SARS-COV-2 comorbidity network and outcome in hospitalized patients in Crema, Italy

Abstract

We report onset, course, correlations with comorbidities, and diagnostic accuracy of nasopharyngeal swab in 539 individuals suspected to carry SARS-COV-2 admitted to the hospital of Crema, Italy. All individuals underwent clinical and laboratory exams, SARS-COV-2 reverse transcriptase-polymerase chain reaction on nasopharyngeal swab, and chest X-ray and/or computed tomography (CT). Data on onset, course, comorbidities, number of drugs including angiotensin converting enzyme (ACE) inhibitors and angiotensin-II-receptor antagonists (sartans), follow-up swab, pharmacological treatments, non-invasive respiratory support, ICU admission, and deaths were recorded. Among 411 SARS-COV-2 patients (67.7% males) median age was 70.8 years (range 5-99). Chest CT was performed in 317 (77.2%) and showed interstitial pneumonia in 304 (96%). Fatality rate was 17.5% (74% males), with 6.6% in 60-69 years old, 21.1% in 70-79 years old, 38.8% in 80-89 years old, and 83.3% above 90 years. No death occurred below 60 years. Non-invasive respiratory support rate was 27.2% and ICU admission 6.8%. Charlson comorbidity index and high C-reactive protein at admission were significantly associated with death. Use of ACE inhibitors or sartans was not associated with outcomes. Among 128 swab negative patients at admission (63.3% males) median age was 67.7 years (range 1-98). Chest CT was performed in 87 (68%) and showed interstitial pneumonia in 76 (87.3%). Follow-up swab turned positive in 13 of 32 patients. Using chest CT at admission as gold standard on the entire study population of 539 patients, nasopharyngeal swab had 80% accuracy. Comorbidity network analysis revealed a more homogenous distribution 60-40 aged SARS-COV-2 patients across diseases and a crucial different interplay of diseases in the networks of deceased and survived patients. SARS-CoV-2 caused high mortality among patients older than 60 years and correlated with pre-existing multiorgan impairment.

Fig 1. Age and sex-stratification of the outcomes (CPAP/NIV, ICU admission, and death) in SARS-CoV-2 patients.

Bars are numbers of patients. Females are represented in the upper softer bars.

Fig 2. Receiver operating characteristic curve analysis of the logistic regression that measured the predictive power of swabs outcome in CT scan classification.

AUC (area under the curve) estimates how a variable is far from the random predictive power (solid black line, AUC = 0.5).

Fig 3. Comorbidity networks.

A cohort of 270 Covid-19 patients was divided in three classes according to the year of birth: a) 1920–1940, b) 1941–1960, c) 1961–1980. For each group a comorbidity network was derived according to the constraint that at least 10% of patients must share two diseases. Nodes are ordered clockwise according to their degree (number of connection each disease has with other nodes), while node size is associated with the number of patients sharing the disease. The color of each nodes refers to the community a disease belongs to.

Fig 4. Comorbidity network of deceased patients.

a) The comorbidity network was obtained according to patients outcome (deceased/survived). It was derived according to the constraint that at least 10% of patients must share two diseases. Nodes are ordered clockwise according to their degree (number of connection each disease has with other nodes), while node size is associated with the number of patients sharing the disease. The color of each nodes refers to the community a disease belongs to; b) Pie chart of the Charlson Comorbidity Index calculated for each diseased patient. The frequency of patients associated with a specific value of the index is reported within the brackets.

Fig 5. Comorbidity network of survived patients.

a) The comorbidity network was obtained according to patients outcome (deceased/survived). It was derived according to the constraint that at least 10% of patients must share two diseases. Nodes are ordered clockwise according to their degree (number of connection each disease has with other nodes), while node size is associated with the number of patients sharing the disease. The color of each nodes refers to the community a disease belongs to; b) Pie chart of the Charlson Comorbidity Index calculated for each survived patient. The frequency of patients associated with a specific value of the index is reported within the brackets.

Fig 6. Statistically significant difference between normalized Laplacian of deceased and survived networks.

The difference resulted significant with p<0.007. The grey distribution is the result of the difference between 500,000 randomization of survived and deceased normalized Laplacian matrices. The red dot shows the difference between the original survived and deceased normalized Laplacian matrices. The dotted vertical lines represent the 0.05 tails of the distribution. Original image created with the software Gelphi 0.9.1 https://gephi.org/.