Quality of Care in Hospitalizations for Sickle Cell Disease in the U.S.: Do Low Patient Volumes Mean Higher Mortality?

Sickle cell disease (SCD) affects about 100,000 Americans and is responsible for approximately 100,000 hospitalizations annually at a cost of $500 million. Quality of care in patients hospitalized for SCD has been insufficiently studied. Therefore, we aimed to examine whether three potential determinants of quality care,

1. hospital volume,

2. hospital teaching status, and

3. patient socioeconomic status (SES) are associated with three quality indicators:

4. length of stay (LOS),

5. cost of hospitalization, and

6. mortality for hospitalized patients with SCD.

The data source was the National Inpatient Sample (NIS) from 2003–2005, which includes data on nearly 8 million hospitalizations annually from 1000 hospitals in 37 states. The NIS comprises a 20% stratified sample of US community hospitals that includes all payers. In the NIS, hospital volumes are categorized as small, medium, or large depending on geography and population. SCD-specific hospital volumes were determined by ascertaining the total number of SCD discharges at each hospital. No patient-specific SES data is available in the NIS, so we utilized an indirect marker of SES, the median household income for the patient’s ZIP code. Cases were identified using all ICD-9CM codes for SCD as either the primary or a secondary diagnosis. Patients of all ages were included. Four severity of illness markers, were included as covariates to control for referral bias. Other independent variables we examined were age, gender, year of admission, insurance type, and weekend admission. Multivariable analyses included multiple logistic regression to examine adjusted odds of mortality and mixed linear models to examine adjusted length of stay and costs. 70,748 SCD discharges occurred from 2003–2005. 425 deaths occurred yielding a mortality rate of 6 deaths per 1000 hospitalizations. The median length of stay was 4 days (5^th–95^th %ile range – 1–15 days), and the median cost per hospitalization was $4,100 (5^th–95^th %ile range – $1,200–20,500). Bivariate analyses revealed an inverse relationship between SCDspecific hospital volume and mortality with odds ratios of 3.1, 2.1, and 1.8 for quartiles 1, 2, and 3, respectively, relative to quartile 4 (highest volume.) Multivariable analyses (table), controlling for severity of illness and several other factors, confirmed that a lower SCD-specific hospital volume was associated with significantly increased adjusted odds of mortality and a decreased length of stay. Additionally, self-pay status for insurance type was associated with increased adjusted odds of mortality and a decreased length of stay. Teaching hospitals had higher adjusted costs of care for all patients but decreased LOS in adults only. There was no significant association with mortality and hospital teaching status. SES was not associated with LOS or cost, but lower SES trended toward higher mortality in adults (not shown). In conclusion, these are the first national data that demonstrate that hospitals with low volumes of SCD hospitalizations have significantly higher adjusted mortality rates for SCD patients than those with high volumes. These findings suggest that identification of the specific factors associated with reduced mortality in high-volume hospitals, such as greater adherence to accepted standards of care and clinical practice guidelines, might have the potential to reduce mortality at low-volume hospitals. We also demonstrate that a lack of health insurance coverage is associated with higher adjusted mortality rates and decreased LOS in SCD, a finding which should prompt greater efforts to provide coverage for this frequently underinsured population.

Table