Linked Records of Children with Traumatic Brain Injury. Probabilistic Linkage without Use of Protected Health Information

Probabilistic Linkage Creates a Novel Database to Study Bronchiolitis Care in the PICU

OBJECTIVES

Lack of a comprehensive database containing diagnosis, patient and clinical characteristics, diagnostics, treatments, and outcomes limits needed comparative effectiveness research (CER) to improve care in the PICU. Combined, the Pediatric Hospital Information System (PHIS) and Virtual Pediatric Systems (VPS) databases contain the needed data for CER, but limits on the use of patient identifiers have thus far prevented linkage of these databases with traditional linkage methods. Focusing on the subgroup of patients with bronchiolitis, we aim to show that probabilistic linkage methods accurately link data from PHIS and VPS without the need for patient identifiers to create the database needed for CER.

METHODS

We used probabilistic linkage to link PHIS and VPS records for patients admitted to a tertiary children's hospital between July 1, 2017 to June 30, 2019. We calculated the percentage of matched records, rate of false-positive matches, and compared demographics between matched and unmatched subjects with bronchiolitis.

RESULTS

We linked 839 of 920 (91%) records with 4 (0.5%) false-positive matches. We found no differences in age (P = .76), presence of comorbidities (P = .16), admission illness severity (P = .44), intubation rate (P = .41), or PICU stay length (P = .36) between linked and unlinked subjects.

CONCLUSIONS

Probabilistic linkage creates an accurate and representative combined VPS-PHIS database of patients with bronchiolitis. Our methods are scalable to join data from the 38 hospitals that jointly contribute to PHIS and VPS, creating a national database of diagnostics, treatment, outcome, and patient and clinical data to enable CER for bronchiolitis and other conditions cared for in the PICU.

Comparison of record linkage software for deduplicating patient identities in California's Prescription Drug Monitoring Program

BACKGROUND

To help prevent overdose deaths involving prescription drugs, accurate linkage of prescription drug monitoring program (PDMP) records for individual patients is essential.

OBJECTIVES

To compare the accuracy of the linkage program used by California's PDMP against various record linkage programs with respect to accuracy in deduplicating patient identities in the PDMP, with implications for identifying high-risk opioid use and outlier behaviors.

RESEARCH DESIGN

We evaluated California's program, Link Plus, LinkSolv, and The Link King on 557 861 PDMP identity records with addresses in two 3-digit zip code areas for patients who filled a controlled substance prescription in 2013. Manual review was performed on a stratified sample of 720 paired records identified as matches by at least one program.

MEASURES

We estimated sensitivity and positive predictive value, and computed PDMP patient alerts for the patient entities identified by each program.

RESULTS

Sensitivity was 95% for LinkSolv and The Link King, 84% for Link Plus, and 73% for California's program; positive predictive value was ≥93% for all programs. The number of patient entities prompting a PDMP alert was similar among the programs for all alerts except multiple provider episodes (obtaining prescriptions from ≥6 prescribers or ≥6 pharmacies in the last 6 months), which were 10.9%, 26.6%, and 16.9% greater using The Link King, Link Plus, and LinkSolv, respectively, compared to California's program.

CONCLUSIONS

PDMPs should assess the accuracy of record linkage algorithms and the impacts of these algorithms on patient safety alerts and develop national best practices for PDMP record linkage.

Novel Claims-Based Outcome Phenotypes in Survivors of Pediatric Traumatic Brain Injury

OBJECTIVE

For children hospitalized with acute traumatic brain injury (TBI), to use postdischarge insurance claims to identify: (1) healthcare utilization patterns representative of functional outcome phenotypes and (2) patient and hospitalization characteristics that predict outcome phenotype.

SETTING

Two pediatric trauma centers and a state-level insurance claim aggregator.

PATIENTS

A total of 289 children, who survived a hospitalization after TBI between 2009 and 2014, were in the hospital trauma registry, and had postdischarge insurance eligibility.

DESIGN

Retrospective cohort study.

MAIN MEASURES

Unsupervised machine learning to identify phenotypes based on postdischarge insurance claims. Regression analyses to identify predictors of phenotype.

RESULTS

Median age 5 years (interquartile range 2-12), 29% (84/289) female. TBI severity: 30% severe, 14% moderate, and 60% mild. We identified 4 functional outcome phenotypes. Phenotypes 3 and 4 were the highest utilizers of resources. Morbidity burden was highest during the first 4 postdischarge months and subsequently decreased in all domains except respiratory. Severity and mechanism of injury, intracranial pressure monitor placement, seizures, and hospital and intensive care unit lengths of stay were phenotype predictors.

CONCLUSIONS

Unsupervised machine learning identified postdischarge phenotypes at high risk for morbidities. Most phenotype predictors are available early in the hospitalization and can be used for prognostic enrichment of clinical trials targeting mitigation or treatment of domain-specific morbidities.

Joining Datasets Without Identifiers: Probabilistic Linkage of Virtual Pediatric Systems and PEDSnet

OBJECTIVES

To 1) probabilistically link two important pediatric data sources, Virtual Pediatric Systems and PEDSnet, 2) evaluate linkage accuracy overall and in patients with severe sepsis or septic shock, and 3) identify variables important to linkage accuracy.

DESIGN

Retrospective linkage of prospectively collected datasets from Virtual Pediatrics Systems, Inc (Los Angeles, CA) and the PEDSnet consortium.

SETTING

Single-center academic PICU.

PATIENTS

All PICU encounters between January 1, 2012, and December 31, 2017, that were deterministically matched between the two datasets.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We abstracted records from Virtual Pediatric Systems and PEDSnet corresponding to PICU encounters and probabilistically linked using 44 features shared by the two datasets. We generated a gold standard deterministic linkage using protected health information elements, which were then removed from datasets. We then calculated candidate pair log-likelihood ratios for all pairs of subjects and selected optimal pairs in a two-stage algorithm. A total of 22,051 gold standard PICU encounter pairs were identified over the study period. The optimal linkage model demonstrated excellent discrimination (area under the receiver operating characteristic curve > 0.99); 19,801 cases (89.9%) were matched with 13 false positives. The addition of two protected health information dates (admission month, birth day-of-year) increased to 20,189 (91.6%) the cases matched, with three false positives. Restricting to patients with Virtual Pediatric Systems diagnosis of severe sepsis or septic shock (n = 1,340 [6.1%]) matched 1,250 cases (93.2%) with zero false positives. Increased number of laboratory values present in the first 12 hours of admission significantly increased log-likelihood ratios, suggesting stronger candidate pair matching.

CONCLUSIONS

We demonstrated the use of probabilistic linkage to accurately join two complementary pediatric critical care datasets at a single academic PICU in the absence of protected health information. Combining datasets with curated diagnoses and granular measurements can validate patient acuity metrics and facilitate multicenter machine learning algorithms. We anticipate these methods will generalize to other common PICU diagnoses.

Review of Clinical Research Informatics

OBJECTIVES

Clinical Research Informatics (CRI) declares its scope in its name, but its content, both in terms of the clinical research it supports-and sometimes initiates-and the methods it has developed over time, reach much further than the name suggests. The goal of this review is to celebrate the extraordinary diversity of activity and of results, not as a prize-giving pageant, but in recognition of the field, the community that both serves and is sustained by it, and of its interdisciplinarity and its international dimension.

METHODS

Beyond personal awareness of a range of work commensurate with the author's own research, it is clear that, even with a thorough literature search, a comprehensive review is impossible. Moreover, the field has grown and subdivided to an extent that makes it very hard for one individual to be familiar with every branch or with more than a few branches in any depth. A literature survey was conducted that focused on informatics-related terms in the general biomedical and healthcare literature, and specific concerns ("artificial intelligence", "data models", "analytics", etc.) in the biomedical informatics (BMI) literature. In addition to a selection from the results from these searches, suggestive references within them were also considered.

RESULTS

The substantive sections of the paper-Artificial Intelligence, Machine Learning, and "Big Data" Analytics; Common Data Models, Data Quality, and Standards; Phenotyping and Cohort Discovery; Privacy: Deidentification, Distributed Computation, Blockchain; Causal Inference and Real-World Evidence-provide broad coverage of these active research areas, with, no doubt, a bias towards this reviewer's interests and preferences, landing on a number of papers that stood out in one way or another, or, alternatively, exemplified a particular line of work.

CONCLUSIONS

CRI is thriving, not only in the familiar major centers of research, but more widely, throughout the world. This is not to pretend that the distribution is uniform, but to highlight the potential for this domain to play a prominent role in supporting progress in medicine, healthcare, and wellbeing everywhere. We conclude with the observation that CRI and its practitioners would make apt stewards of the new medical knowledge that their methods will bring forward.

Probabilistic record linkage of de-identified research datasets with discrepancies using diagnosis codes

We develop an algorithm for probabilistic linkage of de-identified research datasets at the patient level, when only diagnosis codes with discrepancies and no personal health identifiers such as name or date of birth are available. It relies on Bayesian modelling of binarized diagnosis codes, and provides a posterior probability of matching for each patient pair, while considering all the data at once. Both in our simulation study (using an administrative claims dataset for data generation) and in two real use-cases linking patient electronic health records from a large tertiary care network, our method exhibits good performance and compares favourably to the standard baseline Fellegi-Sunter algorithm. We propose a scalable, fast and efficient open-source implementation in the ludic R package available on CRAN, which also includes the anonymized diagnosis code data from our real use-case. This work suggests it is possible to link de-identified research databases stripped of any personal health identifiers using only diagnosis codes, provided sufficient information is shared between the data sources.

Hospital Readmissions After Pediatric Trauma

OBJECTIVES

To determine the rate, etiology, and timing of unplanned and planned hospital readmissions and to identify risk factors for unplanned readmission in children who survive a hospitalization for trauma.

DESIGN

Multicenter retrospective cohort study of a probabilistically linked dataset from the National Trauma Data Bank and the Pediatric Health Information System database, 2007-2012.

SETTING

Twenty-nine U.S. children's hospitals.

PATIENTS

51,591 children (< 18 yr at admission) who survived more than or equal to a 2-day hospitalization for trauma.

MEASUREMENTS AND MAIN RESULTS

The primary outcome was unplanned readmission within 1 year of discharge from the injury hospitalization. Secondary outcomes included any readmission, reason for readmission, time to readmission, and number of readmissions within 1 year of discharge. The primary exposure groups were isolated traumatic brain injury, both traumatic brain injury and other injury, or nontraumatic brain injury only. We hypothesized a priori that any traumatic brain injury would be associated with both planned and unplanned hospital readmission. We used All Patient Refined Diagnosis Related Groups codes to categorize readmissions by etiology and planned or unplanned. Overall, 4,301/49,982 of the patients (8.6%) with more than or equal to 1 year of observation time were readmitted to the same hospital within 1 year. Many readmissions were unplanned: 2,704/49,982 (5.4%) experienced an unplanned readmission in the first year. The most common reason for unplanned readmission was infection (22%), primarily postoperative or posttraumatic infection (38% of readmissions for infection). Traumatic brain injury was associated with lower odds of unplanned readmission in multivariable analyses. Seizure or RBC transfusion during the index hospitalization were the strongest predictors of unplanned, earlier, and multiple readmissions.

CONCLUSIONS

Many survivors of pediatric trauma experience unplanned, and potentially preventable, hospital readmissions in the year after discharge. Identification of those at highest risk of readmission can guide targeted in-hospital or postdischarge interventions.

Functional Outcome After Intracranial Pressure Monitoring for Children With Severe Traumatic Brain Injury

IMPORTANCE

Intracranial pressure (ICP) monitoring is a mainstay of therapy for children with traumatic brain injury (TBI), but its overall association with patient outcome is unclear.

OBJECTIVE

To test the hypothesis that ICP monitoring is associated with improved functional survival of children with severe TBI.

DESIGN, SETTING, AND PARTICIPANTS

A propensity-weighted effectiveness analysis was conducted using 2 linked national databases with data from 30 US children's hospitals from January 1, 2007, to December 31, 2012, on 3084 children with severe TBI. Clinical events including neurosurgical procedures were identified using validated computable phenotypes. Data analysis was conducted from September 1, 2016, to March 1, 2017.

EXPOSURE

Placement of an ICP monitor.

MAIN OUTCOMES AND MEASURES

A composite of hospital mortality, discharge to hospice, or survival with placement of new tracheostomy and gastrostomy tubes.

RESULTS

Of the 3084 children in the study (1128 girls and 1956 boys; mean [SD] age, 7.03 [5.44] years), 1002 (32.4%) underwent ICP monitoring, with substantial hospital variation (6% to 50% by hospital). Overall, 484 children (15.7%) experienced the primary composite outcome. A propensity approach using matching weights generated good covariate balance between those who did and those who did not undergo ICP monitoring. Using a propensity-weighted logistic regression model clustered by hospital, no statistically significant difference was found in functional survival between monitored and unmonitored patients (odds ratio of poor outcome among those who underwent ICP monitoring, 1.31; 95% CI, 0.99-1.74). In a prespecified secondary analysis, no difference in mortality was found (odds ratio, 1.16; 95% CI, 0.89-1.50). Prespecified subgroup analyses of children younger and older than 2 years of age and among those with unintentional and inflicted (intentional) injuries also showed no difference in outcome with ICP monitoring.

CONCLUSIONS AND RELEVANCE

With the use of linked national data and validated computable phenotypes, no evidence was found of a benefit from ICP monitoring on functional survival of children with severe TBI. Intracranial pressure monitoring is a widely but inconsistently used technology with incompletely demonstrated effectiveness. A large prospective cohort study or randomized trial is needed.

Development and Prospective Validation of Tools to Accurately Identify Neurosurgical and Critical Care Events in Children With Traumatic Brain Injury

OBJECTIVE

To develop and validate case definitions (computable phenotypes) to accurately identify neurosurgical and critical care events in children with traumatic brain injury.

DESIGN

Prospective observational cohort study, May 2013 to September 2015.

SETTING

Two large U.S. children's hospitals with level 1 Pediatric Trauma Centers.

PATIENTS

One hundred seventy-four children less than 18 years old admitted to an ICU after traumatic brain injury.

MEASUREMENTS AND MAIN RESULTS

Prospective data were linked to database codes for each patient. The outcomes were prospectively identified acute traumatic brain injury, intracranial pressure monitor placement, craniotomy or craniectomy, vascular catheter placement, invasive mechanical ventilation, and new gastrostomy tube or tracheostomy placement. Candidate predictors were database codes present in administrative, billing, or trauma registry data. For each clinical event, we developed and validated penalized regression and Boolean classifiers (models to identify clinical events that take database codes as predictors). We externally validated the best model for each clinical event. The primary model performance measure was accuracy, the percent of test patients correctly classified. The cohort included 174 children who required ICU admission after traumatic brain injury. Simple Boolean classifiers were greater than or equal to 94% accurate for seven of nine clinical diagnoses and events. For central venous catheter placement, no classifier achieved 90% accuracy. Classifier accuracy was dependent on available data fields. Five of nine classifiers were acceptably accurate using only administrative data but three required trauma registry fields and two required billing data.

CONCLUSIONS

In children with traumatic brain injury, computable phenotypes based on simple Boolean classifiers were highly accurate for most neurosurgical and critical care diagnoses and events. The computable phenotypes we developed and validated can be used in any observational study of children with traumatic brain injury and can reasonably be applied in studies of these interventions in other patient populations.

EEG Monitoring and Antiepileptic Drugs in Children with Severe TBI

BACKGROUND

Traumatic brain injury (TBI) causes substantial morbidity and mortality in US children. Post-traumatic seizures (PTS) occur in 11-42% of children with severe TBI and are associated with unfavorable outcome. Electroencephalographic (EEG) monitoring may be used to detect PTS and antiepileptic drugs (AEDs) may be used to treat PTS, but national rates of EEG and AED use are not known. The purpose of this study was to describe the frequency and timing of EEG and AED use in children hospitalized after severe TBI.

METHODS

Retrospective cohort study of 2165 children at 30 hospitals in a probabilistically linked dataset from the National Trauma Data Bank (NTDB) and the Pediatric Health Information Systems (PHIS) database, 2007-2010. We included children (age <18 years old at admission) with linked NTDB and PHIS records, severe (Emergency Department [ED] Glasgow Coma Scale [GCS] <8) TBI, hospital length of stay >24 h, and non-missing disposition. The primary outcomes were EEG and AED use.

RESULTS

Overall, 31.8% of the cohort had EEG monitoring. Of those, 21.8% were monitored on the first hospital day. The median duration of EEG monitoring was 2.0 (IQR 1.0, 4.0) days. AEDs were prescribed to 52.0% of the cohort, of whom 61.8% received an AED on the first hospital day. The median duration of AED use was 8.0 (IQR 4.0, 17.0) days. EEG monitoring and AED use were more frequent in children with known risk factors for PTS. EEG monitoring and AED use were not related to hospital TBI volume.

CONCLUSION

EEG use is relatively uncommon in children with severe TBI, but AEDs are frequently prescribed. EEG monitoring and AED use are more common in children with known risk factors for PTS.

Seizures in Children With Severe Traumatic Brain Injury

OBJECTIVE

Traumatic brain injury causes substantial morbidity and mortality in children. Posttraumatic seizures may worsen outcomes after traumatic brain injury. Posttraumatic seizures risk factors are not completely understood. Our objective was to clarify posttraumatic seizures risk factors in a large cohort of children with severe traumatic brain injury.

DESIGN

Retrospective cohort study of a probabilistically linked dataset from the National Trauma Data Bank and the Pediatric Health Information Systems database, 2007-2010.

SETTING

Twenty-nine U.S. children's hospitals.

PATIENTS

A total of 2,122 children (age, < 18 yr old at admission) with linked National Trauma Data Bank and Pediatric Health Information Systems records, severe (emergency department Glasgow Coma Scale, < 8) traumatic brain injury, hospital length of stay more than 24 hours, and nonmissing disposition.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

The outcome was posttraumatic seizures, identified using validated International Classification of Diseases, 9th Revision, Clinical Modification diagnosis codes. Prespecified candidate predictors of posttraumatic seizures included age, injury mechanism, emergency department Glasgow Coma Scale, intracranial hemorrhage type, hypoxia, hypotension, and cardiac arrest. Posttraumatic seizures were diagnosed in 25.2% of children with severe traumatic brain injury. In those without abuse/assault or subdural hemorrhage, the posttraumatic seizures rate varied between 36.6% in those less than 2 years old and 16.4% in those 14-17 years old. Age, abusive mechanism, and subdural hemorrhage are each significant predictors of posttraumatic seizures. The risk of posttraumatic seizures has a complex relationship with these predictors. The estimated odds of posttraumatic seizures decrease with advancing age, odds ratio equal to 0.929 (0.905-0.954) per additional year of age with no abuse/assault and no subdural hemorrhage; odds ratio equal to 0.820 (0.730-0.922) per additional year of age when abuse and subdural hemorrhage are present. An infant with accidental traumatic brain injury and subdural hemorrhage has approximately the same estimated probability of posttraumatic seizures as an abused infant without subdural hemorrhage (47% [95% CI, 39-55%] vs 50% [95% CI, 41-58%]; p = 0.69). The triad of young age, injury by abuse/assault, and subdural hemorrhage confers the greatest estimated probability for posttraumatic seizures (60% [95% CI, 53-66%]).

CONCLUSIONS

Posttraumatic seizures risk in children with severe traumatic brain injury is greatest with a triad of younger age, injury by abuse/assault, and subdural hemorrhage. However, posttraumatic seizures are common even in the absence of these factors.

Clinical Databases and Registries in Congenital and Pediatric Cardiac Surgery, Cardiology, Critical Care, and Anesthesiology Worldwide

The growth in large-scale data management capabilities and the successful care of patients with congenital heart defects have coincidentally paralleled each other for the last three decades, and participation in multicenter congenital heart disease databases and registries is now a fundamental component of cardiac care. This manuscript attempts for the first time to consolidate in one location all of the relevant databases worldwide, including target populations, specialties, Web sites, and participation information. Since at least 1,992 cardiac surgeons and cardiologists began leveraging this burgeoning technology to create multi-institutional data collections addressing a variety of specialties within this field. Pediatric heart diseases are particularly well suited to this methodology because each individual care location has access to only a relatively limited number of diagnoses and procedures in any given calendar year. Combining multiple institutions data therefore allows for a far more accurate contemporaneous assessment of treatment modalities and adverse outcomes. Additionally, the data can be used to develop outcome benchmarks by which individual institutions can measure their progress against the field as a whole and focus quality improvement efforts in a more directed fashion, and there is increasing utilization combining clinical research efforts within existing data structures. Efforts are ongoing to support better collaboration and integration across data sets, to improve efficiency, further the utility of the data collection infrastructure and information collected, and to enhance return on investment for participating institutions.

Advancing the Use of Administrative Data for Emergency Department Diagnostic Imaging Research

Administrative data are critical to describing patterns of use, cost, and appropriateness of imaging in emergency care. These data encompass a range of source materials that have been collected primarily for a nonresearch use: documenting clinical care (e.g., medical records), administering care (e.g., picture archiving and communication systems), or financial transactions (e.g., insurance claims). These data have served as the foundation for large, descriptive studies that have documented the rise and expanded role of diagnostic imaging in the emergency department (ED). This article summarizes the discussions of the breakout session on the use of administrative data for emergency imaging research at the May 2015 Academic Emergency Medicine consensus conference, "Diagnostic Imaging in the Emergency Department: A Research Agenda to Optimize Utilization." The authors describe the areas where administrative data have been applied to research evaluating the use of diagnostic imaging in the ED, the common sources for these data, and the strengths and limitations of administrative data. Next, the future role of administrative data is examined for answering key research questions in an evolving health system increasingly focused on measuring appropriateness, ensuring quality, and improving value for health spending. This article specifically focuses on four thematic areas: data quality, appropriateness and value, special populations, and policy interventions.