Clinical Predictors of All-Cause In-Hospital Mortality In Patients with Sickle Cell Disease in the United States- First Reported Results From A Nationally Representative Sample

Sickle cell disease is the most common monogenic abnormality affecting 70,000–100,000 people in the US, with a reported in-hospital mortality rate of 0.3–0.7%. Predictors of mortality in SCD however remain largely undefined and under-explored. In 2000, data from the Cooperative Study of Sickle Cell Disease had proposed dactylitis, severe anemia and leukocytosis as predictors of adverse events, including death in SCD (Miller et al, NEJM, 2000 13;342(2):83–9). Recently, various demographic predictors of in-hospital mortality including hospital volume, hospital teaching status, and patient socioeconomic status were proposed (McCavit et al Amer Jour Hematology, 2011, 86(4):377–80). However, till date no study has explored the clinical predictors of in-hospital mortality in patients with SCD.

This study aims to explore the various clinical predictors of all cause mortality in hospitalized SCD patients using a nationally representative database.

We used the Nationwide Inpatient Sample (NIS) database for the study. NIS is a stratified probability sample of 20% of all hospital discharges among U.S. community hospitals (n=1,044, sampling universe = 90% of all such discharges) sponsored by the Agency for Health Care Quality and Research.

Presence of following ICD-9-CM diagnosis codes were used to define SCD related hospitalizations: 282.41, 282.42, 282.5, 282.60, 282.61, 282.62, 282.63, 282.64, 282.68, and 282.69. ICD-9-CM 99.04, 99.01 were used to identify packed red cell transfusion and exchange transfusion respectively. ICD-9-CM code 517.3 was used to identify ACS and 995.91, 995.92, and 038 were used to identify sepsis as a diagnosis during hospitalization for SCD. Only data on SCD related hospitalizations and mortality were analyzed.

Sampling weights were applied to represent all community hospital discharges in the US for the year 2007. Univariate and multivariable logistic regression was performed for statistical analysis. A 2-sided p-value of <0.05 was considered significant.

In 2007, there were a total of 166,084 admissions with a diagnosis of SCD. Of these, 86,318 discharges had SCD as the primary diagnosis upon admission and 79,766 had SCD as one of the secondary diagnoses during the admission. Of these, males represented 37.5% of SCD related hospitalizations. There were a total of 844 deaths reported in the above admissions with an all-cause in-hospital mortality rate of 0.5%. Of these, 420(49.7%) were females.

In the multivariable logistic regression model, the following clinical factors were independently associated with statistically significantly higher odds of mortality: Male gender, every ten year increase in age, history of at least one PRBC transfusion, history of exchange transfusion, diagnosis of acute chest syndrome and development of sepsis during the admissions (Table 1). On univariate analysis, history of mechanical ventilation (either non-invasive or invasive) was also a highly significant independent predictor of mortality [estimated OR (95% CI) = 120.9 (82.7–176.9)]. However, we did not include this in the final model as it can be an intermediary for other factors in the pathway leading to mortality and thus potentially introduce confounding in our analysis.

This study proposes for the first time, a set of clinical factors which are independent predictors of all –cause mortality in hospitalized patients with SCD. We propose that early and targeted aggressive therapy based on the presence of these factors should guide the management of hospitalized SCD patients for improved mortality.

Krishnamurti:Daiichi Sankyo Company, Ltd. and Eli Lilly and Company: Research Funding.