Using electronic medical records to increase the efficiency of catheter-associated urinary tract infection surveillance for National Health and Safety Network reporting

Development and validation of models for detection of postoperative infections using structured electronic health records data and machine learning

BACKGROUND

Postoperative infections constitute more than half of all postoperative complications. Surveillance of these complications is primarily done through manual chart review, which is time consuming, expensive, and typically only covers 10% to 15% of all operations. Automated surveillance would permit the timely evaluation of and reporting of all operations.

METHODS

The goal of this study was to develop and validate parsimonious, interpretable models for conducting surveillance of postoperative infections using structured electronic health records data. This was a retrospective study using 30,639 unique operations from 5 major hospitals between 2013 and 2019. Structured electronic health records data were linked to postoperative outcomes data from the American College of Surgeons National Surgical Quality Improvement Program. Predictors from the electronic health records included diagnoses, procedures, and medications. Infectious complications included surgical site infection, urinary tract infection, sepsis, and pneumonia within 30 days of surgery. The knockoff filter, a penalized regression technique that controls type I error, was applied for variable selection. Models were validated in a chronological held-out dataset.

RESULTS

Seven percent of patients experienced at least one type of postoperative infection. Models selected contained between 4 and 8 variables and achieved >0.91 area under the receiver operating characteristic curve, >81% specificity, >87% sensitivity, >99% negative predictive value, and 10% to 15% positive predictive value in a held-out test dataset.

CONCLUSION

Surveillance and reporting of postoperative infection rates can be implemented for all operations with high accuracy using electronic health records data and simple linear regression models.

Utilizing technology to increase efficiency of infection prevention data collection: Our experience using electronic medical records for symptom surveillance

Utilizing technology can increase efficiency for infection prevention data collection. The purpose of this article is to explain our experience using electronic medical record (EMR) reports for symptom surveillance to increase understanding of how technology enhances infection prevention abilities. The new EMR report method took on average just over 8 minutes a week to run compared to the traditional manual method which took 7 hours a week indicating a major time decrease for data collection.

Risk Factors for Postoperative Urinary Tract Infections Following Anterior Lumbar Interbody Fusion

BACKGROUND

Although risk factors contributing to UTI have been studied in posterior approaches to lumbar fusion, there is a lack of literature on factors contributing to UTI in anterior lumbar interbody fusion (ALIF). Our purpose was to identify preoperative independent risk factors for postoperative urinary tract infection (UTI) following anterior lumbar interbody fusion (ALIF) so that surgeons may be able to initiate preventative measures and minimize the risk of UTI-related morbidity following ALIF.

METHODS

The American College of Surgeons-National Surgical Quality Improvement Program database was queried to identify 10 232 patients who had undergone ALIF from 2005 to 2016; 144 patients (1.41%) developed a postoperative UTI while 10 088 patients (98.59%) did not. Univariate analyses were conducted to compare the 2 cohorts' demographics and preoperative comorbidities. Multivariate logistic regression models were then utilized to identify significant predictors of postoperative UTI following ALIF while controlling for differences seen in univariate analyses.

RESULTS

Age ≥ 60 years (P = .022), female sex (P < .001), alcohol use (P = .014), open wound or wound infections (P = .019), and steroid use (P = .046) were independent risk factors for postoperative UTI. Longer operative times were also independent predictors for developing UTI: 120 minutes ≤ x < 180 minutes (P = .050), 180 minutes ≤ x < 240 minutes (P = .025), and ≥ 240 minutes (P = .001). Postoperative UTI independently increased the risk for pneumonia, blood transfusions, sepsis, thromboembolic events, and extended length of stay as well.

CONCLUSIONS

Age ≥ 60 years, female sex, alcohol use, steroid use, and open wound or wound infections independently increased the risk for UTI following ALIF. Future work analyzing the efficacy of tapering alcohol and steroid use preoperatively and reducing procedural time with the aim of lowering UTI risk is warranted. Preoperative wound care is strongly encouraged to decrease UTI risk.

LEVEL OF EVIDENCE

III.

Identification of postoperative complications using electronic health record data and machine learning

BACKGROUND

Using the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) complication status of patients who underwent an operation at the University of Colorado Hospital, we developed a machine learning algorithm for identifying patients with one or more complications using data from the electronic health record (EHR).

METHODS

We used an elastic-net model to estimate regression coefficients and carry out variable selection. International classification of disease codes (ICD-9), common procedural terminology (CPT) codes, medications, and CPT-specific complication event rate were included as predictors.

RESULTS

Of 6840 patients, 922 (13.5%) had at least one of the 18 complications tracked by NSQIP. The model achieved 88% specificity, 83% sensitivity, 97% negative predictive value, 52% positive predictive value, and an area under the curve of 0.93.

CONCLUSIONS

Using machine learning on EHR postoperative data linked to NSQIP outcomes data, a model with 163 predictors from the EHR identified complications well at our institution.

Association of the Meaningful Use Electronic Health Record Incentive Program With Health Information Technology Venture Capital Funding

Importance

Although the Health Information Technology for Economic and Clinical Health (HITECH) Act has accelerated electronic health record (EHR) adoption since its passage, clinician satisfaction with EHRs remains low, and the association of HITECH with health care information technology (IT) entrepreneurship has remained largely unstudied.

Objective

To determine whether the passage of the HITECH Act was associated with an increase in key measures of health care IT entrepreneurship.

Design, Setting, and Participants

This economic evaluation of venture capital (VC) activity in the US from 2000 to 2019 examined funding trends in health care IT, EHR-related companies, and all VC investments before and after the passage of HITECH. A difference-in-differences analysis compared investments in health care IT companies with those of companies in 3 categories: general health care (non-IT), IT (non-health care), and all US VC transactions. Data were analyzed from September 2018 to August 2019.

Exposures

Venture capital funding received by US companies before and after the HITECH Act.

Main Outcomes and Measures

Venture capital investment in health care IT companies and the proportion of those investments going to seed-stage companies, a proxy for very early-stage entrepreneurship and innovation.

Results

The data included 70 982 investments, of which 9425 (13.3%) were seed stage, 10 706 (15.1%) were early stage, and 50 851 (71.6%) were growth stage. After passage of the HITECH Act, investment in both health care IT companies and EHR-related companies increased at a rate much faster (13.0% and 11.4%, respectively) than VC as a whole (6.9%). In addition, the proportion of investments going to seed-stage health care IT companies increased compared with both overall VC investments and non-IT health care investments. Health care IT companies saw increased probabilities of transactions being seed-stage of 5.1% (SE, 2.2%; 95% CI, 0.8% to 9.3%; P = .02) compared with the entire sample of VC transactions and 13.6% (SE, 1.9%; 95% CI, 9.9% to 17.2%; P < .001) compared with non-IT health care VC transactions. Health care IT had essentially 0 increased probability of a transaction being seed stage compared with IT companies outside health care (-0.8% probability; SE, 2.4%; 95% CI, -5.4% to 3.9%; P = .75).

Conclusions and Relevance

Although widespread clinician dissatisfaction with EHR systems remains a challenge, the HITECH Act's incentive program may have catalyzed early-stage entrepreneurship in health care IT, suggesting an important role for incentives in promoting innovation.

Identification of urinary tract infections using electronic health record data

BACKGROUND

Population ascertainment of postoperative urinary tract infections (UTIs) is time-consuming and expensive, as it often requires manual chart review. Using the American College of Surgeons National Surgical Quality Improvement Program UTI status of patients who underwent an operation at the University of Colorado Hospital, we sought to develop an algorithm for identifying UTIs using data from the electronic health record.

METHODS

Data were split into training (operations occurring between 2013-2015) and test (operations in 2016) sets. A binomial generalized linear model with an elastic-net penalty was used to fit the model and carry out variables selection. International classification of disease codes, common procedural terminology codes, antibiotics, catheterization, and common procedural terminology-specific UTI event rates were included as predictors. The Youden's J statistic was used to determine the optimal classification threshold.

RESULTS

Of 6,840 patients, 134 (2.0%) had a UTI. The model achieved 92% specificity, 80% sensitivity, 100% negative predictive value, 16% positive predictive value, and an area under the curve of 0.94 using a decision threshold of 0.03.

CONCLUSIONS

A model with 14 predictors from the electronic health record identifies UTIs well, and it could be used to scale up UTI surveillance or to estimate the impact of large-scale interventions on UTI rates.

Electronic nursing documentation interventions to promote or improve patient safety and quality care: A systematic review

AIM

To review the evidence on the effects/impact of electronic nursing documentation interventions on promoting or improving quality care and/or patient safety in acute hospital settings.

BACKGROUND

Electronic documentation has been recommended to improve quality care and patient safety. With the gradual move from paper-based to electronic nursing documentation internationally, there is a need to identify interventions that can effectively improve quality care and patient safety.

EVALUATION

We conducted a systematic review on the effectiveness of electronic nursing documentation interventions on promoting or improving quality care and/or patient safety in acute hospital settings.

KEY ISSUES

Six articles reporting on six individual studies met all eligibility criteria. They were uncontrolled pre/post intervention studies reporting positive impacts on at least one or more outcomes. Most outcomes related to documentation practice and documentation of content.

CONCLUSION

Some evidence from our review indicates that implementing electronic nursing documentation in acute hospital settings is time saving, reduces rates of documentation errors, falls and infections.

IMPLICATIONS FOR NURSING MANAGEMENT

A planned approach from management over time to allow nurses adapt to new electronic systems of documentation would seem a good investment in terms of efficiency of work time, possibly resulting in more time for clinical care.

The Impact of Automated Electronic Surveillance of Electronic Medical Records on Pediatric Inpatient Care

Purpose

To assess the impact of the automated surveillance of the electronic medical record process on clinical interventions among hospitalized children at a tertiary care pediatric center.

Methods

A retrospective chart review of the alerts triggered for central line-associated blood stream infections (CLABSIs), catheter-associated urinary tract infections (CAUTIs), neonatal sepsis, or clinical deterioration through elevated pediatric early warning scores (PEWS) by automated electronic surveillance of the hospital electronic medical record (EMR) over a five-month period among hospitalized children. Interventions instituted in response to the alerts were reviewed from the hospital EMR. Fisher’s exact test was performed to detect any significant difference in the proportion of interventions performed for alerts triggered between groups.

Results

A total of 244 alerts were collected (27 CAUTI, 55 CLABSI, 10 neonatal sepsis, and 152 PEWS alerts). A significant difference in the proportion of interventions instituted after neonatal sepsis and PEWS alerts (9/162, 5.6%) as compared to CLABSI and CAUTI alerts (20/82, 24.4%) was observed (p<0.001; Odds ratio (95% CI): 0.182 (0.079-0.422)). Neonatal sepsis triggered the least number of alerts (10/244, 4.1%) and proportionately fewer interventions than the other clinical alerts.

Conclusions

Alerts for potential device-associated infections resulted in more clinical intervention than less-specific alerts. Neonatal sepsis alerts resulted in minimal interventions undertaken in response to the alert. Identifying and focusing on alerts benefitting the patient can serve as a better allocation of time and resources. Future studies should explore which newer alerts and their accompanying interventions improve patient outcomes.

Electronic Surveillance For Catheter-Associated Urinary Tract Infection Using Natural Language Processing

Catheter-associated urinary tract infection (CAUTI) is a common and costly healthcare-associated infection, yet measuring it accurately is challenging and resource-intensive. Electronic surveillance promises to make this task more objective and efficient in an era of new financial and regulatory imperatives, but previous surveillance approaches have used a simplified version of the definition. We applied a complete definition, including subjective elements identified through natural language processing of clinical notes. Through examination of documentation practices, we defined a set of rules that identified positively and negatively asserted symptoms of CAUTI. Our algorithm was developed on a training set of 1421 catheterizedpatients and prospectively validated on 1567 catheterizedpatients. Compared to gold standard chart review, our tool had a sensitivity of 97.1%, specificity of 94.5% PPV of 66.7% and NPV of 99.6% for identifying CAUTI. We discuss sources of error and suggestions for more computable future definitions.

Electronic Health Records and Quality of Care: An Observational Study Modeling Impact on Mortality, Readmissions, and Complications

Electronic health records (EHRs) were implemented to improve quality of care and patient outcomes. This study assessed the relationship between EHR-adoption and patient outcomes.We performed an observational study using State Inpatient Databases linked to American Hospital Association survey, 2011. Surgical and medical patients from 6 large, diverse states were included. We performed univariate analyses and developed hierarchical regression models relating level of EHR utilization and mortality, readmission rates, and complications. We evaluated the effect of EHR adoption on outcomes in a difference-in-differences analysis, 2008 to 2011.Medical and surgical patients sought care at hospitals reporting no EHR (3.5%), partial EHR (55.2%), and full EHR systems (41.3%). In univariate analyses, patients at hospitals with full EHR had the lowest rates of inpatient mortality, readmissions, and Patient Safety Indicators followed by patients at hospitals with partial EHR and then patients at hospitals with no EHR (P < 0.05). However, these associations were not robust when accounting for other patient and hospital factors, and adoption of an EHR system was not associated with improved patient outcomes (P > 0.05).These results indicate that patients receiving medical and surgical care at hospitals with no EHR system have similar outcomes compared to patients seeking care at hospitals with a full EHR system, after controlling for important confounders.To date, we have not yet seen the promised benefits of EHR systems on patient outcomes in the inpatient setting. EHRs may play a smaller role than expected in patient outcomes and overall quality of care.

Necessary Infrastructure of Infection Prevention and Healthcare Epidemiology Programs: A Review

The scope of a healthcare institution's infection prevention and control/healthcare epidemiology program (IPC/HE) should be driven by the size and complexity of the patient population served, that population's risk for healthcare-associated infection (HAI), and local, state, and national regulatory and accreditation requirements. Essential activities of all IPC/HE programs include but are not limited to the following: ∙ Surveillance.∙ Performance improvement to reduce HAI ∙ Acute event response, including outbreak investigation ∙ Education and training of both healthcare personnel and patients ∙ Reporting of HAI to the Centers for Disease Control and Prevention's National Healthcare Safety Network as well as entities required by law.

Recent Advances in Clinical Natural Language Processing in Support of Semantic Analysis

OBJECTIVES

We present a review of recent advances in clinical Natural Language Processing (NLP), with a focus on semantic analysis and key subtasks that support such analysis.

METHODS

We conducted a literature review of clinical NLP research from 2008 to 2014, emphasizing recent publications (2012-2014), based on PubMed and ACL proceedings as well as relevant referenced publications from the included papers.

RESULTS

Significant articles published within this time-span were included and are discussed from the perspective of semantic analysis. Three key clinical NLP subtasks that enable such analysis were identified: 1) developing more efficient methods for corpus creation (annotation and de-identification), 2) generating building blocks for extracting meaning (morphological, syntactic, and semantic subtasks), and 3) leveraging NLP for clinical utility (NLP applications and infrastructure for clinical use cases). Finally, we provide a reflection upon most recent developments and potential areas of future NLP development and applications.

CONCLUSIONS

There has been an increase of advances within key NLP subtasks that support semantic analysis. Performance of NLP semantic analysis is, in many cases, close to that of agreement between humans. The creation and release of corpora annotated with complex semantic information models has greatly supported the development of new tools and approaches. Research on non-English languages is continuously growing. NLP methods have sometimes been successfully employed in real-world clinical tasks. However, there is still a gap between the development of advanced resources and their utilization in clinical settings. A plethora of new clinical use cases are emerging due to established health care initiatives and additional patient-generated sources through the extensive use of social media and other devices.

Incidence rates of hospital-acquired urinary tract and bloodstream infections generated by automated compilation of electronically available healthcare data

BACKGROUND

Monitoring of hospital-acquired infection (HAI) by automated compilation of registry data may address the disadvantages of laborious, costly and potentially subjective and often random sampling of data by manual surveillance.

AIM

To evaluate a system for automated monitoring of hospital-acquired urinary tract (HA-UTI) and bloodstream infections (HA-BSI) and to report incidence rates over a five-year period in a Danish hospital trust.

METHODS

Based primarily on electronically available data relating to microbiology results and antibiotic prescriptions, the automated monitoring of HA-UTIs and HA-BSIs was validated against data from six previous point-prevalence surveys (PPS) from 2010 to 2013 and data from a manual assessment (HA-UTI only) of one department of internal medicine from January 2010. Incidence rates (infections per 1000 bed-days) from 2010 to 2014 were calculated.

FINDINGS

Compared with the PPSs, the automated monitoring showed a sensitivity of 88% in detecting UTI in general, 78% in detecting HA-UTI, and 100% in detecting BSI in general. The monthly incidence rates varied between 4.14 and 6.61 per 1000 bed-days for HA-UTI and between 0.09 and 1.25 per 1000 bed-days for HA-BSI.

CONCLUSION

Replacing PPSs with automated monitoring of HAIs may provide better and more objective data and constitute a promising foundation for individual patient risk analyses and epidemiological studies. Automated monitoring may be universally applicable in hospitals with electronic databases comprising microbiological findings, admission data, and antibiotic prescriptions.

Data elements and validation methods used for electronic surveillance of health care-associated infections: a systematic review

BACKGROUND

We describe the primary data sources, data elements, and validation methods currently used in electronic surveillance systems (ESS) for identification and surveillance of health care-associated infections (HAIs), and compares these data elements and validation methods with recommended standards.

METHODS

Using Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, a PubMed and manual search was conducted to identify research articles describing ESS for identification and surveillance of HAIs published January 1, 2009-August 31, 2014. Selected articles were evaluated to determine what data elements and validation methods were included.

RESULTS

Among the 509 articles identified in the original literature search, 30 met the inclusion criteria. Whereas the majority of studies (83%) used recommended data sources and validated the numerator (80%), only 10% of studies performed external and internal validation. In addition, there was variation in the ESS data formats used.

CONCLUSIONS

Our findings suggest that the majority of ESS for HAI surveillance use standard definitions, but the lack of widespread internal data, denominator, and external validation in these systems reduces the reliability of their findings. Additionally, advanced programming skills are required to create, implement, and maintain these systems and to reduce the variability in data formats.

Natural Language Processing for Real-Time Catheter-Associated Urinary Tract Infection Surveillance: Results of a Pilot Implementation Trial

BACKGROUND

Incidence of catheter-associated urinary tract infection (CAUTI) is a quality benchmark. To streamline conventional detection methods, an electronic surveillance system augmented with natural language processing (NLP), which gathers data recorded in clinical notes without manual review, was implemented for real-time surveillance.

OBJECTIVE

To assess the utility of this algorithm for identifying indwelling urinary catheter days and CAUTI.

SETTING

Large, urban tertiary care Veterans Affairs hospital.

METHODS

All patients admitted to the acute care units and the intensive care unit from March 1, 2013, through November 30, 2013, were included. Standard surveillance, which includes electronic and manual data extraction, was compared with the NLP-augmented algorithm.

RESULTS

The NLP-augmented algorithm identified 27% more indwelling urinary catheter days in the acute care units and 28% fewer indwelling urinary catheter days in the intensive care unit. The algorithm flagged 24 CAUTI versus 20 CAUTI by standard surveillance methods; the CAUTI identified were overlapping but not the same. The overall positive predictive value was 54.2%, and overall sensitivity was 65% (90.9% in the acute care units but 33% in the intensive care unit). Dissimilarities in the operating characteristics of the algorithm between types of unit were due to differences in documentation practice. Development and implementation of the algorithm required substantial upfront effort of clinicians and programmers to determine current language patterns.

CONCLUSIONS

The NLP algorithm was most useful for identifying simple clinical variables. Algorithm operating characteristics were specific to local documentation practices. The algorithm did not perform as well as standard surveillance methods.