Study Design, Setting, and Participants
This monocentric retrospective cohort is an ancillary study of the Zerah and al cohort which took place from March 13 to April 15 2020 (10). Informed consent was obtained for all participants and this study was approved by the University Hospital of Paris research committee on April 17, 2020 and by the institutional ethics board of Sorbonne University on May 11th, 2020 (2020-CER-2020-43). This report follows the STROBE recommendations.
Inclusion criteria were: age 70 years old and older, hospitalized for COVID-19 in a COVID-19 unit. Non-inclusion criteria were the patients without health insurance or who refused use of their medical data, other diseases or medications that might modify eosinophil blood cell count: immunosuppression (HIV with CD4 < 200/ mm3), chronic corticosteroid use, asthma, chemotherapy or immunosuppressive agents, documented lymphoproliferative or myeloproliferative disorders and documented parasitic diseases.
The diagnosis of COVID-19-19 was confirmed by a nasopharyngeal polymerase chain reaction (RT-PCR) for SARS-CoV-2, according to the WHO.
Data Collection Methods and Data Management
Patients’ medical records were reviewed and analyzed by trained physicians. We included baseline characteristics before COVID-19-19: age, gender, home or nursing home residence, previous medical history, and chronic medications. Comorbidity severity was assessed with the Charlson’s index (11) and functional status was assessed by the Activities of Daily Living (ADL) scale (6 basic human functions: bathing, dressing, toileting, transfer, continence, and feeding; 1 point for each function (9)). ADL was categorized in a binary variable: ADL < 3. Due to the very low eosinophils count on this cohort, cut-off for eosinopenia was < 10/mm3 because it has been associated with poor outcome during sepsis and defined absolute eosinopenia (17).
Acute events were recorded: acute atrial fibrillation, acute heart failure, acute pulmonary edema, altered consciousness defined by Glasgow score < 14, thromboembolic or hemorrhagic event, acute kidney failure identified according to the Kidney Disease Improving Global Outcomes (KDIGO) definition (26), fecal impaction, urinary retention, pressure ulcer, quick Sepsis-related Organ Failure Assessment ≥ 2 (qSOFA), defined by a range 0-3, with 1 point each for systolic hypotension (≤ 100 mmHg), tachypnea (≥ 22/min) or altered consciousness (Glasgow coma score < 14) (27). Biological data during infection were also collected: white blood cell count, neutrophil count, eosinophil count, lymphocytes count, platelets count, hemoglobin level, presence of a liver injury (aspartate aminotransferase or alanine aminotransferase ≥ 2 times normal level) and cholestasis with alkaline phosphatase or gamma-glutamyltransferase ≥ 2 times normal level. Prescription of antibiotic was recorded. Data at discharge were also reported: vital status (alive or dead), length of stay and discharge location (home; nursing home; long term stay unit rehabilitation center or other).
As we included all patients from the center, no power calculation was done a priori. Normality was assessed by a graphical representation of the distribution. Data are presented with mean and standard deviation (SD) for continuous variables and count (percentage) for categorical variables. A t-test was used for continuous variables and chi-squared test or Fisher’s exact test for categorical variables. Patient characteristics are described overall and according to mortality status during the stay in the acute geriatric stay.
We assessed for missing values and their distributions in the two groups (deceased or survivor). They represented overall less than 8% of all of the data and 4% for the variable used in the multivariate analysis. Thus, no imputation of the missing data was performed.
We performed a logistic regression with adjustment on three variables as such: deceased = Intercept + x * Age + y * Sex + z * ADL. No stepwise was done due to the limited power and the number of missing data. ADL was chosen because it was a significant factor in the main study.
Analyses were performed with R V4.0.0.