Chapter 2 HEALTH-RELATED QUALITY OF LIFE AMONG PATIENTS
3.3.2. Racial/ethnic disparities in anaemia complication
Figure 3.4 presents the crude percentage of recorded anaemia (no adjustments) stratified by race/ethnicity from 2008 to 2016. It is notable that Native Americans had the highest percentage of anaemia compared to other races, and that this pattern was consistent over time - a sharp rise in 2011 followed by steady increases between 2011 and 2014, and a stabilised period from 2014 onward (though with fluctuations). A similar pattern could be observed among Hispanic Americans and other races/ethnicities. In contrast, the upward trend of recorded anaemia was less obvious in other groups, with higher recorded anaemia in Asian relative to Black and White Americans.
Figure 3.4. Percentage of patients with recorded anaemia from 2008 to 2016, by race/ethnicity
20 25 30 35 40 45 50 55
2008 2009 2010 2011 2012 2013 2014 2015 2016
White Americans Native Americans Black Americans
Hispanic Americans Asian Americans Other races/ethnicities
99
Table 3.2 presents the results from four sets of logistic regression analysis combined with linear splines to explore the potential racial/ethnic disparities in anaemia complication. Race/ethnicity remained a strong predictor of recorded anaemia after controlling for a range of socio- demographic and clinical factors. In the fully adjusted model, compared to White Americans, the odds of being diagnosed with anaemia was highest for Native Americans (adjusted Odds ratio aOR 1.22, 95%CI 1.18-1.27), and Asian Americans (aOR 1.22, 95%CI 1.20-1.25), followed by Hispanic Americans (aOR 1.13, 95%CI 1.11-1.14) and Black Americans (aOR 1.06, 95%CI 1.05-1.07). Admissions who had iron deficiency, proteinuria, diabetes with complications, higher ACCI score, in private hospital or hospital from urban area, and aged less than 50 years or more than 80 years had a higher risk of anaemia compared to other groups (Appendix 2, Table S2).
Table 3.2. The relative likelihood of inpatient CKD-related anaemia among ESRD patients by races/ethnicities
Model 1 Model 2 Model 3 Model 4
OR 95% CI aOR 95% CI aOR 95% CI aOR 95% CI
White Americans 1.0 (Reference) 1.0 (Reference) 1.0 (Reference) 1.0 (Reference)
Native Americans 1.38*** 1.33 1.42 1.27*** 1.22 1.31 1.32*** 1.28 1.36 1.22*** 1.18 1.27 Black Americans 1.07*** 1.06 1.08 1.06*** 1.05 1.07 1.06*** 1.05 1.06 1.06*** 1.05 1.07 Asian Americans 1.27*** 1.25 1.30 1.23*** 1.21 1.26 1.22*** 1.20 1.25 1.22*** 1.20 1.25 Hispanic Americans 1.18*** 1.17 1.20 1.15*** 1.13 1.16 1.13*** 1.12 1.14 1.13*** 1.11 1.14
Others 0.98 0.96 1.00 0.98* 0.96 1.00 0.97** 0.95 0.99 0.98 0.96 1.00
Note: Model 1: unadjusted model. Model 2: adjusted for demographic and socioeconomic variables (insurance type, gender, age at admission, median household income for patient's ZIP Code, location). Model 3: adjusted for clinical variables (renal replacement therapies, iron deficiency, proteinuria, age adjusted comorbidity score ACCI, hospital characteristics, diabetes with or without complications). Model 4: fully adjusted model for all demographic, socioeconomic and clinical variables. OR: odds ratio, aOR: adjusted odds ratio. * p<0.05, ** p<0.01, ***
p<0.001. CI – Confidence intervals.
100
3.1.1.1. Exploration of anaemia likelihood across different groups
Figure 3.5 and Figure 3.6 depict the likelihood of being diagnosed with anaemia over the 9-year period across races/ethnicities, gender, type of RRTs, income quartiles and type of insurance groups. The values presented in these figures were estimated from Model 4 (Table 3.2). The overall pattern is consistent across all comparisons wherein Native Americans had a higher likelihood of anaemia than Whites and where there was a sharp increase in the likelihood of anaemia for all races/ethnicities in 2011 followed by a steady increment afterwards (Figure 3.5).
Figure 3.5. The predicted probability of anaemia from the fully adjusted logistic regression model across races/ethnicities compared to Whites
Admissions receiving a kidney transplantation had lower likelihood of recorded anaemia compared to those receiving dialysis, and HD showed better anaemia outcomes compared to PD (Figure 3.6). The likelihoods of anaemia in ESRD without RRTs group overlapped those of the transplantation group. Regarding insurance type, the likelihood of anaemia was lowest for those with private insurance or Medicare, followed by Medicaid and other insurances, although the odds of anaemia of those with private and Medicare insurance were not significantly different. Female had the
Native/Black/Hispanic/Asian Americans White Americans
101
higher odds of recorded anaemia compared to male across all years. Finally, the lowest income quartile group had the lowest odds of anaemia, while the second highest income quartile group had the highest odds of anaemia (Figure 3.6).
Figure 3.6. The predicted probability of anaemia from the fully adjusted logistic regression model across gender, insurance types, income levels, and RRT types over time
3.3.2.1. The association of PPS and ACA and the likelihood of recorded anaemia Linear splines with knots placed at each policy change point and logistic regression analysis were used to investigate the association of PPS and ACA and the likelihood of recorded anaemia. Results from the fully adjusted logistic regression model (Model 4, Table 3.2) showed that the odds of recorded anaemia after PPS was 1.20 (95%CI:
1.19-1.20) and after ACA was 1.08 (95%CI: 1.07-1.09) compared to the preceding periods. Difference in difference (DiD) analysis – looking at the experience of a specific race/ethnicity over different periods – combined with marginal effect analysis revealed that the changes in the differences between racial/ethnic groups before and
Male/ No RRTs/ Income 1st quartile/
Other type of insurance
Female/ Transplantation/ Income 2nd quartile/ Medicare
Haemodialysis/ Income 3rd quartile/
Medicaid
Peritoneal dialysis/Income 4th quartile/
Private insurance
102
after the introduction of PPS and ACA were significant, where the gap was most evident for Native Americans and Asian Americans in both cases (Table 3.3). For Native Americans, relative to Whites, the difference in the probability of recorded anaemia was 4.32% before PPS. This difference increased slightly after PPS by 0.28%
(in 2012), 0.46% (in 2013), and 0.59% (after ACA). In the aftermath of ACA while the two continued to diverge the difference in the difference in probability of recorded anaemia increased by 0.4% (in 2015) and 0.41% (in 2016).
Table 3.3. The association of PPS/ACA and racial/ethnic disparities PPS Incremental
probability ACA Incremental probability
White Americans Reference Reference
Native Americans
Before PPS 4.32% Before ACA 4.51%
2012 +0.28%*** 2015 +0.4%***
2013 +0.46%*** 2016 +0.41%***
After ACA +0.59%***
Black Americans
Before PPS 1.22% Before ACA 1.28%
2012 +0.09%*** 2015 +0.14%***
2013 +0.15%*** 2016 +0.15%***
After ACA +0.2%***
Asian Americans
Before PPS 4.29% Before ACA 4.49%
2012 +0.29%*** 2015 +0.4%***
2013 +0.47%*** 2016 +0.42%***
After ACA +0.6%***
Hispanic Americans
Before PPS 2.49% Before ACA 2.61%
2012 +0.18%*** 2015 +0.27%***
2013 +0.3%*** 2016 +0.28%***
After ACA +0.39%***
Note: dy/dx marginal effect analysis results. *** p<0.001.