Development and validation of an arterial blood gas analysis interpretation algorithm for application in clinical laboratory services

Identification of data elements for blood gas analysis dataset: a base for developing registries and artificial intelligence-based systems

Background

One of the challenging decision-making tasks in healthcare centers is the interpretation of blood gas tests. One of the most effective assisting approaches for the interpretation of blood gas analysis (BGA) can be artificial intelligence (AI)-based decision support systems. A primary step to develop intelligent systems is to determine information requirements and automated data input for the secondary analyses. Datasets can help the automated data input from dispersed information systems. Therefore, the current study aimed to identify the data elements required for supporting BGA as a dataset.

Materials and methods

This cross-sectional descriptive study was conducted in Nemazee Hospital, Shiraz, Iran. A combination of literature review, experts’ consensus, and the Delphi technique was used to develop the dataset. A review of the literature was performed on electronic databases to find the dataset for BGA. An expert panel was formed to discuss on, add, or remove the data elements extracted through searching the literature. Delphi technique was used to reach consensus and validate the draft dataset.

Results

The data elements of the BGA dataset were categorized into ten categories, namely personal information, admission details, present illnesses, past medical history, social status, physical examination, paraclinical investigation, blood gas parameter, sequential organ failure assessment (SOFA) score, and sampling technique errors. Overall, 313 data elements, including 172 mandatory and 141 optional data elements were confirmed by the experts for being included in the dataset.

Conclusions

We proposed a dataset as a base for registries and AI-based systems to assist BGA. It helps the storage of accurate and comprehensive data, as well as integrating them with other information systems. As a result, high-quality care is provided and clinical decision-making is improved.

Automatic real-time analysis and interpretation of arterial blood gas sample for Point-of-care testing: Clinical validation

Background

Point-of-care arterial blood gas (ABG) is a blood measurement test and a useful diagnostic tool that assists with treatment and therefore improves clinical outcomes. However, numerically reported test results make rapid interpretation difficult or open to interpretation. The arterial blood gas algorithm (ABG-a) is a new digital diagnostics solution that can provide clinicians with real-time interpretation of preliminary data on safety features, oxygenation, acid-base disturbances and renal profile. The main aim of this study was to clinically validate the algorithm against senior experienced clinicians, for acid-base interpretation, in a clinical context.

Methods

We conducted a prospective international multicentre observational cross-sectional study. 346 sample sets and 64 inpatients eligible for ABG met strict sampling criteria. Agreement was evaluated using Cohen’s kappa index, diagnostic accuracy was evaluated with sensitivity, specificity, efficiency or global accuracy and positive predictive values (PPV) and negative predictive values (NPV) for the prevalence in the study population.

Results

The concordance rates between the interpretations of the clinicians and the ABG-a for acid-base disorders were an observed global agreement of 84,3% with a Cohen’s kappa coefficient 0.81; 95% CI 0.77 to 0.86; p < 0.001. For detecting accuracy normal acid-base status the algorithm has a sensitivity of 90.0% (95% CI 79.9 to 95.3), a specificity 97.2% (95% CI 94.5 to 98.6) and a global accuracy of 95.9% (95% CI 93.3 to 97.6). For the four simple acid-base disorders, respiratory alkalosis: sensitivity of 91.2 (77.0 to 97.0), a specificity 100.0 (98.8 to 100.0) and global accuracy of 99.1 (97.5 to 99.7); respiratory acidosis: sensitivity of 61.1 (38.6 to 79.7), a specificity of 100.0 (98.8 to 100.0) and global accuracy of 98.0 (95.9 to 99.0); metabolic acidosis: sensitivity of 75.8 (59.0 to 87.2), a specificity of 99.7 (98.2 to 99.9) and a global accuracy of 97.4 (95.1 to 98.6); metabolic alkalosis sensitivity of 72.2 (56.0 to 84.2), a specificity of 95.5 (92.5 to 97.3) and a global accuracy of 93.0 (88.8 to 95.3); the four complex acid-base disorders, respiratory and metabolic alkalosis, respiratory and metabolic acidosis, respiratory alkalosis and metabolic acidosis, respiratory acidosis and metabolic alkalosis, the sensitivity, specificity and global accuracy was also high. For normal acid-base status the algorithm has PPV 87.1 (95% CI 76.6 to 93.3) %, and NPV 97.9 (95% CI 95.4 to 99.0) for a prevalence of 17.4 (95% CI 13.8 to 21.8). For the four-simple acid-base disorders and the four complex acid-base disorders the PPV and NPV were also statistically significant.

Conclusions

The ABG-a showed very high agreement and diagnostic accuracy with experienced senior clinicians in the acid-base disorders in a clinical context. The method also provides refinement and deep complex analysis at the point-of-care that a clinician could have at the bedside on a day-to-day basis. The ABG-a method could also have the potential to reduce human errors by checking for imminent life-threatening situations, analysing the internal consistency of the results, the oxygenation and renal status of the patient.

Laboratory test interpretations and algorithms in utilization management

Appropriate assimilation of laboratory test results into patient care is enhanced when pathologist interpretations of the laboratory tests are provided for clinicians, and when reflex algorithm testing is utilized. Benefits of algorithms and interpretations include avoidance of misdiagnoses, reducing the number of laboratory tests needed, reducing the number of procedures, transfusions and admissions, shortening the amount of time needed to reach a diagnosis, reducing errors in test ordering, and providing additional information about how the laboratory results might affect other aspects of a patient's care. Providing interpretations can be challenging for pathologists, therefore mechanisms to facilitate the successful implementation of an interpretation service are described. These include algorithm-based testing and interpretation, optimizing laboratory requisitions and/or order-entry systems, proficiency testing programs that assess interpretations and provide constructive feedback, utilization of a collection of interpretive sentences or paragraphs that can be building blocks ("coded comments") for constructing preliminary interpretations, middleware, and pathology resident participation and education. In conclusion, the combination of algorithms and interpretations for laboratory testing has multiple benefits for the medical care for the patient.

Discretization of continuous features in clinical datasets

BACKGROUND

The increasing availability of clinical data from electronic medical records (EMRs) has created opportunities for secondary uses of health information. When used in machine learning classification, many data features must first be transformed by discretization.

OBJECTIVE

To evaluate six discretization strategies, both supervised and unsupervised, using EMR data.

MATERIALS AND METHODS

We classified laboratory data (arterial blood gas (ABG) measurements) and physiologic data (cardiac output (CO) measurements) derived from adult patients in the intensive care unit using decision trees and naïve Bayes classifiers. Continuous features were partitioned using two supervised, and four unsupervised discretization strategies. The resulting classification accuracy was compared with that obtained with the original, continuous data.

RESULTS

Supervised methods were more accurate and consistent than unsupervised, but tended to produce larger decision trees. Among the unsupervised methods, equal frequency and k-means performed well overall, while equal width was significantly less accurate.

DISCUSSION

This is, we believe, the first dedicated evaluation of discretization strategies using EMR data. It is unlikely that any one discretization method applies universally to EMR data. Performance was influenced by the choice of class labels and, in the case of unsupervised methods, the number of intervals. In selecting the number of intervals there is generally a trade-off between greater accuracy and greater consistency.

CONCLUSIONS

In general, supervised methods yield higher accuracy, but are constrained to a single specific application. Unsupervised methods do not require class labels and can produce discretized data that can be used for multiple purposes.