1 INTRODUCTION

The effect of excess weight on morbidity and mortality has been acknowledged from over 2000 years ago. Hippocrates recognized that “sudden death is more common in those who are naturally fat than in the lean”.1 Over the last few decades, the prevalence of obesity in the United States has increased significantly, bearing dramatic social, clinical and economic implications.2-6 Elevated body mass index (BMI) has been proven over the years as an independent risk factor for various cardio-vascular conditions such as ischemic heart disease, acute coronary syndrome, congestive heart failure, atrial and ventricular arrhythmia and sudden cardiac death.7, 8

Sudden cardiac death (SCD) is responsible for about 50% of the mortality from cardiovascular disease in the United States and other developed countries.9, 10 Different clinical parameters including age, co-morbidities, initial cardiac rhythm, and time to return of spontaneous circulation were investigated as predictors of survival in SCD.11 While obesity has been shown to be associated with increased incidence and severity of major cardiovascular risk factors and elevated risk for SCD,8, 12, 13 studies examining its effect on outcomes in SCD victims have shown conflicting results.14-20 Some studies showed increased mortality in patients with BMI > 30 kg/m2 admitted to the hospital following a sudden cardiac death.17, 18 At the same time, several other studies have implied that the “obesity paradox”, described in various cardio-vascular conditions such as acute myocardial infarction (AMI) and heart failure, applies to patients admitted after a sudden cardiac death, showing lower mortality in obese patients.14, 16, 18, 19

We aimed at describing the BMI distribution and baseline characteristics in a nationwide cohort of patients, admitted for an out of hospital sudden cardiac death (OHSCD) in the United States, and the relationship between BMI and in-hospital mortality.

2 METHODS 2.1 Data source

The data were drawn from the National Inpatient Sample (NIS), the Healthcare Cost and Utilization Project (HCUP), and Agency for Healthcare Research and Quality (AHRQ)21, 22 datasets, consisting only of de-identified information; therefore, this study was deemed exempt from institutional review by the Human Research Committee.

The NIS is the largest collection of all-payer data on inpatient hospitalizations in the United States. The dataset represents an approximate 20% stratified sample of all inpatient discharges from U.S. hospitals.23 This information includes patient-level and hospital-level factors such as patient demographic characteristics, primary and secondary diagnoses and procedures, co-morbidities, length of stay (LOS), hospital region, hospital teaching status, hospital bed size, and cost of hospitalization. National estimates can be calculated using the patient-level and hospital-level sampling weights that are provided by the HCUP.

For the purpose of this study, we obtained data for the years 2015 (last quarter) and 2016. International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) was used from the last quarter of 2015 and thereafter for reporting diagnoses and procedures in the NIS database during the study period. For each index hospitalization, the database provides a principal discharge diagnosis and a maximum of 29 additional diagnoses, in addition to a maximum of 15 procedures. The reason we only included the data coded with ICD-10 codes is that the ICD-10 system includes individual codes for BMI values and ranges.

2.2 Study population and variables

We identified patients 18 years of age or older with a primary diagnosis of sudden cardiac death based on ICD-10-CM codes I46.2, I46.8, or I46.9, who had one of the BMI, Z68.x, codes, among the secondary diagnoses. Of notice, these represented only the successfully resuscitated OHSCD patients, since those who were not successfully resuscitated in the field or died in the emergency departments, were not hospitalized. To have the “cleanest” possible data on patients admitted for successfully resuscitated out of hospital cardiac arrest, we avoided including patients with a secondary diagnosis of a cardiac arrest in our analysis due to the fact that these could represent patients who underwent an in-hospital sudden cardiac death or had a prior history of cardiac arrest included as a secondary diagnosis.

The following codes represent the six BMI subgroups we have created for our study: Z68.1, BMI ≤19 kg/m2, under-weight group; Z68.20–25, BMI 20–25 kg/m2, normal-weight group; Z68.26–30, BMI 26–30 kg/m2, over-weight group; Z68.31–35, BMI 31–35 kg/m2, obese I group; Z68.36–39 kg/m2, BMI 36–39, obese II group; Z68.4, BMI ≥40 kg/m2, extremely obese group. In addition to analyzing the individual BMI subgroups mentioned above, we combined the overweigh, Obese I and Obese II groups to compare the outcomes of these patients to the combined group of all the underweight, normal weight and extremely obese patients.

The following patient demographics were collected from the database: age, sex, and race. Prior comorbidities were identified by measures from the AHRQ. For the purposes of calculating Deyo-Charlson Comorbidity Index (Deyo-CCI), additional comorbidities were identified from the database using ICD-10-CM codes. Deyo-CCI is a modification of the Charlson Comorbidity Index, containing 17 comorbidity conditions with differential weights, with a total score ranging from 0 to 33. (Detailed information on Deyo-CCI provided in the Appendix A table). Higher Deyo-CCI scores indicate a greater burden of comorbid diseases and are associated with mortality, 1 year after admission.24 The index has been used extensively in studies from administrative databases, with proved validity in predicting short- and long-term outcomes.25, 26 Our primary outcome in this study was in-hospital mortality. Length of stay was the secondary outcome we analyzed.

2.3 Statistical analysis

The chi-square (χ2) test and Wilcoxon Rank Sum test were used to compare categorical variables and continuous variables, respectively. The NIS provides discharge sample weights that are calculated within each sampling stratum as the ratio of discharges in the universe to discharges in the sample.27 We generated a weighted logistic regression model to identify independent predictors of in-hospital mortality. Candidate variables included patient-level characteristics, Deyo-CCI and hospital-level factors. We retained all predictor variables that were associated with our primary and secondary outcome with p < .05 in our final multivariable regression model. For all analyses, we used SAS® software version 9.4 (SAS Institute Inc., Cary, NC.) A p value <.05 was considered statistically significant.

3 RESULTS 3.1 Study cohort

A total of 466 hospitalizations for successfully resuscitated out of hospital sudden cardiac death patients across the United States during 2015 (last quarter) and 2016 were included in the analysis. After implementing the weighting method, these represented an estimated total of 2330 hospitalizations for OHSCD, in patients with documented BMI during the index hospitalization. The majority of patients (52.4%) were male and the mean age of the cohort was 62.3 ± 29 years.

As shown in Table 1, 62% of the study population were white, majority of 56% had Medicare coverage, 82.3% were in the lower 75th income percentile. As to the clinical characteristics, 39.5% of the study population had history of hypertension, 40.1% had diabetes mellitus, 33.3% had chronic pulmonary disease, 9.9% of the patients had a prior history of an AMI, 10.9% had peripheral vascular disease. The median BMI in the study was 38 (IQR: 31–41) with 85.6% of the patients with BMI above the normal (>25 kg/m2). The data reveal that 16.3% of all hospitalizations for successfully resuscitated OHSCD included patients with a diagnosis of VT/VF and 9.2% of the patients were diagnosed with an acute myocardial infraction (STEMI or NSTEMI). Twelve percent of the patients underwent a percutaneous coronary angiography and 2.8% of the patients required percutaneous coronary intervention.

TABLE 1. Baseline characteristics of the study population (total and per BMI groups) <20 20–25 26–30 31–35 36–39 ≥40 Total P Value Patients, n Unweighted 42 25 48 74 66 211 466 Weighted 210 125 240 370 330 1055 2330 Age group, % <.001 18–44 9.5 8.0 10.4 8.1 7.6 10.4 9.4 45–59 23.8 16.0 20.8 21.6 25.8 34.1 27.7 60–74 31.0 36.0 58.3 52.7 56.1 45.0 47.4 75 or older 35.7 40.0 10.4 17.6 10.6 10.4 15.5 Gender, % <.001 Male 57.1 64.0 56.2 60.8 43.9 48.8 52.4 Female 42.9 36.0 43.8 37.8 56.1 51.2 47.4 Missing 0.0 0.0 0.0 1.4 0.0 0.0 0.2 Race, % <.001 White 69.0 52.0 47.9 63.5 69.7 63.0 62.4 Non-White 31.0 44.0 41.7 16.2 19.7 26.5 26.8 Other/Missing 0.0 4.0 10.4 20.3 10.6 10.4 10.7 Comorbidity, % Hypertension 26.2 36.0 41.7 50.0 47.0 36.0 39.5 <.001 Congestive heart failure 11.9 12.0 20.8 16.2 27.3 37.0 27.0 <.001 Diabetes Mellitus 9.5 20.0 35.4 55.4 45.5 42.7 40.1 <.001 Renal Failure 19.0 36.0 31.3 33.8 34.8 38.9 34.8 <.001 Peripheral Vascular Disease 14.3 16.0 4.2 12.2 9.1 11.4 10.9 .004 Prior MI 4.8 8.0 18.7 12.2 9.1 8.5 9.9 <.001 VT/VF 9.5 16.0 14.6 27.0 19.7 13.3 16.3 <.001 Deyo-CCI, % .008 0 9.5 12.0 14.6 12.2 15.2 10.9 12.0 1 23.8 16.0 20.8 14.9 13.6 13.7 15.7 2 or higher 66.7 72.0 64.6 73.0 71.2 75.4 72.3 Primary payer, % <.001 Medicare 61.9 72.0 56.3 58.1 53.0 53.1 56.0 Medicaid 11.9 12.0 14.6 10.8 13.6 19.9 15.9 Private insurance 11.9 16.0 14.6 23.0 21.2 22.7 20.4 Self-pay 7.1 0.0 10.4 4.1 3.0 1.9 3.6 No charge 2.4 0.0 0.0 0.0 0.0 0.0 0.2 Other/missing 4.8 0.0 4.2 4.1 9.1 2.4 3.9 Income percentile, % <.001 0 to 25th percentile 33.3 52.0 27.1 31.1 31.8 38.9 35.6 26th to 50th percentile 26.2 12.0 18.8 28.4 34.8 26.1 26.2 51st to 75th percentile 21.4 12.0 20.8 24.3 18.2 22.7 21.5 76th to 100th percentile 19.0 20.0 29.2 13.5 12.1 10.9 14.6 Missing 0.0 4.0 4.2 2.7 3.0 1.4 2.1 Hospital status, % <.001 Urban teaching 69.0 80.0 66.7 67.6 80.3 63.0 68.0 Urban nonteaching 23.8 16.0 31.2 23.0 12.1 28.9 24.7 Rural 7.1 4.0 2.1 9.5 7.6 8.1 7.3 Hospital region, % <.001 South 42.9 40.0 52.1 39.2 47.0 45.0 44.6 West 23.8 32.0 14.6 17.6 15.2 14.7 17.0 Midwest 21.4 16.0 18.8 29.7 19.7 28.9 25.3 Northeast 11.9 12.0 14.6 13.5 18.2 11.4 13.1 Hospital bed size, % .159 Large 57.1 52.0 52.1 56.8 48.5 56.9 54.9 Small/medium 42.9 48.0 47.9 43.2 51.5 43.1 45.1 Note: p Values were generated using Chi-square test and refer to differences between BMI groups within baseline characteristics. 3.2 Patients characteristics by BMI group

Baseline characteristics of the study population in the different BMI groups is presented in detail in Table 1. The distribution of the BMI groups varied significantly based on the country regions as well as income percentiles (p < .001). While about three quarters of the OHSCD patients (74.9%) across the country were obese (BMI > 30 kg/m2), the prevalence of obesity among the study population was highest in the Midwest (81.4%) and lowest in the West coast of the United States (68.3%), p < .001. Female predominance was documented in the obese II and the extremely obese groups. Younger age and higher prevalence of comorbidities including hypertension, diabetes and congestive heart failure were documented in the obese patients (Table 1). Among the obese patients, only 12.1% had an annual income in the highest income quartile, compared to 22.7% among nonobese patients (p < .001).

3.3 Length of stay and mortality by BMI groups

The average LOS in the hospital for the study population was 5.51 ± 0.42 days. As shown in Figure 1; the trend of the correlation between BMI and length of stay was linear in nature with longer hospital stay in obese patients, p < .001 (Figure 1).

The relationship between BMI and the length of hospitalization in the study

The overall rate of in-hospital mortality during the study period was documented at 69.3%. A U-shaped relationship between the BMI and the in-hospital mortality was documented, as described in Figure 2. Following the observation that the over-weight, obese I and obese II patient subgroups (BMI 26–39) exhibited significantly lower in-hospital mortality (61%) compared to the other BMI groups (75%), we performed an additional statistical analysis dividing the patients into these two subgroups (Table 2).

The relationship between BMI and in-hospital mortality in the study population

TABLE 2. In-hospital outcomes for the total study population and per BMI group BMI, kg/m2 ≤20 20–25 26–30 31–35 36–39 ≥40 Total p Value Mortality, % 76.2 80.0 64.6 62.2 56.1 74.4 69.3 <.0001 Length of Stay (days), Mean ± SEM 4.86 ± 0.85 3.96 ± 0.41 5.31 ± 0.27 4.76 ± 0.58 6.30 ± 0.57 5.88 ± 0.56 5.51 ± 0.42 <.0001 Note: p Values were generated using Chi-square test and refer to differences between BMI groups. 3.4 Predictors of in-hospital mortality

In an unadjusted analysis, we found several parameters that significantly increased the odds of in-hospital mortality (Table 3). These included: white race and history of congestive heart failure (all with p < .01). In addition, BMI 26–30, BMI 31–35, BMI 36–39 (compared to BMI 20–25) and having BMI between 26 and 39 compared to the other groups combined (below 25 kg/m2 or above 40 kg/m2), decreased the risk for in-hospital mortality (all with p < .01). Personal history of hypertension and diabetes were found as predictors of improved outcomes in a univariate analysis, before adjusting for potential confounders. While admission with a diagnosis of STEMI or NSTEMI did not predict improved outcomes, the small proportion of patients who underwent coronary intervention (2.8%) were found to have lower mortality, Odds Ratio (OR) –0.21, 95% Confidence Interval (95% CI), (0.17–0.28). After adjusting for potential confounders, BMI below 25 kg/m2 or above 40 kg/m2, compared to BMI 26–39 kg/m2, remained an independent predictor of in-hospital mortality in a multivariate analysis (Table 4). Hypertension and diabetes were not found to be independent predictors in a multivariate analysis, implying an interaction between them and other clinical parameters directly effecting the primary outcome. Congestive heart failure, OR −1.29 (1.01–1.65), and Deyo Comorbidity index of ≥2, OR −1.64 (1.19–2.25), were also found to be independent predictors of mortality.

TABLE 3. Univariate analysis for predictors of in-hospital mortality Predictor Probability (95% CI) Odds Ratio (95% CI) p Value Age group, years .487 18–44 72.73% (66.47,78.20) 1.00 (reference) N/A 45–59 70.54% (66.91,73.93) 0.90 (0.64,1.26) .537 60–74 68.33% (65.52,71.00) 0.81 (0.59,1.12) .198