A total of 148 patients were analyzed. Mean age was 51.29years (S. D= 20.22) with a mean age in survival group 48.4years (S. D =19.94) and mortality group 59.10years (S. D=19.1). The maximum number of people lied in age-group 60-70 years (n=28, 18.9%) followed by 20-30 and >70 years both of which have the same numbers. Data is negatively skewed (-0.217). In the study, there were more females (88, 59.5%) than males (60, 40.5%).
Most of the patients lie in group with RDW >15.6 (n=60, 40.5%). (Figure ‑1). Mean RDW was 15.933 (S.D =2.69). Data for RDW groups was negatively skewed (-0.678).
As data did not follow normal distribution (negatively skewed) nonparametric test (Mann-Whitney U test) was done to test the difference of distribution of Age, RDW, APACHE II and SOFA across the categories of clinical outcome (improved and mortality). The test showed a significant difference between the improved and mortality group with a p-value of 0.005, 0.000, 0.000, 0.002 for age, RDW, APACHE II, SOFA respectively (Table 1).
Table 1 Mann-Whitney U test for predicting mortality among septic patients.
Variable
|
Mann-Whitney U-Test
|
p-value
|
Age
|
2808.5
|
0.005
|
RDW
|
3422.0
|
0.000
|
APACHE II
|
3119.5
|
0.000
|
SOFA
|
2866.5
|
0.002
|
Binary logistic regression analysis was done to analyze the effect of confounding factors like age, sex, presence of septic shock on mortality. Results showed no significant effect of these confounding factors on mortality except for sex (p=0.029, Odds ratio= 2.950, 95% C.I =1.120-7.773) (Table 2). Among the predictive scores viz. RDW, APACHE II, and SOFA scores; only RDW had a significant difference in predicting mortality with an odds ratio of 1.551 (p=0.000003, 95% C.I =1.292-1.863). So RDW is a better prognostic test to predict mortality in septic patients.
Table 2 Binary logistic regression analysis of confounding factors and prognosis predictive scores
|
Outcome
|
p-value
|
Odds Ratio
|
95% C.I
|
Improved/Cured (N=108)
|
Mortality (N=40)
|
Mean (S.D)
|
Row %
|
n
|
Mean (S.D)
|
Row %
|
n
|
|
|
Lower
|
Upper
|
Age (years)
|
48.4 (19.94)
|
73.0%
|
108
|
59.10 (19.1)
|
27.0%
|
40
|
0.101
|
1.250
|
0.958
|
1.632
|
Hematocrit %
|
35.3
(8.8)
|
73.0%
|
108
|
33.6 (10.1)
|
73.0%
|
40
|
0.315
|
0.979
|
0.941
|
1.020
|
SOFA
|
6 (3)
|
73.0%
|
108
|
8 (3)
|
27.0%
|
40
|
0.062
|
1.221
|
0.990
|
1.506
|
APACHE II
|
16 (7)
|
73.0%
|
108
|
21 (7)
|
27.0%
|
40
|
0.157
|
1.053
|
0.983
|
1.131
|
RDW
|
15.2 (2.2)
|
73.0%
|
108
|
17.9 (2.9)
|
27.0%
|
40
|
0.000003
|
1.551
|
1.292
|
1.863
|
Sex
|
Male
|
-
|
65.0%
|
39
|
-
|
35.0%
|
21
|
0.029
|
2.950
|
1.120
|
7.773
|
Female
|
-
|
78.4%
|
69
|
-
|
21.6%
|
19
|
Septic shock
|
Yes
|
-
|
63.6%
|
28
|
-
|
36.4%
|
16
|
0.555
|
0.713
|
0.231
|
2.194
|
No
|
-
|
76.9%
|
80
|
-
|
23.1%
|
24
|
Patients were further divided into two groups: (a) sepsis and (b) septic shock. Out of 44 patients with septic shock 16 died (36.4 %) and among 104 patients without septic shock, 24 died (23.1%) with odds ratio of 0.713 (p=0.555, 95% C.I= 0.231-2.194)(Table 2). Overall mortality was 27.02% (n=40).
Table 3 Binary logistic regression of RDW group and outcome
RDW Classification
|
Improved/Cured (N=108)
|
Mortality (N=40)
|
Odds Ratio
|
p-Value
|
95% C.I
|
N
|
%
|
N
|
%
|
|
|
Lower
|
Upper
|
|
<13.1
|
10
|
9.3%
|
0
|
0.0%
|
0.000
|
0.003
|
0.000
|
0.000
|
|
>13.1-14
|
27
|
25.0%
|
1
|
2.5%
|
0.00
|
0.999
|
0
|
0
|
|
>14-15.6
|
39
|
36.1%
|
11
|
27.5%
|
0.042
|
0.003
|
0.005
|
0.332
|
|
>15.6
|
32
|
29.6%
|
28
|
70.0%
|
0.332
|
0.008
|
0.139
|
0.746
|
|
RDW group analysis showed no mortality in RDW <13.1 group, 3.6% mortality in RDW >13.1-14 group, 22.0% mortality in RDW >14-15.6 group and 46.7% mortality in >15.6) RDW group (Table 3). Significant mortality difference was seen in >14-15.6 and >15.6 RDW groups with p-value 0.003 and 0.008 respectively. This shows an increasing trend of mortality with the increase in RDW value and vice-versa.
Receiver Operating Characteristic (ROC) curve was used to test the efficacy of different clinical scores viz. RDW, SOFA, APACHE II to predict mortality in septic patients (Figure‑2). Area under the ROC curve was analyzed which shows RDW, APACHE II and SOFA were fair tests to predict mortality in sepsis with AUC of 0.734 (95% C.I =0.649-0.818; p-value=0.000), 0.7.28 ( 95% C. I= 0.637 to 0.819; p-value=0.000), and 0.680 (95% C.I 0.591-0.770; p-value=0.001) respectively. (Table 4). AUC of RDW is >0.7 which is considered a fair test.
Table 4 Area under the ROC curve for RDW, APACHE II, SOFA to predict mortality of sepsis
Test Variable(s)
|
Area
|
Sig.
|
95% Confidence Interval
|
|
Lower Bound
|
Upper Bound
|
|
SOFA
|
0.680
|
0.001
|
0.591
|
0.770
|
|
RDW
|
0.734
|
0.000
|
0.649
|
0.818
|
|
APACHE II
|
0.728
|
0.000
|
0.637
|
0.819
|
|
RDW value of 15.05 has a sensitivity of 73% (positive likelihood ratio= 1.82) and specificity of 60% (negative likelihood ratio =0.45) while RDW value of 16.1 has sensitivity of 56% (positive likelihood ratio=2.07) and specificity of 73% (negative likelihood ratio = 0.6). Youden Index was maximum (37%) at RDW value 14.75 which has a sensitivity of 83% (positive likelihood ratio=1.81) and specificity of 54% (negative likelihood ratio=0.32). Increasing the value of RDW decreases the sensitivity of the test and increases the specificity of the test.