Data Source and Study Cohort Selection
Our study cohort included five, common serious infection hospitalizations in people with SSc in the U.S. NIS 1998-2016 sample. The NIS is a 20% stratified sample of discharge records from all participating community hospitals from all participating states [6]. The NIS is the largest publicly available, de-identified all-payer inpatient health care database in the U.S. It has been used for epidemiological studies of hospitalization, mortality and costs, since it represents all hospitalizations in the U.S. The Institutional Review Board at the University of Alabama at Birmingham (UAB) approved this study.
We identified five types of serious infections based on the presence of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes in the primary diagnosis position for hospitalization: (1) opportunistic infections (OI; 010.xx –018.xx, 031.xx, 078.5, 075.xx, 053.xx, 112.4, 112.5, 112.81, 112.83, 130.xx, 136.3, 117.5, 027.0, 039.xx, 117.3, 114.xx, 115.xx, 116.0); (2) skin and soft tissue infections (SSTI; 040.0, 569.61, 681.xx, 682.xx, 785.4, 728.86, and 035.xx); (3) urinary tract infection (UTI; 590.xx); (4) pneumonia (003.22, 481.0, 513.0, 480.xx, 482.xx, 483.xx, 485.xx, 486.xx); and (5) sepsis/bacteremia (038.xx and 790.7), as previously [7, 8]. These diagnostic codes have been shown to be valid in administrative datasets, with positive predictive values of 70% to 100% in people with rheumatoid arthritis [9]. We also used the ICD-10-CM codes for infections for the 2015-2016 data due to a coding system change to ICD-10-CM in 2015 in the U.S. (Appendix 1). Composite infection was defined as any of the serious infection occurring as primary diagnosis for hospitalization. SSc was identified based on the presence of an International Classification of Diseases, ninth or tenth revision, clinical modification (ICD-9-CM or ICD-10-CM) code 710.1 or M34 (includes progressive systemic sclerosis, CREST syndrome, scleroderma, acrosclerosis, but excludes circumscribed scleroderma) in a non-primary position during the index hospitalization. A previous study showed sensitivity of 81% and specificity of 95% using a diagnostic code approach for SSc [10].
Covariates and Outcomes
We adjusted each regression model for covariates/confounders, including age, sex, race, serious infection type (OI, SSTI, UTI, pneumonia, sepsis [reference]), median household income, the insurance payer, hospital characteristics (region, location/teaching status, bed size) and Deyo-Charlson comorbidity index, a validated measure of medical comorbidity that includes 17 comorbidities (myocardial infarction, congestive heart failure, cerebrovascular disease, dementia, renal disease, liver disease, chronic pulmonary disease, diabetes, etc.), based on the presence of ICD-9-CM codes at index admission [11], with higher score indicating more comorbidity load. Deyo-Charlson index was categorized as none, one or two or above, as previously [12-14].
We examined healthcare utilization and in-hospital mortality, details are as follows: (1) health care utilization: total hospital charges above the median for each calendar year; the length of hospital stay above the median of 3 days; and discharge to non-home settings (rehabilitation or an inpatient facility); and (2) in-hospital mortality.
Statistical Analyses
We followed the survey analysis procedures that account for the weights, clusters and strata as defined in NIS, including the modified weights with the change in sampling in 2012. We compared the summary statistics using chi-square or student’s t-test, as appropriate. Rates were calculated per 100,000 NIS claims. We analyzed time-trends in rate of each serious infection using Cochran Armitage test. We performed multivariable-adjusted logistic regression analyses for each study outcome, adjusting for all covariates listed in the section above. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. We used SAS 9.3 (Cary, N.C.) for all analyses. We considered a p-value <0.05 to be statistically significant, which corresponds to a 95% CI that excludes unity.