Does nurse use of a standardized flowsheet to document communication with advanced providers provide a mechanism to detect pulse oximetry failures? A retrospective study of electronic health record data

BACKGROUND

Pulse oximetry guides clinical decisions, yet does not uniformly identify hypoxemia. We hypothesized that nursing documentation of notifying providers, facilitated by a standardized flowsheet for documenting communication to providers (physicians, nurse practitioners, and physician assistants), may increase when hypoxemia is present, but undetected by the pulse oximeter, in events termed "occult hypoxemia."

OBJECTIVE

To compare nurse documentation of provider notification in the 4 h preceding cases of occult hypoxemia, normal oxygenation, and evident hypoxemia confirmed by an arterial blood gas reading.

METHODS

We conducted a retrospective study using electronic health record data from patients with COVID-19 at five hospitals in a healthcare system with paired SpO2 and SaO2 readings (measurements within 10 min of oxygen saturation levels in arterial blood, SaO2, and by pulse oximetry, SpO2). We applied multivariate logistic regression to assess if having any nursing documentation of provider notification in the 4 h prior to a paired reading confirming occult hypoxemia was more likely compared to a paired reading confirming normal oxygen status, adjusting for characteristics significantly associated with nursing documentation. We applied conditional logistic regression to assess if having any nursing documentation of provider notification was more likely in the 4-hour window preceding a paired reading compared to the 4-hour window 24 h earlier separately for occult hypoxemia, visible hypoxemia, and normal oxygenation.

RESULTS

There were data from 1910 patients hospitalized with COVID-19 who had 44,972 paired readings and an average of 26.5 (34.5) nursing documentation of provider notification events. The mean age was 63.4 (16.2). Almost half (866/1910, 45.3 %) were White, 701 (36.7 %) were Black, and 239 (12.5 %) were Hispanic. Having any nursing documentation of provider notification was 46 % more common in the 4 h before an occult hypoxemia paired reading compared to a normal oxygen status paired reading (OR 1.46, 95 % CI: 1.28-1.67). Comparing the 4 h immediately before the reading to the 4 h one day preceding the paired reading, there was a higher likelihood of having any nursing documentation of provider notification for both evident (OR 1.45, 95 % CI 1.24-1.68) and occult paired readings (OR 1.26, 95 % CI 1.04-1.53).

CONCLUSION

This study finds that nursing documentation of provider notification significantly increases prior to confirmed occult hypoxemia, which has potential in proactively identifying occult hypoxemia and other clinical issues. There is potential value to encouraging standardized documentation of nurse concern, including communication to providers, to facilitate its inclusion in clinical decision-making.

The accuracy of fully-automated algorithms for the surveillance of central venous catheter-related bloodstream infection in hospitalised patients

BACKGROUND

Continuous surveillance for healthcare-associated infections such as central venous catheter-related bloodstream infections (CVC-BSI) is crucial for prevention. However, traditional surveillance methods are resource-intensive and prone to bias. This study aimed to develop and validate fully-automated surveillance algorithms for CVC-BSI.

METHODS

Two algorithms were developed using electronic health record data from 1000 admissions with a positive blood culture (BCx) at Karolinska University Hospital from 2017: (1) Combining microbiological findings in BCx and CVC cultures with BSI symptoms; (2) Only using microbiological findings. These algorithms were validated in 5170 potential CVC-BSI-episodes from all admissions in 2018-2019, and results extrapolated to all potential CVC-BSI-episodes within this period (n = 181,354). The reference standard was manual record review according to ECDC's definition of microbiologically confirmed CVC-BSI (CRI3-CVC).

RESULTS

In the potential CVC-BSI-episodes, 51 fulfilled ECDC's definition and the algorithms identified 47 and 49 episodes as CVC-BSI, respectively. Both algorithms performed well in assessing CVC-BSI. Overall, algorithm 2 performed slightly better with in the total period a sensitivity of 0.880 (95%-CI 0.783-0.959), specificity of 1.000 (95%-CI 0.999-1.000), PPV of 0.918 (95%-CI 0.833-0.981) and NPV of 1.000 (95%-CI 0.999-1.000). Incidence according to the reference and algorithm 2 was 0.33 and 0.31 per 1000 in-patient hospital-days, respectively.

CONCLUSIONS

Both fully-automated surveillance algorithms for CVC-BSI performed well and could effectively replace manual surveillance. The simpler algorithm, using only microbiology data, is suitable when BCx testing adheres to recommendations, otherwise the algorithm using symptom data might be required. Further validation in other settings is necessary to assess the algorithms' generalisability.

Electronic Health Record Recording of Patient Pain: Challenges and Discrepancies

PURPOSE OF REVIEW

In the present review, various categories of pain, clinician-observed pain scales, and patient-reported pain scales are evaluated to better understand factors that impact patient pain perceptions. Additionally, the expansion of areas that require further research to determine the optimal way to evaluate pain scale data for treatment and management are discussed.

RECENT FINDINGS

Electronic health record (EHR) data provides a starting point for evaluating whether patient predictors influence postoperative pain. There are several ways to assess pain and choosing the most effective form of pain treatment. Identifying individuals at high risk for severe postoperative pain enables more effective pain treatment. However, there are discrepancies in patient pain reporting dependent on instruments used to measure pain and their storage in the EHR. Additionally, whether administered by a physician or another healthcare practitioner, differences in patient pain perception occur. While each scale has distinct advantages and limitations, pain scale data is a valuable therapeutic tool for assisting clinicians in providing patients with optimal pain control. Accurate assessment of patient pain perceptions by data extraction from electronic health records provides a potential for pain alleviation improvement. Predicting high-risk postoperative pain syndromes is a difficult clinical challenge. Numerous studies have been conducted on factors that impact pain prediction. Postoperative pain is significantly predicted by the kind of operation, the existence of prior discomfort, patient anxiety, and age.

Identifying populations with chronic pain in primary care: developing an algorithm and logic rules applied to coded primary care diagnostic and medication data

BACKGROUND

Estimates of chronic pain prevalence using coded primary care data are likely to be substantially lower than estimates derived from community surveys. Most primary care studies have estimated chronic pain prevalence using data searches confined to analgesic medication prescriptions. Increasingly, following recent NICE guideline recommendations, patients and doctors opt for non-drug treatment of chronic pain thus excluding these patients from prevalence estimates based on medication codes. We aimed to develop and test an algorithm combining medication codes with selected diagnostic codes to estimate chronic pain prevalence using coded primary care data.

METHODS

Following a scoping review 4 criteria were developed to identify cohorts of people with chronic pain. These were (1) people with one of 12 ('tier 1') conditions that almost always results in the individual having chronic pain (2) people with one of 20 ('tier 2') conditions included when there are also 3 or more prescription-only analgesics issued in the last 12 months (3) chronic neuropathic pain, or (4) 4 or more prescription-only analgesics issued in the last 12 months. These were translated into 8 logic rules which included 1,932 SNOMED CT codes.

RESULTS

The algorithm was run on primary care data from 41 GP Practices in Lambeth. The total population consisted of 386,238 GP registered adults ≥ 18 years as of the 31st March 2021. 64,135 (16.6%) were identified as people with chronic pain. This definition demonstrated notably high rates in Black ethnicity females, and higher rates in the most deprived, and older population.

CONCLUSIONS

Estimates of chronic pain prevalence using structured healthcare data have previously shown lower prevalence estimates for chronic pain than reported in community surveys. This has limited the ability of researchers and clinicians to fully understand and address the complex multifactorial nature of chronic pain. Our study demonstrates that it may be possible to establish more representative prevalence estimates using structured data than previously possible. Use of logic rules offers the potential to move systematic identification and population-based management of chronic pain into mainstream clinical practice at scale and support improved management of symptom burden for people experiencing chronic pain.

Interpretable physiological forecasting in the ICU using constrained data assimilation and electronic health record data

OBJECTIVE

Prediction of physiological mechanics are important in medical practice because interventions are guided by predicted impacts of interventions. But prediction is difficult in medicine because medicine is complex and difficult to understand from data alone, and the data are sparse relative to the complexity of the generating processes. Computational methods can increase prediction accuracy, but prediction with clinical data is difficult because the data are sparse, noisy and nonstationary. This paper focuses on predicting physiological processes given sparse, non-stationary, electronic health record data in the intensive care unit using data assimilation (DA), a broad collection of methods that pair mechanistic models with inference methods.

METHODS

A methodological pipeline embedding a glucose-insulin model into a new DA framework, the constrained ensemble Kalman filter (CEnKF) to forecast blood glucose was developed. The data include tube-fed patients whose nutrition, blood glucose, administered insulins and medications were extracted by hand due to their complexity and to ensure accuracy. The model was estimated using an individual's data as if they arrived in real-time, and the estimated model was run forward producing a forecast. Both constrained and unconstrained ensemble Kalman filters were estimated to compare the impact of constraints. Constraint boundaries, model parameter sets estimated, and data used to estimate the models were varied to investigate their influence on forecasting accuracy. Forecasting accuracy was evaluated according to mean squared error between the model-forecasted glucose and the measurements and by comparing distributions of measured glucose and forecast ensemble means.

RESULTS

The novel CEnKF produced substantial gains in robustness and accuracy while minimizing the data requirements compared to the unconstrained ensemble Kalman filters. Administered insulin and tube-nutrition were important for accurate forecasting, but including glucose in IV medication delivery did not increase forecast accuracy. Model flexibility, controlled by constraint boundaries and estimated parameters, did influence forecasting accuracy.

CONCLUSION

Accurate and robust physiological forecasting with sparse clinical data is possible with DA. Introducing constrained inference, particularly on unmeasured states and parameters, reduced forecast error and data requirements. The results are not particularly sensitive to model flexibility such as constraint boundaries, but over or under constraining increased forecasting errors.

Blood pressure and cholesterol measurements in primary care: cross-sectional analyses in a dynamic cohort

Background

Guidelines on cardiovascular risk management (CVRM) recommend blood pressure (BP) and cholesterol measurements every 5 years in men aged ≥40 years and (post-menopausal) women aged ≥50 years.

Aim

To evaluate CVRM guideline implementation.

Design & setting

Cross-sectional analyses in a dynamic cohort using primary care electronic health record (EHR) data from the Julius General Practitioners’ Network (JGPN) (n = 388 929).

Method

Trends (2008–2018) were assessed in the proportion of patients with at least one measurement (BP and cholesterol) every 1, 2, and 5 years, in those with:

1. a history of cardiovascular disease (CVD) and diabetes mellitus (DM);

2. a history of DM only;

3. a history of CVD only;

4. a cardiovascular risk assessment (CRA) indication based on other medical history, or;

5. no CRA indication.

Trends were evaluated over time using logistic regression mixed-model analyses.

Results

Trends in annual BP and cholesterol measurement increased for patients with a history of CVD from 37.0% to 48.4% (P<0.001) and 25.8% to 40.2% (P<0.001). In the 5-year window from 2014–2018, BP and cholesterol measurements were performed respectively in 78.5% and 74.1% of all men aged ≥40 years and 82.2% and 78.5% of all women aged ≥50 years. Least measured were patients without a CRA indication (men 60.2% and 62.4%; women 55.5% and 59.3%).

Conclusion

The fairly high frequency of CVRM measurements available in the EHR of patients in primary care suggests an adequate implementation of the CVRM guideline. As nearly all individuals visit the GP at least once within a 5-year time window, improvement of CVRM remains possible, especially in those without a CRA indication.

Unified concept and assertion detection using contextual multi-task learning in a clinical decision support system

Assertions, such as negation and speculation, alter the meaning of clinical findings ('concepts') in Electronic Health Records. Accurate assertion detection is vital to the identification of target findings in clinical decision support systems. Diverse clinical concepts and assertion modifiers embedded within longer sentences add to the challenge of error-free detection. Recent approaches leveraging biomedical contextual embeddings lead to standalone concept and assertion models that do not effectively utilize inter-task knowledge transfer. We propose a novel neural model integrating task-specific fine-tuning and multi-task learning in a coherent framework based on the hierarchical relationship between the tasks. We show that such a unified framework enhances both the tasks using several real-world clinical notes' datasets (n2c2 2010, n2c2 2012, NegEx). Concept task performance enhanced by +1.69 F1 on n2c2 2010 and +2.96 F1 on n2c2 2012 compared to standalone baselines. Assertion recognition improved by +2.89 F1 and +3.77 F1, respectively. Negation detection under low-resource settings increased significantly (+2.4 F1, p-value = 3.11E-05, McNemar's test), demonstrating the impact of inter-task knowledge transfer. The integrated architecture enhanced the generalization performance of speculation detection (+2.09 F1). To the best of our knowledge, this model is the first demonstration of a contextual multi-task system for unified detection of concepts and assertions in clinical decision support applications.

Accuracy of identifying hospital acquired venous thromboembolism by administrative coding: implications for big data and machine learning research

Big data analytics research using heterogeneous electronic health record (EHR) data requires accurate identification of disease phenotype cases and controls. Overreliance on ground truth determination based on administrative data can lead to biased and inaccurate findings. Hospital-acquired venous thromboembolism (HA-VTE) is challenging to identify due to its temporal evolution and variable EHR documentation. To establish ground truth for machine learning modeling, we compared accuracy of HA-VTE diagnoses made by administrative coding to manual review of gold standard diagnostic test results. We performed retrospective analysis of EHR data on 3680 adult stepdown unit patients identifying HA-VTE. International Classification of Diseases, Ninth Revision (ICD-9-CM) codes for VTE were identified. 4544 radiology reports associated with VTE diagnostic tests were screened using terminology extraction and then manually reviewed by a clinical expert to confirm diagnosis. Of 415 cases with ICD-9-CM codes for VTE, 219 were identified with acute onset type codes. Test report review identified 158 new-onset HA-VTE cases. Only 40% of ICD-9-CM coded cases (n = 87) were confirmed by a positive diagnostic test report, leaving the majority of administratively coded cases unsubstantiated by confirmatory diagnostic test. Additionally, 45% of diagnostic test confirmed HA-VTE cases lacked corresponding ICD codes. ICD-9-CM coding missed diagnostic test-confirmed HA-VTE cases and inaccurately assigned cases without confirmed VTE, suggesting dependence on administrative coding leads to inaccurate HA-VTE phenotyping. Alternative methods to develop more sensitive and specific VTE phenotype solutions portable across EHR vendor data are needed to support case-finding in big-data analytics.

Protocol for a national blood transfusion data warehouse from donor to recipient

INTRODUCTION

Blood transfusion has health-related, economical and safety implications. In order to optimise the transfusion chain, comprehensive research data are needed. The Dutch Transfusion Data warehouse (DTD) project aims to establish a data warehouse where data from donors and transfusion recipients are linked. This paper describes the design of the data warehouse, challenges and illustrative applications.

STUDY DESIGN AND METHODS

Quantitative data on blood donors (eg, age, blood group, antibodies) and products (type of product, processing, storage time) are obtained from the national blood bank. These are linked to data on the transfusion recipients (eg, transfusions administered, patient diagnosis, surgical procedures, laboratory parameters), which are extracted from hospital electronic health records.

APPLICATIONS

Expected scientific contributions are illustrated for 4 applications: determine risk factors, predict blood use, benchmark blood use and optimise process efficiency. For each application, examples of research questions are given and analyses planned.

CONCLUSIONS

The DTD project aims to build a national, continuously updated transfusion data warehouse. These data have a wide range of applications, on the donor/production side, recipient studies on blood usage and benchmarking and donor-recipient studies, which ultimately can contribute to the efficiency and safety of blood transfusion.