Data from the year 2020 was extracted from the National In-patient Sample (NIS). The Agency supported the Healthcare Cost Utilization Project (HCUP) for Healthcare Research and Quality (AHRQ), and the NIS is a portion of a family of databases generated by the Healthcare Cost Utilization Project (HCUP). The NIS was designed and sustained by the AHRQ and is the largest publicly available all-payer inpatient database designed to produce U.S. regional and national estimates of patient utilization, admission, fee, quality, and outcomes. It was designed as a stratified probability sample representing all non-federal acute care hospitals nationwide. The attributes of the design & description of NIS can be found at https://www.hcup-us.ahrq.gov/nisoverview.jsp.
Discharge details comprise patient demographics, primary payer, hospital features, principal diagnosis, secondary diagnoses, and procedural diagnoses. Next, hospitals are stratified according to ownership/control, bed dimensions, teaching class, urban/rural location, and geographic area. Then, a 20% probability sample of all hospitals within each stratum is collected. Finally, all discharges from these hospitals are recorded and weighted to represent nationally. Data from 48 statewide data organizations (47 States plus the District of Columbia) encompassing more than 97% of the U.S. population was included in the NIS sampling frame.
Inclusion Criteria and study variables
We conducted a retrospective study design of hospitalizations using NIS year 2020 with a principal diagnosis of NF with a secondary diagnosis of with and without COVID-19 infection in acute-care hospitals across the United States. Hospitalizations were selected from the NIS database (online at http://www.hcup-us.ahrq.gov). The study population consisted of all inpatient hospitalizations recorded in the NIS 2020 for patients 18 years and above, meeting our diagnostic criteria (Table below). Study variables included age, gender, race, hospital characteristics (teaching vs. nonteaching, bed size; small, medium, and large), hospital region (Northeast, Midwest, south, and west), insurance (Medicare, Medicaid, private and others), Median annual income expected for patient’s zip code, medical comorbidities, primary and secondary outcomes (outlined below). We used the following ICD-10 codes to identify principal and secondary diagnoses: Neutropenic Fever ICD 10 codes; D709, R50.81, and COVID-19 Infection; U071 (supplementary table 1). In addition, we studied baseline characteristics, inpatient mortality predictors, and outcomes (primary and secondary) for Neutropenic Fever hospitalization with COVID-19 vs. without COVID-19. The patient charge represents the charge the hospital billed for the stay but does not reflect the cost of care. In addition, the healthcare cost and utilization project provides data containing hospital-specific cost-to-charge ratios based on all-payer inpatient costs. This cost information is obtained from the hospital accounting reports the Centers for Medicare and Medicaid Services collected. The cost information was obtained from the hospital accounting reports collected by the Centers for Medicare and Medicaid Services [6].
Outcomes Measured
The primary outcome was inpatient mortality among patients principally admitted with neutropenic fever with vs. without a secondary diagnosis of COVID-19 infection. Secondary outcomes included respiratory failure, hemorrhagic shock, septic shock, acute kidney injury (AKI), patient disposition, and health economic burden defined as longer length of stay (LOS), higher hospital cost, and patient charge.
Baseline patient characteristics included demographics (age, sex, race), primary expected payer, median household income for the patient’s ZIP code, hospital characteristics (teaching vs. nonteaching, bed size; small, medium, and large), hospital region (Northeast, Midwest, south, and west), Elixhauser comorbidities as defined by the AHRQ, which include: Congestive Heart Failure, Myocardial Infarction, Peripheral Vascular Disease, Valvular heart disease, Cardiac arrhythmias, Cerebrovascular Disease, Dementia, Chronic Pulmonary Disease, Obesity, Rheumatic Disease, Peptic Ulcer Disease, Liver Disease, Diabetes without chronic complication, Diabetes with chronic complication, Hypothyroidism, Coagulopathy, fluid and electrolyte disorders, Hemiplegia or paraplegia, Renal disease, Any Malignancy (solid, Leukemia, Lymphoma except skin malignancy), metastatic tumor, AIDS/HIV, Alcohol abuse, Drug abuse, Depression, and Psychoses.
Statistical Analysis
Analyses were performed using STATA (Statistics and Data Science), version 17.0 NP-Parallel Edition (Stata Corp, Texas, USA). Continuous variables were compared using the independent Student t-test, while Fisher exact test was used for proportions variables. Logistic regression was used for binary or dichotomous, or categorical variables. Poisson regression was used for discrete variables due to not normal variable distribution. Linear regression was used for continuous variables. A univariate logistic regression, linear regression, and Poisson regression model analysis were used for unadjusted outcome variables. A univariate model was used to calculate unadjusted odds ratios (ORs) for the primary and secondary outcomes. In contrast, multivariable logistic, linear, and Poisson regression was used to calculate adjusted odds ratios (ORs) for the primary and secondary outcomes. Multiple imputations were used for less than 0.5 percent of missing data for the race variable. All variables with P-values < 0.1 with our univariate analysis were included in a multivariable logistic regression model. All P values were 2-sided, and a P-value < 0.05 was considered statistically significant in the multivariable analysis. Elixhauser comorbidity index was used to adjust for comorbidity burden for the primary and secondary outcomes. The severity of comorbidities was quantified using the Elixhauser comorbidity index. The original weighted Elixhauser scores, developed by Van Walraven et al., were computed and further stratified into groups ( 0, 1, 2, and ≥ 3) [7]. The comorbidity score was then calculated for each patient by summing the individual weights of all comorbidities. Weighted estimates were calculated by applying discharge weight to the unweighted discharge records. Weighted estimates were used for all statistical analyses.
Covariates included in the adjusted models were age, gender, race, insurance, current chemotherapy use, aspirin use, steroid use, nicotine use, end-stage renal disease on dialysis, and Elixhauser comorbidity index which include: congestive heart failure, myocardial infarction, peripheral vascular disease, valvular heart disease, cardiac arrhythmias, cerebrovascular disease, dementia, chronic pulmonary disease, obesity, rheumatic disease, peptic ulcer disease, liver disease, diabetes without chronic complication, diabetes with chronic complication, hypothyroidism, coagulopathy, fluid and electrolyte disorders, hemiplegia or paraplegia, renal disease, any malignancy (solid, leukemia, lymphoma except skin malignancy), metastatic tumor, HIV/AIDS, alcohol abuse, drug abuse, depression, and psychoses.