Methods:
Data from the National Inpatient Sample (NIS) were used for this study.
NIS database has been made possible through sponsorship of federal
agency for Healthcare Research and Quality (AHRQ). The main purpose of
AHRQ is to enhance the quality, appropriateness, and effectiveness of
health care services (9). NIS is publicly available all-payer
administrative claims-based database. National estimates of the entire
US hospitalized population were calculated using the Agency for
Healthcare Research and Quality sampling and weighting method.
Institutional review board approval was not required for this study
given the de-identified nature of the NIS and its public availability.
We analyzed the NIS data from January 2007 to December 2017 using the
International Classification of Disease, 9th Revision, Clinical
Modification (ICD-9-CM) and International Classification of Disease,
10th Revision, Clinical Modification (ICD-10-CM) codes. Patients who
sustained in-hospital SCA were identified by applying ICD-9-CM and
ICD-10-CM codes of 99.60, 99.63 and 5A12012, respectively to any
procedure field. These codes indicate utilization of cardiopulmonary
resuscitation (CPR) and well representative of in-hospital SCA from
administrative datasets as shown by the earlier studies (10,11). ESRD
patients were then subsequently identified using ICD-9-CM and ICD-10-CM
codes of 585.6 and N18.6, respectively. Patients were excluded if they
were less than 18 years of age or had acute kidney injury (AKI) and
prior history of renal transplantation. Baseline characteristics and
outcomes were compared in ESRD patients who sustained in-hospital SCA to
non-ESRD patients with in-hospital SCA. Propensity score matching was
also done to balance confounding variables and outcomes were again
assessed in both groups. Trends in in-patient mortality and length of
stay (LOS) were also assessed. Predictors of in-patient mortality in
ESRD patients after a SCA event were also analyzed.
Age, race, median income, urban/rural hospital, US region and Elixhauser
comorbidities were selected for analysis. Descriptive statistics were
presented as frequencies with percentages for categorical variables and
as means with standard deviations or median with interquartile range as
appropriate for continuous variables. Baseline characteristics were
compared using a Pearson𝜒2 test for categorical variables and
independent samples t-test or non-parametric tests for continuous
variables as appropriate. Median LOS, median cost of stay and mortality
were calculated. The median cost of stay was adjusted for inflation (in
comparison to December 2017). Simple linear regression or Chi-Square
test was used for trend analysis over the study years as appropriate. To
mitigate the risk of confounding and selection bias, a nearest neighbor
1:1 propensity score (PS) matching was done using a caliper width of
0.2. In this way, ESRD and non-ESRD patients were well matched with
respect to demographic variables as shown in table 1. Predictors of
mortality in ESRD patients who sustained in-hospital SCA were analyzed
using a logistic regression model. A forward stepwise entry model was
used for this purpose. Initially, all variables, which were
significantly associated with mortality with a p value of less than 0.05
in univariate analysis, were entered in the model from the baseline
table. Subsequently only those variables are retained in the model which
were associated with mortality with a p value of less than 0.10 during
forward entry. A type I error of less than 0.05 was considered
statistically significant. All statistical analyses were performed using
statistical package for social science (SPSS) version 26 (IBM Corp) and
R version 3.5.