The accuracy of clinical information systems

Clinical Outcomes and Mortality Impact of Hyperbaric Oxygen Therapy in Patients With Carbon Monoxide Poisoning

OBJECTIVES

Carbon monoxide poisoning affects 50,000 per year in the United States alone. Mortality is approximately 3%, and up to 40% of survivors suffer from permanent neurocognitive and affective deficits. Hyperbaric oxygen therapy has shown benefit on reducing the long-term neurologic sequelae of carbon monoxide poisoning but has not demonstrated improved survival. The objective of this study is to assess the efficacy of hyperbaric oxygen for acute and long-term mortality in carbon monoxide poisoning using a large clinical databank.

DESIGN

Retrospective analysis.

SETTING

University of Pittsburgh Medical Center healthcare system (Pittsburgh, PA).

PATIENTS

One-thousand ninety-nine unique encounters of adult patients with carbon monoxide poisoning.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Baseline demographics, laboratory values, hospital charge transactions, discharge disposition, and clinical information from charting were obtained from the electronic medical record. In propensity-adjusted analysis, hyperbaric oxygen therapy was associated with a reduction in inpatient mortality (absolute risk reduction, 2.1% [3.7-0.9%]; p = 0.001) and a reduction in 1-year mortality (absolute risk reduction, 2.1% [3.8-0.4%]; p = 0.013).

CONCLUSIONS

These data demonstrate that hyperbaric oxygen is associated with reduced acute and reduced 1-year mortality. Further studies are needed on the mortality effects of hyperbaric oxygen therapy in carbon monoxide poisoning.

Secondary EMR data for quality improvement and research: A comparison of manual and electronic data collection from an integrated critical care electronic medical record system

PURPOSE

This study measured the quality of data extracted from a clinical information system widely used for critical care quality improvement and research.

MATERIALS AND METHODS

We abstracted data from 30 fields in a random sample of 207 patients admitted to nine adult, medical-surgical intensive care units. We assessed concordance between data collected: (1) manually from the bedside system (eCritical MetaVision) by trained auditors, and (2) electronically from the system data warehouse (eCritical TRACER). Agreement was assessed using Cohen's Kappa for categorical variables and intraclass correlation coefficient (ICC) for continuous variables.

RESULTS

Concordance between data sets was excellent. There was perfect agreement for 11/30 variables (35%). The median Kappa score for the 16 categorical variables was 0.99 (IQR 0.92-1.00). APACHE II had an ICC of 0.936 (0.898-0.960). The lowest concordance was observed for SOFA renal and respiratory components (ICC 0.804 and 0.846, respectively). Score translation errors by the manual auditor were the most common source of data discrepancies.

CONCLUSIONS

Manual validation processes of electronic data are complex in comparison to validation of traditional clinical documentation. This study represents a straightforward approach to validate the use of data repositories to support reliable and efficient use of high quality secondary use data.

Assessment and perceptions of intensive care data quality, reporting and use: a survey of ICU directors

It is becoming increasingly common for government bodies, healthcare providers, funders and consumers to seek measures of the quality of critical care. It is important to ensure the quality of intensive care unit (ICU) data is high so these stakeholders can confidently use quality of care measures in decision-making. This paper aims to evaluate the quality of data collected for and submitted to the Australian and New Zealand Intensive Care Society Adult Patient Database, and to investigate the perceptions of NSW ICU directors in relation to ICU data quality, reporting and usage. A survey tool was developed based on an existing framework that consisted of procedures for assessing data quality in medical registries. The survey was distributed to the directors of all NSW ICUs that submitted data in the 2007/2008 financial year. Overall, completeness of the data and its quality was perceived to be good. Participants were less likely to engage in activities involving the detection and correction of data errors, feedback of data or use of data for local purposes. A number of barriers and enablers to good quality ICU data as well as strategies to improve data quality were identified. Inadequate staff, training and resources for data collection were widespread concerns. NSW ICU directors believe more work is required to achieve high quality data and appropriate use of the data collected. Strategies targeting increased resources including updated technology and improved staffing and training, as well as low-cost solutions such as audit, feedback and clinician engagement, have been highlighted.

[Qualitative evaluation of blood products records in a hospital]

PURPOSE OF THE STUDY

This study aimed at evaluating the qualitative performance of blood products traceability from paper and electronic medical records in a hospital.

STUDY DESIGN

Quality of date/time documentation was assessed by detection, for 20minutes or more, of chronological errors and inter-source inconsistencies, in a random sample of 168 blood products transfused during 2009.

RESULTS

A receipt date/time was confirmed in 52% of paper records; a data entry error was attested in 25% of paper records, and 21% of electronic records. A transfusion date/time was notified in 93% of paper records, with a data entry error in 26% of paper records and 25% of electronic records. The patient medical record held at least one date/time error in 18% and 17%, for receipt and transfusion respectively. Environmental factors (clinical setting, urgency, blood product category) did not contributed to data error rates.

CONCLUSION

Although blood products traceability has good quantitative results, the recorded documentation is not qualitative. In our study, data entry errors are similar in electronic or paper records, but the global failure rate is lesser in electronic records because omissions are controlled.

Doppler-defined pulmonary hypertension in medical intensive care unit patients: Retrospective investigation of risk factors and impact on mortality

Pulmonary hypertension (PH) is poorly characterized in the critically ill. No prior studies describe the burden of or outcomes associated with PH in a general medical intensive care unit population. We hypothesize that PH is an important comorbidity prevalent in the modern medical intensive care unit. We undertook a preliminary investigation to define the consequences of Doppler-defined PH in the critically ill. A single-center retrospective case–control study of medical intensive care patients admitted over a 1-year period was conducted. Eligible patients had an echocardiogram within 4 days of admission. PH was defined to include both pulmonary arterial and venous hypertension and required a tricuspid regurgitant jet velocity ≥3 m/sec. Cases and controls were compared for comorbidities, illness severity, diagnoses, and mortality. Multivariable regression was performed to identify clinical features associated with PH and mortality. 299 (21% of admissions) patients had an eligible echocardiogram. Patients with PH (N=126) had a higher unadjusted mortality than did controls (N=173) (37% vs. 25%, P=0.04) and PH remained significantly associated with mortality after controlling for other clinical factors (HR=1.59, 95% CI=1.03–2.44, P=0.036). Low ejection fraction (OR=2.21, 95% CI=1.19–4.11, P=0.012) and pulmonary embolism (OR=4.28, 95% CI=1.59–11.5, P=0.004) were independently associated with PH. Doppler-defined PH is associated with mortality in the critically ill. Prospective studies are needed to define the prevalence of pulmonary venous hypertension versus pulmonary arterial hypertension, and the clinical consequences of each, in a general medical intensive care unit population.

Machine learning techniques to examine large patient databases

Computerization in healthcare in general, and in the operating room (OR) and intensive care unit (ICU) in particular, is on the rise. This leads to large patient databases, with specific properties. Machine learning techniques are able to examine and to extract knowledge from large databases in an automatic way. Although the number of potential applications for these techniques in medicine is large, few medical doctors are familiar with their methodology, advantages and pitfalls. A general overview of machine learning techniques, with a more detailed discussion of some of these algorithms, is presented in this review.

The Surgical Intensive Care-infection Registry: a research registry with daily clinical support capabilities

Infections in the surgical and trauma intensive care unit (STICU) are responsible for significant patient morbidity and mortality. Research into these infectious complications often uses administrative databases or clinical information systems designed for documenting and billing daily patient care. Neither of these sources is intended for research, and many investigators have questioned their accuracy. The Surgical Intensive Care-Infection Registry (SIC-IR) was developed as a research data repository to use to monitor STICU infections. SIC-IR is a relational database application designed to collect quality data and to integrate with daily patient care. SIC-IR prospectively collects and archives more than 100 clinical variables daily on each STICU patient to ensure completeness and correctness of the registry. Furthermore, SIC-IR aids in clinical activities by providing patient summaries and medical record documentation. SIC-IR provides accurate data for STICU infection research and enables the users to easily undertake quality-of-care improvement initiatives.