Back

Oxygen Saturation or breathlessness? How best to predict pneumonia in patients presenting with a cough to primary care

Chambers, E.; Stuart, B.; Moore, M.; Lown, M.

2020-09-11 primary care research
10.1101/2020.09.09.20191650 medRxiv
Show abstract

SYNOPSISO_ST_ABSBackgroundC_ST_ABSDistinguishing pneumonia from acute bronchitis has an important bearing on treatment decisions. The best validated diagnostic decision rule was derived from the GRACE-09 study. The 3Cs study provides oxygen saturations in addition to clinical signs and symptoms. Oxygen saturation is a predictor in hospital settings and may be beneficial in primary care. ObjectivesThis project evaluates whether the data from GRACE-09 and 3Cs studies can be mapped onto each other and be combined dataset to refine the diagnostic accuracy of the GRACE-09 clinical prediction rule for the diagnosis of pneumonia. Patients and MethodsGRACE-09 included 3106 patients with acute cough in a multicentre European study. 3Cs is a UK based study of 28883 patients with acute cough. Binary logistic regression was used in the 3Cs dataset to determine which variables significantly predicted low oxygen saturation (<95%). Statistically significant predictors were determined using area under curve (AUC). This equation was used to model the probability of low oxygen saturation in the GRACE-09 dataset and subsequently included in the GRACE-09 model to determine whether low oxygen saturation is a comparable predictor to breathlessness in predicting pneumonia. ResultsFor 3Cs, using all seven variables achieved an AUC for low oxygen saturation of 0.73 (95% CI 0.72-0.74). Substituting low oxygen saturation into GRACE-09 model gives AUC 0.75 (95% CI 0.69-0.82) compared to original GRACE-09 using breathlessness and gives AUC 0.77 (95% CI 0.73-0.82). ConclusionsLow oxygen saturation and breathlessness are similarly predictive and could be used interchangeably in a modified GRACE-09 prediction model. KEY MESSAGESCan pulse oximetry replace subjective breathlessness assessment in the GRACE-09 prediction rule? Pulse oximetry gives an objective measurement that can be used to assist the diagnosis of pneumonia in primary care.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.