.8%, and 43.6% of the records, respectively. Less than a fourth of the records provided any details about the cause of trauma. Accordingly, the current, paper-based medical record system at Beni-Suef University Hospital is insufficient for the development of a practical trauma registry. More efforts are needed to develop efficient and comprehensive documentation of trauma data in order to inform and improve patient care.

Risk factors associated with short term mortality changes over time, after arrival to the emergency department

Background

Preventing death is the most important outcome pursued in the Emergency Department. Prompt accurate assessment, followed by competent and efficient investigation and treatment is the recipe sought. Abnormal physiological measurements are common antecedents to deterioration and therefore a cornerstone in many risk stratification tools. Some risk factors have their impact during the first few days after admittance, others have higher impact on 30 day mortality. Understanding the variance in impact of risk factors is relevant for future composition of risk stratification models.

Methods

We included patients aged 18 years or older, registered at the Emergency Department at Odense University Hospital from April 1st 2012 to September 30th 2014. We performed multivariate logistic regressions, adjusted for age, gender and comorbidity, to describe the relationship between potential risk factors and measures of short term mortality.

Results

A total of 43,178 were eligible for analysis. Median age was 56 (IQR 36–72) and 48.3% were males. The over-all 30-day-mortality was 4%. One third of deaths occurred within the first 2 days.

Higher age, male gender and comorbidity are all associated with immediate, 0-2 day, 3-7 day and 8–30 day mortality. The degree of acuteness at arrival defined by urgency-level, physician-assisted transfer to the Emergency Department and abnormal vital parameters are associated with 0-2 day mortality. High temperature at arrival shows no association in either mortality-group. Missing values are associated with immediate and 0–2 day mortality, but no association with mortality after 7 days.

Discussion

Abnormal vital parameters and degree of acuity at admission were strongly associated with mortality in the first hours and days after admission, where after the association decreased. The effect of other risk factors such as male gender, comorbidity and high age were time stable or even increasing over time..

Conclusions

The over-all 30-day mortality was 4%. Physiology–related risk factors varied in strength of association throughout different mortality outcome measures.

Electronic supplementary material

The online version of this article (10.1186/s13049-018-0493-2) contains supplementary material, which is available to authorized users.

Imputation strategies in the trauma registration

BACKGROUND

Trauma databases often contain relatively high proportions of missing physiologic values. Multiple imputation (MI) could be a possible adequate solution for the missing values. This study aimed to demonstrate the influence of more simplified imputation models on standardized W statistic (Ws) (number of excess survivors per hundred patients that would be achieved if the study center treated identically the same case mix as the reference population).

METHODS

Data from three trauma care networks in the Netherlands were used to investigate local differences in missing data. Five different imputation models (MI 1 to 5) were created based on literature and expert opinion. A sixth database was created using maximal single imputation and a seventh database with only complete case analysis (CCA). The Ws values were calculated for the three regions separately.

RESULTS

A total of 8,853, 24,487, and 8,599 observations were examined in region 1, region 2, and region 3, respectively. The Ws in region 1 ranged from -0.48 (95% confidence interval [CI], -1.71 to 0.80) for CCA to 0.53 (95% CI, -0.19 to 1.26) for MI 4 and a range of 0.40 (95% CI, -0.91 to 0.10) for CCA to -0.32 (-0.69, 0.04) for MI 1 and MI 4 was found in region 2. The Ws for region 3 ranged from -0.19 (-0.83 to 0.45) in all MI data sets to -0.12 (-0.76 to 0.52) in the CCA data set. Although there were no significant differences between the Ws of the imputation data sets and the CCA analysis, large differences were found in the region with the most missing values.

CONCLUSION

Different imputation strategies did influence Ws values. Supplementary variables showed no additional value for the imputation process and a more simplified imputation model could be used to adequately impute missing data.

LEVEL OF EVIDENCE

Prognostic, level II.

Individual and Neighborhood Characteristics of Children Seeking Emergency Department Care for Firearm Injuries Within the PECARN Network

OBJECTIVE

The objective was to describe the characteristics of children seeking emergency care for firearm injuries within the PECARN network and assess the influence of both individual and neighborhood factors on firearm-related injury risk.

METHODS

This was a retrospective, multicenter cross-sectional analysis of children (<19 years old) presenting to 16 pediatric EDs (2004-2008). ICD-9-CM E-codes were used to identify and categorize firearm injuries by mechanism/intent. Neighborhood variables were derived from home address data. Multivariable analysis examined the influence of individual and neighborhood factors on firearm-related injuries compared to nonfirearm ED visits. Injury recidivism was assessed.

RESULTS

A total of 1,758 pediatric ED visits for firearm-related injuries were analyzed. Assault (51.4%, n = 904) and unintentional injury (33.2%, n = 584) were the most common injury mechanisms. Among children with firearm injuries, 68.3% were older adolescents (15-19 years old), 82.3% were male, 68.2% were African American, and 76.3% received public insurance/were uninsured. Extremity injuries were most common (75.9%), with 20% sustaining injuries to multiple body regions, 48.1% requiring admission and 1% ED mortality. Multivariable analysis identified firearm injury risk factors, including adolescent age (p < 0.001), male sex (p < 0.001), non-Caucasian race/ethnicity (p < 0.001), public payer/uninsured status (p < 0.001), and higher levels of neighborhood disadvantage (p < 0.001). Among children with firearm injuries, 12-month ED recidivism for any reason was 22.4%, with < 1% returning for another firearm injury.

CONCLUSION

Among children receiving ED treatment within the PECARN network, there are distinct demographic and neighborhood factors associated with firearm injuries. Among younger children (<10 years old), unintentional injuries predominate, while assault-type injuries were most common among older adolescents. Overall, among this PECARN patient population, male adolescents living in neighborhoods characterized by high levels of concentrated disadvantage had an elevated risk for firearm injury. Public health efforts should focus on developing and implementing initiatives addressing risk factors at both the individual and the community level, including ED-based interventions to reduce the risk for firearm injuries among high-risk pediatric populations.

Lead-Time Bias and Interhospital Transfer after Injury: Trauma Center Admission Vital Signs Underpredict Mortality in Transferred Trauma Patients

BACKGROUND

Admission physiology predicts mortality after injury, but may be improved by resuscitation before transfer. This phenomenon, which has been termed lead-time bias, may lead to underprediction of mortality in transferred patients and inaccurate benchmarking in centers receiving large numbers of transfer patients. We sought to determine the impact of using vital signs on arrival at the referring center vs on arrival at the trauma center in mortality prediction models for transferred trauma patients.

STUDY DESIGN

We performed a retrospective cohort study using a state-wide trauma registry including all patients age 16 years or older, with Abbreviated Injury Scale scores ≥ 3, admitted to level I and II trauma centers in Pennsylvania, from 2011 to 2014. The primary outcomes measure was the risk-adjusted association between mortality and interhospital transfer (IHT) when adjusting for physiology (as measured by Revised Trauma Score [RTS]) using the referring hospital arrival vital signs (model 1) compared with trauma center arrival vital signs (model 2).

RESULTS

After adjusting for patient and injury factors, IHT was associated with reduced mortality (odds ratio [OR] 0.85; 95% CI 0.77 to 0.93) using the RTS from trauma center admission, but with increased mortality (OR 1.15; 95% CI 1.05 to 1.27) using RTS from the referring hospital. The greater the number of transfer patients seen by a center, the greater the difference in center-level mortality predicted by the 2 models (β -0.044; 95% CI -0.044 to -0.0043; p ≤ 0.001).

CONCLUSIONS

Trauma center vital signs underestimate mortality in transfer patients and may lead to incorrect estimates of expected mortality. Where possible, benchmarking efforts should use referring hospital vital signs to risk-adjust IHT patients.

Classifying, measuring and improving the quality of data in trauma registries: A review of the literature

INTRODUCTION

Globally, injury is a major cause of death and disability. Improvements in trauma care have been driven by trauma registries. The capacity of a trauma registry to inform improvements in the quality of trauma care is dependent upon the quality of data. The literature on data quality in disease registries is inconsistent and ambiguous; methods used for classifying, measuring, and improving data quality are not standardised. The aim of this study was to review the literature to determine the methods used to classify, measure and improve data quality in trauma registries.

METHODS

A scoping review of the literature was performed. Databases were searched using the term "trauma registry" and its synonyms, combined with multiple terms denoting data quality. There was no restriction on year. Full-length manuscripts were included if the classification, measurement or improvement of data quality in one or more trauma registries was a study objective. Data were abstracted regarding registry demographics, study design, data quality classification, and the reported methods used to measure and improve the pre-defined data quality dimensions of accuracy, completeness and capture.

RESULTS

Sixty-nine publications met the inclusion criteria. Four publications classified data quality. The most frequently described methods for measuring data accuracy (n=47) were checks against other datasets (n=18) and checks of injury coding (n=17). The most frequently described methods for measuring data completeness (n=47) were the percentage of included cases, for a given variable or list of variables, for which there was an observation in the registry (n=29). The most frequently described methods for measuring data capture (n=37) were the percentage of cases in a linked reference dataset that were also captured in the primary dataset being evaluated (n=24). Most publications dealing with the measurement of a dimension of data quality did not specify the methods used; most publications dealing with the improvement of data quality did not specify the dimension being targeted.

CONCLUSION

The classification, measurement and improvement of data quality in trauma registries is inconsistent. To maintain confidence in the usefulness of trauma registries, the metrics and reporting of data quality need to be standardised.

Hamsi scoring in the prediction of unfavorable outcomes from tuberculous meningitis: results of Haydarpasa-II study

Predicting unfavorable outcome is of paramount importance in clinical decision making. Accordingly, we designed this multinational study, which provided the largest case series of tuberculous meningitis (TBM). 43 centers from 14 countries (Albania, Croatia, Denmark, Egypt, France, Hungary, Iraq, Italy, Macedonia, Romania, Serbia, Slovenia, Syria, Turkey) submitted data of microbiologically confirmed TBM patients hospitalized between 2000 and 2012. Unfavorable outcome was defined as survival with significant sequela or death. In developing our index, binary logistic regression models were constructed via 200 replicates of database by bootstrap resampling methodology. The final model was built according to the selection frequencies of variables. The severity scale included variables with arbitrary scores proportional to predictive powers of terms in the final model. The final model was internally validated by bootstrap resampling. A total of 507 patients' data were submitted among which 165 had unfavorable outcome. Eighty-six patients died while 119 had different neurological sequelae in 79 (16%) patients. The full model included 13 variables. Age, nausea, vomiting, altered consciousness, hydrocephalus, vasculitis, immunosuppression, diabetes mellitus and neurological deficit remained in the final model. Scores 1-3 were assigned to the variables in the severity scale, which included scores of 1-6. The distribution of mortality for the scores 1-6 was 3.4, 8.2, 20.6, 31, 30 and 40.1%, respectively. Altered consciousness, diabetes mellitus, immunosuppression, neurological deficits, hydrocephalus, and vasculitis predicted the unfavorable outcome in the scoring and the cumulative score provided a linear estimation of prognosis.

A survey of trauma database utilisation in England

Trauma registries are used worldwide to coordinate patient care as well as provide data for audit and research purposes. National registries collect this data, producing research opportunities, outcome standards and a means by which to benchmark trauma centre performance. The Trauma Audit and Research Network (TARN) is the UK national registry, with data upload being mandatory from all major trauma centres (MTCs), a process which is manual and time and resource intensive. A telephone survey was carried out to collect data from all 26 MTCs in England. A questionnaire was designed to identify how data was collected at a local level, what software and methods were used and what resources were allocated to collect and upload trauma data to the TARN. Further information on hospital size and number of beds was collected from internet searches. Twenty-three MTCs were contacted in total. The majority used Microsoft Excel, with the next most common programme being Bluespier. Other commercially available registries used included Collector, VTOMS and McKesson. One trust created its own software and three used no electronic database at all. Electronic patient record integration was variable and limited to some commercially available registries. The mean number of TARN data collectors was two per centre, with a mean duration of data collection of 4.5 years. The wide range of software options and their lack of integration with the hospital electronic patient records results in the duplication of data as well as requiring time and resources. This may also be due to the difference in data required for coordinating on-going patient care and that required for upload to the TARN. Whilst some of these programmes do have the capabilities for automatic data upload, further efforts must be made to provide a cohesive system that provides the required integration and customisability in order to improve efficiency and ultimately trauma care.

Identification of dynamic prehospital changes with continuous vital signs acquisition

OBJECTIVE

In most trauma registries, prehospital trauma data are often missing or unreliable because of the difficult dual task consigned to prehospital providers of recording vital signs and simultaneously resuscitating patients. The purpose of this study was to test the hypothesis that the analysis of continuous vital signs acquired automatically, without prehospital provider input, improves vital signs data quality, captures more extreme values that might be missed with conventional human data recording, and changes Trauma Injury Severity Scores compared with retrospectively compiled prehospital trauma registry data.

METHODS

A statewide vital signs collection network in 6 medevac helicopters was deployed for prehospital vital signs acquisition using a locally built vital signs data recorder (VSDR) to capture continuous vital signs from the patient monitor onto a memory card. VSDR vital signs data were assessed by 3 raters, and intraclass correlation coefficients were calculated to test interrater reliability. Agreement between VSDR and trauma registry data was compared with the methods of Altman and Bland including corresponding calculations for precision and bias.

RESULTS

Automated prehospital continuous VSDR data were collected in 177 patients. There was good agreement between the first recorded vital signs from the VSDR and the trauma registry value. Significant differences were observed between the highest and lowest heart rate, systolic blood pressure, and pulse oximeter from the VSDR and the trauma registry data (P< .001). Trauma Injury Severity Scores changed in 12 patients (7%) when using data from the VSDR.

CONCLUSION

Real-time continuous vital signs monitoring and data acquisition can identify dynamic prehospital changes, which may be missed compared with vital signs recorded manually during distinct prehospital intervals. In the future, the use of automated vital signs trending may improve the quality of data reported for inclusion in trauma registries. These data may be used to develop improved triage algorithms aimed at optimizing resource use and enhancing patient outcomes.

Gunshot injuries in children served by emergency services

OBJECTIVE

To describe the incidence, injury severity, resource use, mortality, and costs for children with gunshot injuries, compared with other injury mechanisms.

METHODS

This was a population-based, retrospective cohort study (January 1, 2006-December 31, 2008) including all injured children age ≤ 19 years with a 9-1-1 response from 47 emergency medical services agencies transporting to 93 hospitals in 5 regions of the western United States. Outcomes included population-adjusted incidence, injury severity score ≥ 16, major surgery, blood transfusion, mortality, and average per-patient acute care costs.

RESULTS

A total of 49,983 injured children had a 9-1-1 emergency medical services response, including 505 (1.0%) with gunshot injuries (83.2% age 15-19 years, 84.5% male). The population-adjusted annual incidence of gunshot injuries was 7.5 cases/100,000 children, which varied 16-fold between regions. Compared with children who had other mechanisms of injury, those injured by gunshot had the highest proportion of serious injuries (23%, 95% confidence interval [CI] 17.6-28.4), major surgery (32%, 95% CI 26.1-38.5), in-hospital mortality (8.0%, 95% CI 4.7-11.4), and costs ($28,510 per patient, 95% CI 22,193-34,827).

CONCLUSIONS

Despite being less common than other injury mechanisms, gunshot injuries cause a disproportionate burden of adverse outcomes in children, particularly among older adolescent males. Public health, injury prevention, and health policy solutions are needed to reduce gunshot injuries in children.

Patient choice in the selection of hospitals by 9-1-1 emergency medical services providers in trauma systems

OBJECTIVES

Reasons for undertriage (transporting seriously injured patients to nontrauma centers) and the apparent lack of benefit of trauma centers among older adults remain unclear; understanding emergency medical services (EMS) provider reasons for selecting certain hospitals in trauma systems may provide insight to these issues. In this study, the authors evaluated reasons cited by EMS providers for selecting specific hospital destinations for injured patients, stratified by age, injury severity, field triage status, and prognosis.

METHODS

This was a retrospective cohort study of injured children and adults transported by 61 EMS agencies to 93 hospitals (trauma and nontrauma centers) in five regions of the western United States from 2006 through 2008. Hospital records were probabilistically linked to EMS records using trauma registries, state discharge data, and emergency department data. The seven standardized reasons cited by EMS providers for selecting hospital destinations included closest facility, ambulance diversion, physician choice, law enforcement choice, patient or family choice, specialty resource center, and other. "Serious injury" was defined as an Injury Severity Score (ISS) ≥ 16, and unadjusted in-hospital mortality was considered as a marker of prognosis. All analyses were stratified by age in 10-year increments, and descriptive statistics were used to characterize the findings.

RESULTS

A total of 176,981 injured patients were evaluated and transported by EMS over the 3-year period, of whom 5,752 (3.3%) had ISS ≥ 16 and 2,773 (1.6%) died. Patient or family choice (50.6%), closest facility (20.7%), and specialty resource center (15.2%) were the most common reasons indicated by EMS providers for selecting destination hospitals; these frequencies varied substantially by patient age. The frequency of patient or family choice increased with increasing age, from 36.4% among 21- to 30-year-olds to 75.8% among those older than 90 years. This trend paralleled undertriage rates and persisted when restricted to patients with serious injuries. Older patients with the worst prognoses were preferentially transported to major trauma centers, a finding that was not explained by field triage protocols.

CONCLUSIONS

Emergency medical services transport patterns among injured patients are not random, even after accounting for field triage protocols. The selection of hospitals appears to be heavily influenced by patient or family choice, which increases with patient age and involves inherent differences in patient prognosis.

A comparison of missing data procedures for addressing selection bias in HIV sentinel surveillance data

BACKGROUND

Selection bias is common in clinic-based HIV surveillance. Clinics located in HIV hotspots are often the first to be chosen and monitored, while clinics in less prevalent areas are added to the surveillance system later on. Consequently, the estimated HIV prevalence based on clinic data is substantially distorted, with markedly higher HIV prevalence in the earlier periods and trends that reveal much more dramatic declines than actually occur.

METHODS

Using simulations, we compare and contrast the performance of the various approaches and models for handling selection bias in clinic-based HIV surveillance. In particular, we compare the application of complete-case analysis and multiple imputation (MI). Several models are considered for each of the approaches. We demonstrate the application of the methods through sentinel surveillance data collected between 2002 and 2008 from India.

RESULTS

Simulations suggested that selection bias, if not handled properly, can lead to biased estimates of HIV prevalence trends and inaccurate evaluation of program impact. Complete-case analysis and MI differed considerably in their ability to handle selection bias. In scenarios where HIV prevalence remained constant over time (i.e. β = 0), the estimated β^1 derived from MI tended to be biased downward. Depending on the imputation model used, the estimated bias ranged from -1.883 to -0.048 in logit prevalence. Furthermore, as the level of selection bias intensified, the extent of bias also increased. In contrast, the estimates yielded by complete-case analysis were relatively unbiased and stable across the various scenarios. The estimated bias ranged from -0.002 to 0.002 in logit prevalence.

CONCLUSIONS

Given that selection bias is common in clinic-based HIV surveillance, when analyzing data from such sources appropriate adjustment methods need to be applied. The results in this paper suggest that indiscriminant application of imputation models can lead to biased results.

The impact of missing trauma data on predicting massive transfusion

BACKGROUND

Missing data are inherent in clinical research and may be especially problematic for trauma studies. This study describes a sensitivity analysis to evaluate the impact of missing data on clinical risk prediction algorithms. Three blood transfusion prediction models were evaluated using an observational trauma data set with valid missing data.

METHODS

The PRospective Observational Multicenter Major Trauma Transfusion (PROMMTT) study included patients requiring one or more unit of red blood cells at 10 participating US Level I trauma centers from July 2009 to October 2010. Physiologic, laboratory, and treatment data were collected prospectively up to 24 hours after hospital admission. Subjects who received 10 or more units of red blood cells within 24 hours of admission were classified as massive transfusion (MT) patients. Correct classification percentages for three MT prediction models were evaluated using complete case analysis and multiple imputation. A sensitivity analysis for missing data was conducted to determine the upper and lower bounds for correct classification percentages.

RESULTS

PROMMTT study enrolled 1,245 subjects. MT was received by 297 patients (24%). Missing percentage ranged from 2.2% (heart rate) to 45% (respiratory rate). Proportions of complete cases used in the MT prediction models ranged from 41% to 88%. All models demonstrated similar correct classification percentages using complete case analysis and multiple imputation. In the sensitivity analysis, correct classification upper-lower bound ranges per model were 4%, 10%, and 12%. Predictive accuracy for all models using PROMMTT data was lower than reported in the original data sets.

CONCLUSION

Evaluating the accuracy clinical prediction models with missing data can be misleading, especially with many predictor variables and moderate levels of missingness per variable. The proposed sensitivity analysis describes the influence of missing data on risk prediction algorithms. Reporting upper-lower bounds for percent correct classification may be more informative than multiple imputation, which provided similar results to complete case analysis in this study.

Seven deadly sins in trauma outcomes research: an epidemiologic post mortem for major causes of bias

BACKGROUND

Because randomized clinical trials in trauma outcomes research are expensive and complex, they have rarely been the basis for the clinical care of trauma patients. Most published findings are derived from retrospective and occasionally prospective observational studies that may be particularly susceptible to bias. The sources of bias include some common to other clinical domains, such as heterogeneous patient populations with competing and interdependent short- and long-term outcomes. Other sources of bias are unique to trauma, such as rapidly changing multisystem responses to injury that necessitate highly dynamic treatment regimens such as blood product transfusion. The standard research design and analysis strategies applied in published observational studies are often inadequate to address these biases.

METHODS

Drawing on recent experience in the design, data collection, monitoring, and analysis of the 10-site observational PRospective Observational Multicenter Major Trauma Transfusion (PROMMTT) study, 7 common and sometimes overlapping biases are described through examples and resolution strategies.

RESULTS

Sources of bias in trauma research include ignoring (1) variation in patients' indications for treatment (indication bias), (2) the dependency of intervention delivery on patient survival (survival bias), (3) time-varying treatment, (4) time-dependent confounding, (5) nonuniform intervention effects over time, (6) nonrandom missing data mechanisms, and (7) imperfectly defined variables. This list is not exhaustive.

CONCLUSION

The mitigation strategies to overcome these threats to validity require epidemiologic and statistical vigilance. Minimizing the highlighted types of bias in trauma research will facilitate clinical translation of more accurate and reproducible findings and improve the evidence-base that clinicians apply in their care of injured patients.

Evaluating the validity of multiple imputation for missing physiological data in the national trauma data bank

Background:

The National Trauma Data Bank (NTDB) is plagued by the problem of missing physiological data. The Glasgow Coma Scale score, Respiratory Rate and Systolic Blood Pressure are an essential part of risk adjustment strategies for trauma system evaluation and clinical research. Missing data on these variables may compromise the feasibility and the validity of trauma group comparisons.

Aims:

To evaluate the validity of Multiple Imputation (MI) for completing missing physiological data in the National Trauma Data Bank (NTDB), by assessing the impact of MI on 1) frequency distributions, 2) associations with mortality, and 3) risk adjustment.

Methods:

Analyses were based on 170,956 NTDB observations with complete physiological data (observed data set). Missing physiological data were artificially imposed on this data set and then imputed using MI (MI data set). To assess the impact of MI on risk adjustment, 100 pairs of hospitals were randomly selected with replacement and compared using adjusted Odds Ratios (OR) of mortality. OR generated by the observed data set were then compared to those generated by the MI data set.

Results:

Frequency distributions and associations with mortality were preserved following MI. The median absolute difference between adjusted OR of mortality generated by the observed data set and by the MI data set was 3.6% (inter-quartile range: 2.4%-6.1%).

Conclusions:

This study suggests that, provided it is implemented with care, MI of missing physiological data in the NTDB leads to valid frequency distributions, preserves associations with mortality, and does not compromise risk adjustment in inter-hospital comparisons of mortality.

Cancellations of (helicopter-transported) mobile medical team dispatches in the Netherlands

Background

The trauma centre of the Trauma Center Region North-West Netherlands (TRNWN) has consensus criteria for Mobile Medical Team (MMT) scene dispatch. The MMT can be dispatched by the EMS-dispatch centre or by the on-scene ambulance crew and is transported by helicopter or ground transport. Although much attention has been paid to improve the dispatch criteria, the MMT is often cancelled after being dispatched. The aim of this study was to assess the cancellation rate and the noncompliant dispatches of our MMT and to identify factors associated with this form of primary overtriage.

Methods

By retrospective analysis of all MMT dispatches in the period from 1 July 2006 till 31 December 2006 using chart review, we conducted a consecutive case review of 605 dispatches. Four hundred and sixty seven of these were included for our study, collecting data related to prehospital triage, patient’s condition on-scene and hospital course.

Results

Average age was 35.9 years; the majority of the patients were male (65.3%). Four hundred and thirty patients were victims of trauma, sustaining injuries in most cases from blunt trauma (89.3%). After being dispatched, the MMT was cancelled 203 times (43.5%). Statistically significant differences between assists and cancellations were found for overall mortality, mean RTS, GCS and ISS, mean hospitalization, length and amount of ICU admissions (p < 0.001). All dispatches were evaluated by using the MMT-dispatch criteria and mission appropriateness criteria. Almost 26% of all dispatches were neither appropriate, nor met the dispatch criteria. Fourteen missions were appropriate, but did not meet the dispatch criteria. The remaining 318 dispatches had met the dispatch criteria, of which 135 (30.3%) were also appropriate. The calculated additional costs of the cancelled dispatches summed up to a total of € 34,448, amounting to 2.2% of the total MMT costs during the study period.

Conclusion

In our trauma system, the MMT dispatches are involved with high rates of overtriage. After being dispatched, the MMT is cancelled in almost 50% of all cases. We found an undertriage rate of 4%, which we think is acceptable. All cancellations were justified. The additional costs of the cancelled missions were within an acceptable range. According to this study, it seems to be possible to reduce the overtriage rate of the MMT dispatches, without increasing the undertriage rate to non-acceptable levels.

A multiple imputation model for imputing missing physiologic data in the national trauma data bank

BACKGROUND

Like most trauma registries, the National Trauma Data Bank (NTDB) is limited by the problem of missing physiologic data. Multiple imputation (MI) has been proposed to simulate missing Glasgow Coma Scale (GCS) scores, respiratory rate (RR), and systolic blood pressure (SBP). The aim of this study was to develop an MI model for missing physiologic data in the NTDB and to provide guidelines for its implementation.

STUDY DESIGN

The NTDB 7.0 was restricted to patients admitted in 2005 with at least one anatomic injury code. A series of auxiliary variables thought to offer information for the imputation process was selected from the NTDB by literature review and expert opinion. The relation of these variables to physiologic variables and to the fact that they were missing was examined using logistic regression. The MI model included all auxiliary variables that had a statistically significant association with physiologic variables or with the fact that they were missing (Bonferroni-corrected p value <0.05).

RESULTS

The NTDB sample included 373,243 observations. Glasgow Coma Scale, respiratory rate, and systolic blood pressure were missing for 20.3%, 3.9%, and 8.5% of data observations, respectively. The MI model included information on the following: gender, age, anatomic injury severity, transfer status, injury mechanism, intubation status, alcohol and drug test results, emergency department disposition, total length of stay, ICU length of stay, duration of mechanical ventilation, and discharge disposition. The MI model offered good discrimination for predicting the value of physiologic variables and the fact that they were missing (areas under the receiver operating characteristic curve between 0.832 and 0.999).

CONCLUSIONS

This article proposes an MI model for imputing missing physiologic data in the NTDB and provides guidelines to facilitate its use. Implementation of the model should improve the quality of research involving the NTDB. The methodology can also be adapted to other trauma registries.

Despite overall low pediatric head injury mortality, disparities exist between races

BACKGROUND

To continually improve quality of care, it is important for centers caring for children with head injury to evaluate their overall outcome and that among diverse patient groups.

METHODS

Data on children with head injuries were extracted from the National Trauma Data Bank of the American College of Surgeons and our local trauma registry. Unadjusted mortality, as well as stratified analysis and logistic regression modeling, was used to evaluate overall and race-specific mortality.

RESULTS

There were 13,363 children in the National Trauma Data Base and 3111 in our database included. Our overall mortality rate compared favorably with the national data (3.2% vs 6.8%, P < .05). Our local data, however, showed a significant difference in mortality between white and African American (AA) children (2.2% vs 5.3%, P < .05), which was not identified in the national data. After stratification, the disparities by race persisted. Finally, multivariate regression modeling revealed that AA race was an independent predictor of mortality among our patient population, with an odds ratio of 3.1 (95% confidence interval, 1.2-7.8).

CONCLUSION

Despite excellent outcomes for children with head injuries, we have uncovered unsettling inequities between AA and white children. These findings support the need to evaluate outcomes among specific groups to identify disparities that require further careful investigation.

Managing the common problem of missing data in trauma studies

PURPOSE

To provide guidance for managing the problem of missing data in clinical studies of trauma in order to decrease bias and increase the validity of findings for subsequent use.

ORGANIZING CONSTRUCT

A thoughtful approach to missing data is an essential component of analysis to promote the clear interpretation of study findings.

METHODS

Integrative review of relevant biostatistics, medical and nursing literature, and case exemplars of missing data analyses using multiple linear regression based upon data from the National Study on the Costs and Outcomes of Trauma (NSCOT) was used as an example.

FINDINGS AND CONCLUSIONS

In studies of traumatically injured people, multiple imputed values are often superior to complete case analyses that might have significant bias. Multiple imputation can improve accuracy of the assessment and might also improve precision of estimates. Sensitivity analyses which implements repeated analyses using various scenarios may also be useful in providing information supportive of further inquiry. This stepwise approach of missing data could also be valid in studies with similar types or patterns of missing data.

CLINICAL RELEVANCE

In interpreting and applying findings of studies with missing data, clinicians need to ensure that researchers have used appropriate methods for handling this issue. If suitable methods were not employed, nurse clinicians need to be aware that the findings may be biased.

Predicting Discharge of Trauma Survivors to Rehabilitation

OBJECTIVES

To conduct a population-based survey among trauma survivors on accessibility to rehabilitation services in metropolitan, urban, and rural areas in Quebec (Canada), we attempted to use trauma registries as a sampling frame of subjects discharged to rehabilitation. Discharge destinations were inaccurate in many registries, preventing straightforward identification of the survey subjects. Using the best registry data, we aimed to identify predictors of rehabilitation discharge and to use them to specify a reliable sampling frame for the survey.

DESIGN

A logistic predictive model of rehabilitation discharge was developed. This model was applied to data from metropolitan, urban, and rural trauma centers to identify all subjects predicted to be discharged to a rehabilitation facility.

RESULTS

Age, acute-care length of stay, injury-severity score, lower-limb injuries, and seven other predictors were included in the model that generated an area under the ROC curve (AUC) of 0.83 and a classification accuracy of 76.6%. The metropolitan, urban, and rural frames were slightly different. They included, respectively, 808, 798, and 929 subjects.

CONCLUSIONS

The procedure helped us bypass largely inaccurate data from trauma registries. The sampling frames reflected severely injured trauma survivors who were likely to have been referred to postacute rehabilitation.

Endotracheal intubation increases out-of-hospital time in trauma patients

OBJECTIVES

Prior efforts have linked field endotracheal intubation (ETI) with increased out of hospital (OOH) time, but it is not clear if the additional time delay is due to the procedure, patient acuity, or transport distance. We sought to assess the difference in OOH time among trauma patients with and without OOH-ETI after accounting for distance and other clinical variables.

METHODS

Retrospective cohort analysis of trauma patients 14 years or older transported by ground or air to one of two Level 1 trauma centers from January 2000 to December 2003. Geographical data were probabilistically linked to trauma registry records for transport distance. Trauma registry OOH time (interval from 9-1-1 call to hospital arrival) was validated against a subset of linked ambulance records using Bland-Altman plots and tested by using the Spearman rank correlation coefficient. Based on the validation, the sample was restricted to patients with OOH time 100 minutes or less. The propensity for OOH-ETI was calculated by using field vital signs, demographics, mechanism, transport mode, comorbidities, Abbreviated Injury Scale head injury 3 or greater, injury severity score, blood transfusion, and major surgery. Multivariable linear regression (outcome = total OOH time) was used to assess the time increase (minutes) associated with OOH-ETI after adjusting for distance, propensity for OOH-ETI, and mode of transport.

RESULTS

A total of 8,707 patients were included in the analysis, of which 570 (6.5%) were intubated in the field. Adjusted only for distance, OOH times averaged 6.1 minutes longer (95% CI 4.2-7.9) among patients intubated with RSI. After including other covariates, OOH time was 10.7 minutes (95% CI 7.7-13.8) longer among patients with RSI and 5.2 minutes (95% CI 2.2-8.1) longer among patients with conventional ETI. The time difference was greatest farther from the hospital.

CONCLUSIONS

Patients with OOH-ETI have increased total OOH time, especially among those using RSI, even after accounting for distance and other clinical factors. Injured patients may benefit from airway management techniques that require less time for execution.

Disparities in child abuse mortality are not explained by injury severity

BACKGROUND

Unadjusted abuse-related mortality has been demonstrated to be nearly 4-fold higher for African American (AA) children. Little is known about the etiology of this disparity. This study examines the importance of injury severity and initial presentation in explaining the observed disparity.

METHOD

Our trauma database was reviewed to identify all abused patients admitted over a 10-year period. Outcomes among white and AA children were compared with specific attention to injury severity scores and initial presentation. Logistic regression and Cox proportional hazard analyses were performed to evaluate the impact of race on outcome.

RESULTS

There were 443 abused children identified. Thirty-eight percent of the group was AA. The overall mortality was 7.7%; however, the AA mortality was significantly higher than white children (14.8% vs 3.3%; P < .05). After controlling for injury severity and physiology at presentation, the odds ratio of mortality for an AA child was 9.14 (95% confidence interval, 1.97-42.43). Survival analysis confirmed the disparity after revealing a hazard ratio of dying for AA children of 6.51 (95% confidence interval, 2.74-15.47) compared with white children.

CONCLUSION

Despite attempts to control for the clinical presentation and injury severity of abused children, significant differences in mortality persist between AA and white children.

Predicting hospital mortality among injured children using a national trauma database

PURPOSE

The purpose of this study was to develop a model that accurately predicts mortality among injured children based on components of the initial patient evaluation and that is generalizable to diverse acute care settings. Important predictive variables obtained in an emergency setting are frequently missing in even large national databases, limiting their effectiveness for developing predictions. In this study, a model predicting pediatric trauma mortality was developed using a national database and methods to handle missing data that may avoid biases that can occur restricting analyses to complete cases.

METHODS

Records of pediatric patients included in the National Pediatric Trauma Registry (NPTR) between 1996 and 1999 were used as a training set in a logistic regression model to predict hospital mortality using vital signs, Glasgow Coma Scale (GCS) score, and intubation status. Multiple imputation was applied to handle missing data. The model was tested using independent data from the NPTR and National Trauma Data Bank (NTDB).

RESULTS

Complete case analysis identified only GCS-eye and intubation status as predictors of mortality. A model based on complete case analysis had good discrimination (c-index = 0.784) and excellent calibration (Hosmer-Lemeshow c-statistic, 6.8) (p > 0.05). Using multiple imputation, three additional predictors of mortality (systolic blood pressure, pulse, and GCS-motor) were identified and improved model performance was observed. The model developed using multiple imputation had excellent discrimination (c-index, 0.947-0.973) in both test datasets. Calibration was better in the NPTR testing set than in the NTDB (Hosmer-Lemeshow c-statistic, 9.2 for NPTR [p > 0.05] and 258.2 for NTDB [p < 0.05]). At a probability cutoff that minimized misclassification in the training set, the false-negative and false-negative rates of the model were better than those obtained with either the Revised Trauma Score (RTS) or Pediatric Trauma Score using data from the NPTR testing set. Although the false-positive rates were lower with the RTS using data from the NTDB, the false-negative rates of the proposed model and the RTS were similar in this test dataset.

CONCLUSIONS

Using multiple imputation to handle missing data, a model predicting pediatric trauma mortality was developed that compared favorably with existing trauma scores. Application of these methods may produce predictive trauma models that are more statistically reliable and applicable in clinical practice.