RESULTS
Patient and clinical characteristics
Of the more than 20,000 patients enrolled in the CHECK program, 373 had SCD. Twelve outliers with inpatient expenditures more than $100,000 in any CHECK year were excluded from analyses because such patients were expected to have unique medical problems beyond their SCD5,33. Therefore, the analytic sample included 361 cases. Fifty-two percent of these 361 SCD patients were “engaged” for CHECK services, which were tailored to their individual needs. Table 1 shows the demographics and comorbidities in these 361 participants stratified by High utilization (n=32), Medium utilization (n=173) and Low utilization (n=156) groups. Statistical tests were conducted across risk utilization groups at baseline. Age and percent male did not differ significantly across the three risk groups, nor did the percentage of patients who were engaged versus enrolled in the CHECK program. Splenic sequestration history was not significantly different across the three risk groups. Only stroke and respiratory disease varied significantly across the SCD hospitalization utilization risk groups: (Stroke, High utilization risk group= 15.6%, Medium utilization risk group=4.6%, and Low utilization risk group= 1.9%, p=.0007; Respiratory Disease High utilization risk group= 81.2%, Medium utilization risk group=60.7%, and Low utilization risk group= 46.2%, p<.001). For all the comorbidities, symptoms were significantly highest in the High risk group and lowest in the Low risk group
Analysis of inpatient expenditures
The utilization of acute care services was predicted to be associated with total expenditures because published studies show that acute inpatient expenditures are the dominant cost in SCD39,40. As expected, inpatient expenditures mirrored the trends of total expenditures across the three risk group categories (see Table 2). Many SCD patients had no inpatient expenditures. A two-part analysis accommodated the semi-continuous expenditure data - fitting a continuous model allowing for data with excess zeros. The results suggested that the effect of utilization risk group on inpatient expenditure varied by CHECK year. For utilization risk group comparisons, the first part of the analysis estimated the percent expenditure reduction for each CHECK year while the second part estimated the odds of having zero expenditures for each CHECK year (see Table 3).
The results suggested that the effect of utilization risk on expenditure varied by CHECK year. In the Baseline year, both Medium and Low utilization risk groups had lower expenditures compared to High utilization risk groups. During the first year in CHECK, the odds of having zero inpatient expenditures for patients in the Low risk group was 7.34 times those in the High risk group and the odds of having zero inpatient expenditure for patients in the Medium risk group was 3.54 times those in the High risk group. At baseline, 95% of patients in the low risk utilization group had zero expenditure compared to 22% in the high risk group.
Looking at expenses a different way, Figure 1 shows the frequency distribution of logarithm transformed expenditures for children in the three tiers of utilization. High utilizers (n=32; Panel A) began with a broad range of expenditures, then all but a few had reduced expenditures over the next two years, ending with a bimodal distribution. Wilcoxon pairwise tests suggested that inpatient expenditures during the second year in CHECK were significantly lower compared to Baseline year (adjusted p-value = 0.02). The other two comparisons (Year 1 compared to Baseline year (p=0.209) and Year 2 compared to Year 1 (p=0.42)) were not significantly different because the small sample size of high utilization risk group limits statistical power.
Figure 1, Panel B shows that Medium utilizers (n=173) began with a bimodal distribution. Inpatient expenditures during the second year in CHECK were significantly reduced compared to the first year (p = 0.004) and Baseline year (p = 0.002). The first year vs. baseline year was not significant (p=0.675). Figure 1, Panel C shows that Low utilizers (n=156) also began with a bimodal distribution of expenditures. Expenditures for the low utilizer tier increased over time. As expected, higher expenditures were associated with more hospital days; some were elective hospitalizations such as tonsillectomy and others were hospitalizations for unpredictable sickle cell complications. Using pairwise comparisons, inpatient expenditures for the second and first years in CHECK were significantly increased compared to the Baseline year (p < 0.001). The second compared to the first-year expenditures were not significantly different (p = 0.672).