A systematic review of prediction models to diagnose COVID-19 in adults admitted to healthcare centers

Development of Machine Learning Models for the Identification of Elevated Ketone Bodies During Hyperglycemia in Patients with Type 1 Diabetes

Aims: Diabetic ketoacidosis (DKA) is a serious life-threatening condition caused by a lack of insulin, which leads to elevated plasma glucose and metabolic acidosis. Early identification of developing DKA is important to start treatment and minimize complications and risk of death. The aim of the present study is to develop and test prediction model(s) that gives an alarm about their risk of developing elevated ketone bodies during hyperglycemia. Methods: We analyzed data from 138 type 1 diabetes patients with measurements of ketone bodies and continuous glucose monitoring (CGM) data from over 30,000 days of wear time. We utilized a supervised binary classification machine learning approach to identify elevated levels of ketone bodies (≥0.6 mmol/L). Data material was randomly divided at patient level in 70%/30% (training/test) dataset. Logistic regression (LR) and random forest (RF) classifier were compared. Results: Among included patients, 913 ketone samples were eligible for modeling, including 273 event samples with ketone levels ≥0.6 mmol/L. An area under the receiver operating characteristic curve from the RF classifier was 0.836 (confidence interval [CI] 90%, 0.783-0.886) and 0.710 (CI 90%, 0.646-0.77) for the LR classifier. Conclusions: The novel approach for identifying elevated ketone levels in patients with type 1 diabetes utilized in this study indicates that CGM could be a valuable resource for the early prediction of patients at risk of developing DKA. Future studies are needed to validate the results.

Combination of rRT-PCR and Clinical Features to Predict Coronavirus Disease 2019 for Nosocomial Infection Control

Background

Coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), immediately became a pandemic. Therefore, nosocomial infection control is necessary to screen for patients with possible COVID-19.

Objective

This study aimed to investigate commonly measured clinical variables to predict COVID-19.

Methods

This cross-sectional study enrolled 1087 patients in the isolation ward of a university hospital. Conferences were organized to differentiate COVID-19 from non-COVID-19 cases, and multiple nucleic acid tests were mandatory when COVID-19 could not be excluded. Multivariate logistic regression models were employed to determine the clinical factors associated with COVID-19 at the time of hospitalization.

Results

Overall, 352 (32.4%) patients were diagnosed with COVID-19. The majority of the non-COVID-19 cases were predominantly caused by bacterial infections. Multivariate analysis indicated that COVID-19 was significantly associated with age, sex, body mass index, lactate dehydrogenase, C-reactive protein, and malignancy.

Conclusion

Some clinical factors are useful to predict patients with COVID-19 among those with symptoms similar to COVID-19. This study suggests that at least two real-time reverse-transcription polymerase chain reactions of SARS-CoV-2 are recommended to exclude COVID-19.

Systematic reviews of machine learning in healthcare: a literature review

INTRODUCTION

The increasing availability of data and computing power has made machine learning (ML) a viable approach to faster, more efficient healthcare delivery.

METHODS

A systematic literature review (SLR) of published SLRs evaluating ML applications in healthcare settings published between1 January 2010 and 27 March 2023 was conducted.

RESULTS

In total 220 SLRs covering 10,462 ML algorithms were reviewed. The main application of AI in medicine related to the clinical prediction and disease prognosis in oncology and neurology with the use of imaging data. Accuracy, specificity, and sensitivity were provided in 56%, 28%, and 25% SLRs respectively. Internal and external validation was reported in 53% and less than 1% of the cases respectively. The most common modeling approach was neural networks (2,454 ML algorithms), followed by support vector machine and random forest/decision trees (1,578 and 1,522 ML algorithms, respectively).

EXPERT OPINION

The review indicated considerable reporting gaps in terms of the ML's performance, both internal and external validation. Greater accessibility to healthcare data for developers can ensure the faster adoption of ML algorithms into clinical practice.

Evaluation of hematological parameters alterations in different waves of COVID-19 pandemic: A cross-sectional study

BACKGROUND

The occurrence of variations in routine hematological parameters is closely associated with disease progression, the development of severe illness, and the mortality rate among COVID-19 patients. This study aimed to investigate hematological parameters in COVID-19 hospitalized patients from the 1st to the 5th waves of the current pandemic.

METHODS

This cross-sectional study included a total of 1501 hospitalized patients with laboratory-confirmed COVID-19 based on WHO criteria, who were admitted to Shahid Sadoughi Hospital (SSH) in Yazd, Iran, from February 2020 to September 2021. Throughout, we encountered five COVID-19 surge waves. In each wave, we randomly selected approximately 300 patients and categorized them based on infection severity during their hospitalization, including partial recovery, full recovery, and death. Finally, hematological parameters were compared based on age, gender, pandemic waves, and outcomes using the Mann-Whitney U and Kruskal-Wallis tests.

RESULTS

The mean age of patients (n = 1501) was 61.1±21.88, with 816 (54.3%) of them being men. The highest mortality in this study was related to the third wave of COVID-19 with 21.3%. There was a significant difference in all of the hematological parameters, except PDW, PLT, and RDW-CV, among pandemic waves of COVID-19 in our population. The highest rise in the levels of MCV and RDW-CV occurred in the 1st wave, in the 2nd wave for lymphocyte count, MCHC, PLT count, and RDW-SD, in the 3rd wave for WBC, RBC, neutrophil count, MCH, and PDW, and in the 4th wave for Hb, Hct, and ESR (p < 0.01). The median level of Hct, Hb, RBC, and ESR parameters were significantly higher, while the mean level of lymphocyte and were lower in men than in women (p < 0.001). Also, the mean neutrophil in deceased patients significantly was higher than in those with full recovered or partial recovery (p < 0.001).

CONCLUSION

The findings of our study unveiled notable variations in hematological parameters across different pandemic waves, gender, and clinical outcomes. These findings indicate that the behavior of different strains of the COVID-19 may differ across various stages of the pandemic.

Conventional and Novel Diagnostic Tools for the Diagnosis of Emerging SARS-CoV-2 Variants

Accurate identification at an early stage of infection is critical for effective care of any infectious disease. The "coronavirus disease 2019 (COVID-19)" outbreak, caused by the virus "Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)", corresponds to the current and global pandemic, characterized by several developing variants, many of which are classified as variants of concern (VOCs) by the "World Health Organization (WHO, Geneva, Switzerland)". The primary diagnosis of infection is made using either the molecular technique of RT-PCR, which detects parts of the viral genome's RNA, or immunodiagnostic procedures, which identify viral proteins or antibodies generated by the host. As the demand for the RT-PCR test grew fast, several inexperienced producers joined the market with innovative kits, and an increasing number of laboratories joined the diagnostic field, rendering the test results increasingly prone to mistakes. It is difficult to determine how the outcomes of one unnoticed result could influence decisions about patient quarantine and social isolation, particularly when the patients themselves are health care providers. The development of point-of-care testing helps in the rapid in-field diagnosis of the disease, and such testing can also be used as a bedside monitor for mapping the progression of the disease in critical patients. In this review, we have provided the readers with available molecular diagnostic techniques and their pitfalls in detecting emerging VOCs of SARS-CoV-2, and lastly, we have discussed AI-ML- and nanotechnology-based smart diagnostic techniques for SARS-CoV-2 detection.

Development of Clinical Risk Scores for Detection of COVID-19 in Suspected Patients During a Local Outbreak in China: A Retrospective Cohort Study

Objectives: To develop and internally validate two clinical risk scores to detect coronavirus disease 2019 (COVID-19) during local outbreaks.

Methods: Medical records were extracted for a retrospective cohort of 336 suspected patients admitted to Baodi hospital between 27 January to 20 February 2020. Multivariate logistic regression was applied to develop the risk-scoring models, which were internally validated using a 5-fold cross-validation method and Hosmer-Lemeshow (H-L) tests.

Results: Fifty-six cases were diagnosed from the cohort. The first model was developed based on seven significant predictors, including age, close contact with confirmed/suspected cases, same location of exposure, temperature, leukocyte counts, radiological findings of pneumonia and bilateral involvement (the mean area under the receiver operating characteristic curve [AUC]:0.88, 95% CI: 0.84–0.93). The second model had the same predictors except leukocyte and radiological findings (AUC: 0.84, 95% CI: 0.78–0.89, Z = 2.56, p = 0.01). Both were internally validated using H-L tests and showed good calibration (both p > 0.10).

Conclusion: Two clinical risk scores to detect COVID-19 in local outbreaks were developed with excellent predictive performances, using commonly measured clinical variables. Further external validations in new outbreaks are warranted.

Modelling of a triage scoring tool for SARS-COV-2 PCR testing in health-care workers: data from the first German COVID-19 Testing Unit in Munich

Background

Numerous scoring tools have been developed for assessing the probability of SARS-COV-2 test positivity, though few being suitable or adapted for outpatient triage of health care workers.

Methods

We retrospectively analysed 3069 patient records of health care workers admitted to the COVID-19 Testing Unit of the Ludwig-Maximilians-Universität of Munich between January 27 and September 30, 2020, for real-time polymerase chain reaction analysis of naso- or oropharyngeal swabs. Variables for a multivariable logistic regression model were collected from self-completed case report forms and selected through stepwise backward selection. Internal validation was conducted by bootstrapping. We then created a weighted point-scoring system from logistic regression coefficients.

Results

4076 (97.12%) negative and 121 (2.88%) positive test results were analysed. The majority were young (mean age: 38.0), female (69.8%) and asymptomatic (67.8%). Characteristics that correlated with PCR-positivity included close-contact professions (physicians, nurses, physiotherapists), flu-like symptoms (e.g., fever, rhinorrhoea, headache), abdominal symptoms (nausea/emesis, abdominal pain, diarrhoea), less days since symptom onset, and contact to a SARS-COV-2 positive index-case. Variables selected for the final model included symptoms (fever, cough, abdominal pain, anosmia/ageusia) and exposures (to SARS-COV-positive individuals and, specifically, to positive patients). Internal validation by bootstrapping yielded a corrected Area Under the Receiver Operating Characteristics Curve of 76.43%. We present sensitivity and specificity at different prediction cut-off points. In a subgroup with further workup, asthma seems to have a protective effect with regard to testing result positivity and measured temperature was found to be less predictive than anamnestic fever.

Conclusions

We consider low threshold testing for health care workers a valuable strategy for infection control and are able to provide an easily applicable triage score for the assessment of the probability of infection in health care workers in case of resource scarcity.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-022-07627-5.

PARIS score for evaluation of probability of SARS-CoV-2 infection in cancer patients

Control of transmissible diseases as COVID-19 needs a testing and an isolation strategy. The PARIS score developed by Torjdman et al. was aimed at improving patient selection for testing and quarantining but was derived from a general population. We performed a retrospective analysis of the validity of the PARIS score in a cancer patient population. We included 164 patients counting for 181 visits at the emergency department of the Jules Bordet Institute between March 10th and May 18th which had a SARS-CoV-2 RT-PCR test at admission. Twenty-six cases (14.3%) were tested positive with a higher proportion of positive tests among hematological patients compared to those with solid tumors (26% vs 11% p = 0.02). No clinical symptoms were associated with a positive SARS-CoV-2 PCR. No association between anticancer treatment and SARS-CoV-2 infection was found. The PARIS score failed to differentiate SARS-CoV-2-positive and SARS-CoV-2-negative groups (AUC 0.61 95% CI 0.48–0.73). The negative predictive value of a low probability PARIS score was 0.89 but this concerned only 11% of the patients. A high probability PARIS score concerned 49% patients but the positive predictive value was 0.18. CT scan had a sensitivity of 0.77, specificity 0.51, a positive predictive value of 0.24, and a negative predictive value of 0.92. The performance of the PARIS score is thus very poor in this cancer population. A low-risk score can be of some utility but this concerns a minority of patients.

Clinical prediction models for diagnosis of COVID-19 among adult patients: a validation and agreement study

BACKGROUND

Since the beginning of the pandemic, hospitals have been constantly overcrowded, with several observed waves of infected cases and hospitalisations. To avoid as much as possible this situation, efficient tools to facilitate the diagnosis of COVID-19 are needed.

OBJECTIVE

To evaluate and compare prediction models to diagnose COVID-19 identified in a systematic review published recently using performance indicators such as discrimination and calibration measures.

METHODS

A total of 1618 adult patients present at two Emergency Department triage centers and for whom qRT-PCR tests had been performed were included in this study. Six previously published models were reconstructed and assessed using diagnostic tests as sensitivity (Se) and negative predictive value (NPV), discrimination (Area Under the Roc Curve (AUROC)) and calibration measures. Agreement was also measured between them using Kappa's coefficient and IntraClass Correlation Coefficient (ICC). A sensitivity analysis has been conducted by waves of patients.

RESULTS

Among the 6 selected models, those based only on symptoms and/or risk exposure were found to be less efficient than those based on biological parameters and/or radiological examination with smallest AUROC values (< 0.80). However, all models showed good calibration and values above > 0.75 for Se and NPV but poor agreement (Kappa and ICC < 0.5) between them. The results of the first wave were similar to those of the second wave.

CONCLUSION

Although quite acceptable and similar results were found between all models, the importance of radiological examination was also emphasized, making it difficult to find an appropriate triage system to classify patients at risk for COVID-19.