This study is the first empirical examination of differences in inpatient costs, length of stay, and in-hospital mortality between patients with and without SCD undergoing high-volume surgical procedures. Cholecystectomy and hip replacement are clinically important for patients with SCD, in that the prevalence of cholelithiasis ranges from 30% to 70% in these patients and avascular necrosis of the hips occurs in up to 50% [16
Patients with SCD who underwent cholecystectomy incurred 46% higher costs and 73% greater length of stay than patients without SCD. Adjustment for patient, procedural, and hospital characteristics mitigated this effect somewhat, but costs remained 43% higher and length of stay 70% greater. Patients with SCD who underwent hip replacement incurred 40% higher unadjusted costs and 82% greater length of stay than patients without SCD. Adjustment for other potentially explanatory variables had a greater impact for hip replacement than for cholecystectomy, but adjusted costs remained significantly higher by 25% and length of stay 56% greater for patients with SCD. When we included length of stay and discharge disposition as covariates in the cost models, the impact of SCD was greatly attenuated, suggesting that the higher costs associated with SCD are due largely to longer inpatient stays. However, it is important to note that a significant association between SCD and inpatient costs persisted, even after adjustment for length of stay.
Inclusion of sickle cell crisis as a covariate in the regression models revealed that sickle cell crisis is responsible for a substantial portion of SCD burden. Among patients with SCD, sickle cell crisis was associated with a 27% or greater incremental increase in costs and a 53% or greater incremental increase in length of stay for both procedures. Nevertheless, even after factoring out the incremental impact of sickle cell crisis, costs and length of stay were significantly greater for patients with SCD.
We restricted this analyses to patients aged 18 to 64 years who were treated at hospitals where at least 1 patient with SCD had undergone one of the procedures of interest. Although we believe the selection criteria strengthened our ability to make valid comparisons, some differences in patient and hospital characteristics remained. For example, the age distributions of patients with and without SCD were significantly different. In addition, the comorbid conditions prevalent in the non-SCD cohorts (ie, diabetes mellitus, hypertension, and obesity) reflected the older age of the patients as compared with the SCD cohorts. This observation is consistent with findings by Adams et al [15
] that cholecystectomy and hip replacement were not associated with body mass index in patients with SCD, despite such trends in the general population. A study of Medicare beneficiaries with SCD in Tennessee found that men aged 20 to 49 years had significantly higher overall mortality than women [3
], which may be partially explained by our observation of a 72% higher mortality rate in men undergoing cholecystectomy compared with women undergoing the same procedure.
The higher prevalence of pulmonary circulatory disorders may reflect the presence of pulmonary hypertension or symptoms of acute chest syndrome, both of which are complications of SCD. It is likely that controlling for comorbid conditions associated with SCD decreased the observed association of SCD with inpatient outcomes in the multivariable analyses. Patients with SCD had much higher rates of coagulation and pulmonary circulatory disorders. These conditions were associated with greater than 100% higher costs for cholecystectomy and 35% higher costs for hip replacement. These comorbid conditions might more appropriately be considered part of the burden of SCD rather than factors to be controlled in a regression model. In addition, some reports have indicated that screening for certain comorbid conditions, such as pulmonary hypertension, may occur more often in patients with SCD [24
]. Increased screening in patients with SCD would result in a positive correlation between these conditions and SCD, which would decrease the magnitude of any perceived effect of SCD on costs, length of stay, and mortality. Thus, controlling for comorbid conditions may have resulted in underestimates of the direct effects of SCD on inpatient outcomes.
This study also provided an opportunity to evaluate the use of perioperative transfusion. Although some reports have suggested that laparoscopic cholecystectomy can be performed without transfusion in patients with SCD [20
], a recent study found that most children with SCD who undergo either a laparoscopic or open abdominal procedure receive a preoperative or postoperative transfusion [21
]. In our study, 35.2% of patients with SCD who underwent laparoscopic cholecystectomy and 46.7% of those who underwent hip replacement received a blood transfusion during the hospital stay. These figures are likely an underrepresentation of the number of patients who received preoperative transfusions, because practitioners may administer transfusions before the hospital admission.
Variables omitted from the regression models could have biased the results if they were correlated with SCD status and reflected the circumstances or severity of the need for surgery, the quality of the procedure, differences in hospital practices, and differences in baseline clinical characteristics that would affect outcomes. To mitigate the effects of hospital characteristics, we limited the analysis to patients with and without SCD from the same hospitals and controlled for remaining differences in hospital characteristics. Other differences between patients with and without SCD, including the expected primary payer, age, and comorbid conditions, were included in the model to adjust for potential confounding. Because there were differences in age distributions between the SCD and non-SCD cohorts even after the exclusion of patients who were not aged 18 to 65 years, we conducted the analyses separately for 3 age groups (18 to 30 years, 31 to 44 years, 45 to 64 years) and had similar results.
Our study has some limitations. First, the NIS data represent hospital discharges, not individual patients. As a result, patients could have been counted more than once. However, given the procedures of interest, we assumed that this possibility had minimal impact. Also, the NIS is limited to in-hospital outcomes; we could not evaluate rehospitalization rates or postdischarge mortality. Therefore, the mortality rates, total costs, and inpatient days associated with surgery may be underestimates.
Compared with other studies using data from the NIS, our sample size is small. This result largely reflects the low prevalence of the combination of SCD with specific surgical procedures. Furthermore, our findings likely underestimate the number of patients with SCD who underwent cholecystectomy or hip replacement, because we excluded patients who were coded as having sickle cell trait only. Although it was our intention to exclude these individuals from the analyses, a previous analysis of children in the NIS reported that at least one third of cases identified with vaso-occlusive and pain crises were coded as having sickle cell trait—presumably in error, as sickle cell trait is not associated with these complications [22
]. However, given the large numbers of patients in the non-SCD cohorts, misclassifying some patients as not having SCD is expected to have a small effect on the results.
Finally, we did not apply the NIS sample weights in this study because we did not intend to make national projections. By limiting the non-SCD cohort to patients treated at hospitals where patients with SCD underwent the same procedures, national estimates would not have been meaningful.
Patients with SCD underwent cholecystectomy and hip replacement at much younger ages than patients without SCD. After adjustment for patient, procedural, and hospital characteristics, patients with SCD were hospitalized for longer periods and incurred higher inpatient costs. Nevertheless, risk of in-hospital death was similar among patients with and without SCD. Future studies of length of stay among patients with SCD might help to identify means by which care for these patients can be managed more effectively.