Construction and validation of a nomogram to predict mortality risk

in HIV/AIDS patients undergoing maintenance hemodialysis

Zhu-rui Xian
Chengdu Medical College
Xiao-fei Song
Public Health Clinical Center of Chengdu
Yong-fu Wang
Public Health Clinical Center of Chengdu
Ting-ting Yang
Public Health Clinical Center of Chengdu
Nan Mao (  [email protected] )
Chengdu Medical College

Research Article

Keywords: HIV/AIDS, hemodialysis, mortality risk factors, predictive model

Posted Date: February 14th, 2024

DOI: https://doi.org/10.21203/rs.3.rs-3910626/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License

Additional Declarations: No competing interests reported.

Page 1/16
Abstract
Objective: This study was aimed to explore the death risk factors in HIV/AIDS patients undergoing maintenance hemodialysis, and
constructed a mortality prediction risk model.

Methods: In this study, we retrospectively collected clinical data of HIV/AIDS patients who received hemodialysis in Chengdu Public
Health Clinical Medical Center between June 6,2017 and June 6,2023, and were divided into survival group and mortality group on the
basis of the follow-up result. Besides, we separated all patients into training set, which was used for model construction, and validation
set for model verification according to 8:2 ratio. The t-test, non-parametric test, chi-square test, fisher’s precise test and ROC analysis
were used for variable selection, and the logistic regression analysis was used for exploring the relationship between variables and
death. And then, we used the stepwise logistic regression to construct a mortality risk prediction model in HIV/AIDS patients undergoing
maintenance hemodialysis, and next, we used R software to visualize the prediction model which called a nomogram. And last, ROC
analysis, calibration curve and decision curve were used for model evaluation, and meanwhile, we used a independent internal validation
set for model verification.

Result: In this study, we collected clinical data of 166 HIV/AIDS patients undergoing maintenance hemodialysis, including 123 patients
in the training set(55 mortalities and 68 survivals)and 43 patients in the validation set(20mortalities and 23survivals). Stepwise Logistic
regression showed that education level [OR(95%CI): 3.754 (1.247-11.300), p=0.019], dialysis age after diagnosis of HIV/AIDS
[OR(95%CI):0.050 (0.013-0.187),p=0.000], creatine kinase isoenzyme (CK-MB)[OR(95%CI): 7.666 (2.237-26.271),p=0.001],neutrophil and
lymphocyte counts ratio (NLR)[OR(95%CI):3.864 (1.266-11.795),p=0.018], magnesium (Mg2+)[OR(95%CI): 4.883 (1.479-
16.118),p=0.009],HIV-RNA[OR(95%CI): 17.616 (3.797-81.721),p=0.000] were independent risk factors of HIV/AIDS patients undergoing
hemodialysis, and afterwards, we constructed a nomogram based on the 6 independent risk factors. The AUC of the prediction model in
ROC analysis was0.921 (95%CI 0.883~0.968), indicating that this nomogram had a good efficacy in predicting mortality. In addition, the
calibration curve and decision curve both showed that the nomogram had good clinical application. Futhermore, there was a same result
in the validation set.

Conclusion: In present study, the nomogram model had a good performance in predicting the mortality of HIV/AIDS patients undergoing
maintenance hemodialysis, which is worth promoting in clinical practice.

Introduction
Acquired immune deficiency syndrome(AIDS)was caused by human immunodeficiency virus (HIV) that attack the immune system of the
human body, leading to a state of immune regulation disorder[1]. There were 38.6 million people had infected HIV, and in which, about
2.5 million patients were dead according to statistics[2].In recent years, because of the globallypopularization of antiretroviral treatment
(ART), the survival rate of HIV/AIDS patients was improved greatly, but still lower than non-HIV infected people. On the basis of available
data, the mortality of HIV/AIDS patients is attributed to AIDS associated complication[3], including cryptococcal meningitis (CM)[4],
pneumocystis pneumonia (PCP)[5], tuberculosis[6], kaposis sarcoma[7], and so on. Additionally, about 30% of HIV/AIDS patients had
abnormal renal function, and 3.5–48.5%[8]of which developed into end stage renal disease (ESRD)[9] eventually, and the incidence tends
to raise year by year in the world[10, 11]. The human immunodeficiency virus and uremia toxin together attacked HIV/AIDS patients with
uremia, which lowered the quality of these patients’ life[12]. In Atta MG, et al’s[13] research, the survivals of HIV/AIDS patients undergoing
hemodialysis were obviously less than patients receiving hemodialysis without HIV/AIDS, and HIV/AIDS patients with ESRD became one
of the cause of death of HIV/AIDS patients, therefore, early identification of mortality risk and intervention is crucial to reduce the
mortality rate and improve life quality of HIV/AIDS patients undergoing maintenance hemodialysis.

There were no researches that discussing the mortality risk of HIV/AIDS patients undergoing hemodialysis until now, and the objects of
existing studies were confined to only HIV/AIDS patients or hemodialysis patients. The previous studies suggested that age, education
level,serum albumin, and anemia are associated with the mortality in HIV/AIDS patients[14–21] or hemodialysis patients[22, 23]. HIV/AIDS
or hemodialysis patients with higher age, lower education level, lower serum albumin, and anemia had shorter survival time[14–15][22, 23].
In addition, CD4 counts and combined opportunistic infection and mortality of HIV/AIDS patients were closely related [14–15]. However,
the efficacy of using single indicator to predict mortality was low or medium. Therefore, researchers began to construct mortality
prediction model by uniting multiple indicators in order to solve this problem. Fang FJ et al[24]construct a mortality prediction model of
HIV/AIDS adult patients receiving ART, risk factors including CD4 counts, BMI and hemoglobin (HB), the AUC of which is 0.831. Han OY
et al[25]constructed a mortality prediction model of hemodialysis patients, of which the AUC is 0.779, risk factors including age,use of
Page 2/16
autologousarteriovenous fistula (AVF), emergency temporary dialysis catheter placement, diabetes, cardiovascular disease,hemoglobin,
and no caregiver.

In conclusion, although the studies above had constructed mortality prediction models of HIV/AIDS patients and hemodialysis patients,
the efficacy was low. More importantly, there were no researches that exploring the mortality risk of HIV/AIDS patients undergoing
hemodialysis, and constructing mortality prediction model. Therefore, this study was aimed to construct a mortality prediction model of
HIV/AIDS patients undergoing hemodialysis, and provide scientific evidence for clinical diagnosis and treatment.

Methods
Participants and Grouping
This study retrospectively screened HIV/AIDS patients receiving hemodialysis in Chengdu Public Health Clinical Medical Center between
June 6, 2017 and June 6,2023, and separated all participants into training set for model construction and validation set for model
verification according to 8:2 ratio, and divided into mortality group and survival group on basis of follow-up result of the patients. The
inclusion criteria were: (1) Patients age over 18 years old. (2) Hemodialysis patients. (3) HIV/AIDS patients. The exclusion criteria were:
(1) Patients with acute kidney injury. (2)Loss to follow-up. (3) Absence of clinical data. This study was approved by the Research Ethics
Board of Chengdu Public Health Clinical Medical Center, and had got the permission of all participants.
Data collection
We collected all clinical data of HIV/AIDS patients of the hemodialysis center in Chengdu Public Health Clinical Medical Center at their
first visit through electronic mdical record system, which included demographic characteristics (age, sex, BMI and education level), past
medical history, vascular access, hemodialysis frequency, predialysis and postdialysis systolic bloodpressures, and laboratory test
results including blood routine examination, liver function test, renal function test,serum electrolyte examination, myocardial enzyme
examination, serum lipid, amino N-terminal brain natriuretic peptide (BNP), fibrinogen to albumin ratio (FAR), neutrophil count to albumin
ratio (NAR), neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), systemic immune
inflammation index (SII), lymphocyte subsets ( CD3 cell count, CD4 cell count, CD8 cell count, CD3 + CD4+%, CD3 + CD8+%, CD4 to CD8
ratio) and HIV-RNA, and we got the follow-up results by telephone or electronic medical record.

Statistical analysis
We use the SPSS software (version 26.0, IBM, Chicago, IL, USA) and R software were used to perform a statistical analysis of the data.
Descriptive variables includes continuous variables, representing as mean ± SD, and categorical variables, representing as n%. And chi-
square test or fisher’s precise test were used to compare difference between groups. We turned continuous variables into categorical
variables for subsequent analysis by best cut-off value acquired by ROC analysis, which correspond to the biggest Youden index. Then
we used ROC analysis to evaluate the prediction efficacy of every single variable, and chose variables whose AUC > 0.6 for subsequent
analysis. We used single factor logistic regression analysis to explore the relationship between variables and mortality. And afterwards,
we used stepwise logistic regression analysis to construct mortality prediction model of HIV/AIDS patients undergoing hemodialysis,
and used "rms" and "mstate" of R software to visualize the logistic regression model, and turned into a nomogram. At last, we used ROC
analysis, calibration curve and decision curve for model evaluation and internal validation.

Results
Patient characteristics
This study screened 166 HIV/AIDS patients undergoing maintenance hemodialysis, 75 mortalities and 91 survivals. And we separated
all patients into training set, including 123 patients (55 mortalities and 68 survivals ), and validation set, including 43 patients (20
mortalities and 23 survivals). As shown in Table 1, there were no significant differences in age, gender, BMI, hemodialysisfrequency,
education level, hypertension, diabetes mellitus, cardiovascular disease, smoking, drinking, vascularaccess, predialysis and postdialysis
systolic bloodpressures, and dialysis age after diagnosis of HIV/AIDS between the training set and validation set (p > 0.05), which meant
the two data sets were mutual independence, and mutual authentication was practicable. In training set, there were differences in age,
BMI, hemodialysis frequency,education level, vascular access, and dialysis age after diagnosis of HIV/AIDS between survival group and
mortality group (p < 0.05). Additionally, There were no differences in gender, hypertension, diabetes mellitus, cardiovascular disease,

Page 3/16
smoking, drinking, predialysis and postdialysis systolic blood pressures between survival group and mortality group (p>0.05). The
specific demographic data were shown in Table 1.

Table 1
Patient characteristics in training set and validation set
Training set Validation set

Variables ALL(n Survival(n Mortality(n P ALL(n Survival(n Mortality(n p P

= 123) = 68) = 55) Value = 43) = 23) = 20) Value Value

Age,year 53.36 48.97 ± 58.71 ± 0.001 53.93 48.48 ± 60.20 ± 0.007 0.819
± 12.91 11.81 12.99 ± 14.42 12.24 14.45

Gender

Male 100 56(82.35%) 44(80.00%) 36 19(82.61%) 17(85.00%)

Female 23 12(17.65%) 11(20.00%) 0.739 7 4(17.39%) 3(15.00%) 0.832 0.723

BMI(kg/cm²) 22.39 23.00 ± 21.64 ± 0.043 21.47 21.73 ± 21.17 ± 0.529 0.104
± 3.69 3.10 4.22 ± 29.24 3.38 2.35

Hemodialysis.frequency

<3 times/week 68 18(26.47%) 6(10.91%) 8 6(26.09%) 2(10.00%)

3 times/week 55 50(73.53%) 49(89.09%) 0.030 35 17(73.91%) 18(90.00%) 0.176 0.897

Education.level

Primary school diploma 64 26(38.24%) 38(69.09%) 25 11(47.83%) 14(70.00%)

High education 59 42(61.76%) 17(30.91%) 0.001 18 12(52.17%) 6(30.00%) 0.142 0.489

Hypertension 72 42(61.76%) 30(54.55%) 0.419 27 17(73.91%) 10(50.00%) 0.106 0.625

Diabetes mellitus 34 21(30.88%) 13(23.64%) 0.372 10 7(30.43%) 3(15.00%) 0.232 0.575

Cardiovascular disease 8 2(2.94%) 6(10.91%) 0.079 5 1(4.35%) 4(20.00%) 0.132 0.222

Smoking 61 31(45.59%) 30(54.55%) 0.323 19 10(43.48%) 9(45.00%) 0.920 0.541

Drinking 51 24(35.29%) 27(49.09%) 0.123 16 9(39.13%) 7(35.00%) 0.780 0.625

Vascular.access

AVF 59 45(66.18%) 14(25.45%) 19 16(69.57%) 3(15.00%)

NCC 41 9(13.24%) 32(58.18%) 19 4(17.39%) 15(75.00%)

TCC 23 14(20.58%) 9(16.37%) 0.000 5 3(13.04%) 2(10.00%) 0.000 0.354

Predialysis systolic 155.94 155.50 ± 156.49 ± 9.920 156.63 153.74 ± 159.95 ± 0.997 0.726
blood ± 21.29 21.77 20.87 ± 20.65 20.23 21.15
pressures(mmHg)

Postdialysis systolic 133.50 133.38 ± 133.64 ± 0.618 132.91 134.78 ± 130.75 ± 0.343 0.830
blood ± 15.90 15.12 16.94 ± 15.25 14.42 16.25
pressures(mmHg)

Dialysis age after 542.76 715.00 ± 329.80 ± 0.000 416.65 648.35 ± 150.20 ± 0.003 0.062
diagnosis of ± 587.99 523.50 ± 751.36 289.32
HIV/AIDS(Day) 590.04 629.91

Laboratory test results in training set

We collected laboratory test results of all participants through Electronic Medical Record System, including 54 variables, which were
white blood cell count (WBC), neutrophil count (NEU), lymphocyte count (LYM), monocyte count (MONO), eosinophil count (EOS),
basophil count (BASO), red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), mean corpuscular hemoglobin contentration
(MCHC), mean corpuscular hemoglobin (MCH), mean corpuscular volume (MCV), platelet count (PLT), mean platelet volume (MPV), c-
reactive protein (CRP), alanine aminotransferase (ALT), aspartate transaminase (AST), total bilirubin (TBIL), alkaline phosphatase (ALP),
Page 4/16
glutamyl transpeptidase (GGT), albumin (ALB), globulin ratio (A/G), blood urea nitrogen (BUN), blood creatinine (Scr), blood uric acid
(UA), serum cystatin (Cys-C), serum ferritin (SF), serum calcium (Ca2+), serum magnesium (Mg2+), serum phosphorus (P), serum sodium
(Na+), serum potassium (K+), serum chlorine (Cl−), carbon dioxide combining power (CO2), lactic dehydrogenase (LDH), hydroxybutyrate
dehydrogenase (HBDH), creatine kinase isoenzyme (CK-MB), triglyceride (TG), total cholesterol (TC), high density lipoprotein (HDLC), low
density lipoprotein (LDLC), amino N-terminal brain natriuretic peptide (BNP), fibrinogen to albumin ratio (FAR), neutrophil count to
albumin ratio (NAR), neutrophil count to lymphocyte count ratio (NLR), platelet count to lymphocyte count ratio (PLR), systemic immune
inflammation index (SII), lymphocyte subset (CD3 cell count, CD4 cell count, CD8 cell count, CD3 + CD4+%, CD3 + CD8+%, CD4 to CD8
ratio) and HIV-RNA.In training set, we screened 11 meaningful variables through t-test or non-parametric test, they were WBC, NEU,
MCHC, CRP, Cys-C, Mg2+, Cl−, CK-MB, NLR, CD4/CD8 ratio, HIV-RNA (p<0.05). The specific results were shown in Table 2.

Page 5/16
 Table 2
Laboratory test results in training set
Variables ALL(n = 123) Survival(n = 68) Mortality(n = 55) p Value

WBC(×109/L) 5.90 ± 2.37 5.41 ± 1.85 6.48 ± 2.78 0.013

NEU(×109/L) 4.63 ± 2.89 4.04 ± 1.72 5.34 ± 3.76 0.013

LYM(×109/L) 0.93 ± 0.45 0.91 ± 0.43 0.95 ± 0.47 0.626

MONO(×109/L) 0.38 ± 0.22 0.38 ± 0.24 0.39 ± 0.20 0.808

EOS(×109/L) 0.15 ± 0.23 0.14 ± 0.14 0.17 ± 0.31 0.142

BASO(×109/L) 0.02 ± 0.02 0.02 ± 0.01 0.02 ± 0.02 0.621

RBC(×1012/L) 2.43 ± 0.67 2.38 ± 0.74 2.49 ± 0.58 0.383

HGB(g/L) 76.06 ± 17.55 75.56 ± 19.37 76.71 ± 15.00 0.723

HCT,% 23.53 ± 5.61 23.21 ± 6.17 23.94 ± 4.82 0.467

MCH,fl 31.47 ± 4.25 30.87 ± 4.41 32.32 ± 3.96 0.078

MCHC(g/L) 323.85 ± 13.95 326.24 ± 13.79 320.89 ± 13.70 0.034

MCV,fl 98.78 ± 10.73 98.82 ± 11.49 98.74 ± 9.83 0.964

PLT(×109/L) 161.00 ± 72.63 168.69 ± 84.03 151.49 ± 54.71 0.193

MPV,fl 10.21 ± 1.46 10.39 ± 1.57 9.99 ± 1.28 0.128

CRP(mg/L) 23.67 ± 41.09 15.44 ± 22.26 34.06 ± 55.11 0.002

ALT(U/L) 22.16 ± 14.77 22.43 ± 13.68 21.83 ± 16.11 0.830

AST(U/L) 25.04 ± 13.60 24.87 ± 13.54 25.26 ± 13.81 0.875

GGT(U/L) 42.39 ± 29.62 42.45 ± 26.70 42.32 ± 33.08 0.584

ALP(U/L) 100.82 ± 51.41 107.66 ± 57.80 91.84 ± 40.37 0.098

ALB(g/L) 32.20 ± 5.37 32.68 ± 3.31 31.61 ± 5.43 0.277

A/G 1.10 ± 0.30 1.13 ± 0.31 1.07 ± 0.29 0.284

TBIL(umol/L) 3.56 ± 2.27 3.55 ± 2.51 3.57 ± 1.95 0.975

BUN(mmol/L) 30.94 ± 11.03 30.45 ± 10.61 31.53 ± 11.61 0.594

Scr(mmol/L) 1095.05 ± 393.78 1092.03 ± 395.26 1098.79 ± 395.56 0.925

UA(mmol/L) 549.88 ± 176.13 533.32 ± 173.84 570.35 ± 178.39 0.249

Cys-C(mg/L) 4.91 ± 1.27 4.70 ± 0.96 5.17 ± 1.54 0.041

SF(ng/ml) 278.88 ± 201.89 274.41 ± 200.34 284.28 ± 205.58 0.855

Ca(mmol/L) 1.94 ± 0.25 1.97 ± 0.26 1.91 ± 0.23 0.210

Mg2+(mmol/L) 1.00 ± 0.21 0.96 ± 0.21 1.06 ± 0.20 0.008

P(mmol/L) 2.21 ± 0.71 2.17 ± 0.73 2.27 ± 0.70 0.454

CO2(mmol/L) 15.83 ± 5.36 16.49 ± 5.23 15.01 ± 5.46 0.129

Na+(mmol/L) 138.47 ± 6.10 139.20 ± 6.69 137.55 ± 5.21 0.127

K+(mmol/L) 4.56 ± 0.90 4.48 ± 0.90 4.66 ± 0.89 0.260

Page 6/16
 Variables ALL(n = 123) Survival(n = 68) Mortality(n = 55) p Value

Cl−(mmol/L) 104.39 ± 7.16 105.72 ± 7.25 102.74 ± 6.75 0.021

LDH(U/L) 286.63 ± 106.75 282.09 ± 111.71 292.76 ± 100.51 0.593

HBDH(U/L) 220.23 ± 90.60 215.71 ± 76.01 225.89 ± 106.59 0.563

CK-MB(U/L) 21.94 ± 28.33 22.31 ± 35.51 21.47 ± 15.37 0.033

TC(mmol/L) 4.14 ± 1.07 4.14 ± 1.19 4.15 ± 0.94 0.984

TG(mmol/L) 1.86 ± 0.76 1.83 ± 0.71 1.89 ± 0.82 0.704

HDLC(mmol/L) 1.08 ± 0.33 1.04 ± 0.30 1.12 ± 0.35 0.218

LDLC(mmol/L) 2.34 ± 0.75 2.34 ± 0.72 2.35 ± 0.80 0.972

BNP(pg/ml) 16739.80 ± 14125.87 17379.41 ± 14298.53 15886.98 ± 13996.70 0.659

FAR 0.14 ± 0.05 0.14 ± 0.05 0.13 ± 0.05 0.298

NAR 0.14 ± 0.07 0.14 ± 0.07 0.14 ± 0.07 0.982

NLR 5.72 ± 4.37 4.72 ± 2.49 6.86 ± 5.64 0.008

PLR 203.19 ± 137.51 209.89 ± 138.69 194.85 ± 136.88 0.555

SII 847.34 ± 684.27 825.68 ± 638.84 875.08 ± 744.08 0.709

CD3(cells/ul) 672.46 ± 347.56 678.83 ± 363.18 663.96 ± 329.43 0.827

CD4(cells/ul) 242.52 ± 149.84 249.11 ± 150.98 234.12 ± 149.44 0.595

CD8(cells/ul) 432.86 ± 248.30 400.04 ± 237.11 473.36 ± 258.23 0.136

CD3 + CD4+,% 26.36 ± 11.37 27.60 ± 12.41 24.74 ± 9.74 0.187

CD3 + CD8+,% 47.26 ± 13.42 46.26 ± 14.13 48.58 ± 12.44 0.355

CD4/CD8 0.78 ± 0.69 0.85 ± 0.69 0.69 ± 0.68 0.033

HIV-RNA 24345.17 ± 74004.75 6183.73 ± 25699.47 46423.78 ± 102741.60 0.014

Aberation: WBC white blood count, NEU neutrophil count, LYM lymphocyte count, MONO monocyte count, EOS eosinophil coun, BASO
basophil count, RBC red blood cell count, HGB hemoglobin, HCT hematocrit, MCHC mean corpuscular hemoglobin contentration, MCH
mean corpuscular hemoglobin, MCV mean corpuscular volume, PLT platelet count, MPV mean platelet volume, CRP c-reactive protein,
ALT alanine aminotransferase, AST aspartate transaminase, TBIL total bilirubin, ALP alkaline phosphatase, GGT glutamyl
transpeptidase, ALB albumin, A/G globulin ratio, BUN blood urea nitrogen, Scr blood creatinine, UA blood uric acid, Cys-C serum cystatin,
SF serum ferritin, Ca2+ serum calcium, Mg2+ serum magnesium, P serum phosphorus, Na+ serum sodium, K+ serum potassium, Cl−
serum chlorine, CO2 carbon dioxide combining power, LDH lactic dehydrogenase, HBDH hydroxybutyrate dehydrogenase, CK-MB creatine
kinase isoenzyme, TG triglyceride, TC total cholesterol, HDLC high density lipoprotein, LDLC low density lipoprotein, BNP amino N-
terminal brain natriuretic peptide, FAR fibrinogen to albumin ratio, NAR neutrophil count to albumin ratio, NLR neutrophil count to
lymphocyte count ratio, PLR platelet count to lymphocyte count ratio, SII systemic immune inflammation index, lymphocyte subset (CD3
cell count, CD4 cell count, CD8 cell count, CD3 + CD4+, CD3 + CD8+, CD4/CD8 ratio).
ROC analysis in training set
We screened 17 meaningful variables through t-test, non-parametric test, chi-square test and fisher's precise test, and chose variables
that AUC > 0.6 by ROC analysis for following analysis, including age, BMI, education level, vascular access, dialysis age after diagnosis
of HIV/AIDS, WBC, NEU, MCHC, CRP, CK-MB, Mg2+, Cl−, NLR, CD4/CD8 ratio and HIV-RNA, the specific AUC results were showed in Table
3.

Page 7/16
 Table 3
ROC analysis in training set
Vavriables AUC 95%CI p value

Age 0.680 0.583–0.777 0.001

BMI 0.668 0.570–0.765 0.001

Education level 0.654 0.557–0.752 0.003

Hemodialysis.frequency 0.578 0.477–0.679 0.139

Vascular.access 0.655 0.555–0.754 0.003

Dialysis age after diagnosis of HIV/AIDS 0.732 0.640–0.824 0.000

WBC 0.614 0.514–0.715 0.029

NEU 0.615 0.514–0.716 0.028

MCHC 0.625 0.525–0.724 0.018

CRP 0.680 0.584–0.776 0.001

Mg 0.653 0.555–0.751 0.004

Cl 0.622 0.522–0.723 0.020

Cysc 0.597 0.495-0.700 0.064

CK-MB 0.626 0.526–0.727 0.016

NLR 0.627 0.527–0.727 0.016

CD4/CD8 0.648 0.549–0.748 0.005

HIV-RNA 0.654 0.554–0.754 0.003

Aberation: WBC white blood count, NEU neutrophil count, MCHC mean corpuscular hemoglobin contentration, CRP c-reactive protein,
Mg2+ serum magnesium, Cl− serum chlorine, Cys-C serum cystatin, CK-MB creatine kinase isoenzyme, NLR neutrophil count to
lymphocyte count ratio.

Construction of a predictive nomogram model

We screened 6 independent risk factors fromabove 15 variables through stepwise logistic regression, including education level
[OR(95%CI): 3.754 (1.247–11.300),p = 0.019], dialysis age after diagnosis of HIV/AIDS [OR(95%CI):0.050 (0.013–0.187),p = 0.000],
creatine kinase isoenzyme (CK-MB)[OR(95%CI): 7.666 (2.237–26.271),p = 0.001],neutrophil and lymphocyte counts ratio (NLR)
[OR(95%CI):3.864 (1.266–11.795),p = 0.018], magnesium (Mg2+)[OR(95%CI): 4.883 (1.479–16.118),p = 0.009],HIV-RNA[OR(95%CI):
17.616 (3.797–81.721),p = 0.000]. It showed that education level, CK-MB, NLR, Mg2+and HIV-RNAwere mortality independent risk factors
of HIV/AIDS patients undergoing hemodialysis, while dialysis age after diagnosis of HIV/AIDS were independent protective factor of
HIV/AIDS patients undergoing hemodialysis. We constructed a mortality risk factor prediction model of HIV/AIDS patients undergoing
hemodialysis on basis of these independent risk factor above, and turned it into a nomogram (Fig. 1A).

Assessment and validation of model

In training set, the AUC of the prediction model we constructed was 0.921(95%CI 0.883 ~ 0.968)(Fig. 1B), and calibration curve showed
that the calibration value between the actual model and prediction model had a good performance(Fig. 2A), and decision curve showed
that the model had good practicability in clinical(Fig. 2B). Besides, we got the same result in validation set, which AUC was 0.859(95%CI
0.748 ~ 0.969) (Fig. 3).

Discussion
According to the statistics, about 30% HIV/AIDS patients had abnormal renal function, and of which about 3.5–48.5%[8]patients
developed into ESRD eventually[9], and had increasingtrend globally year by year[10, 11]. Patients with HIV/AIDS and ESRD had a very

Page 8/16
poor medical prognosis owing tothe attack of both human immunodeficiency virus and uremic toxin. Therefore, in order to improve life
quality of these patients, this study was aimed to discuss the mortality risk factors of HIV/AIDS patients undergoing hemodialysis, and
construct a mortality prediction nomogram model, expecting to help clinicians have a early identification of mortality risk of HIV/AIDS
patients with ESRD, and improve the life quality and lower the mortality rate of these patients by early intervention.

There were no researchers constructing mortality prediction model of HIV/AIDS patients undergoing hemodialysis as so far, and the
present prediction model constructed by this study could fill up thevacancy just in time, and the prediction model had a good
performance in predicting mortality. Additionally, calibration curve and decision curve both showed the model had widely applied
prospect in clinical practice. There were 6 variables included in this model, they wereeducation level, dialysis age after diagnosis of
HIV/AIDS, neutrophil count to lymphocyte count ratio, CK-MB, Mg2+, HIV-RNA. According to the result of stepwise logistic regression,
HIV/AIDS patients undergoing hemodialysis with lower education level, shorter dialysis age after diagnosis of HIV/AIDS, higher
concentration of Mg2+, NLR, HIV-RNA had a higher mortality risk.

In this study, we found that HIV/AIDS patients undergoing hemodialysis with lower education level had higher mortality risk, and the
correlation between education level and life quality is positive in patients undergoing maintenance hemodialysis or with HIV infection in
some researches[26, 27]. Patients with higher education level had a significant higher life quality than those who had lower education or
are illiteracy, which was roughly consistent with this present study. We thought that the reason may attribute to patients with higher
education level were more likely have a stable job and stable income in modern society, so they may have a higher socioeconomic status
(SES). Some studies showed that the dietary habit of patients with lower socioeconomic status was usually unhealthy, such as food
with lot of fat, sugar or salt, which may have a bad impact on human body, especially patients with chronic kidney disease[28]. In
addition, patients with chronic disease usually need strong financial support, and patients with higher socioeconomic status may have
the ability to earn more money and choose a healthier life style, so that they could lower mortality risk accordingly[29]. On the other hand,
patients with higher education level had good self-management and self-monitoring, and they were likely to visit doctors on regular time,
in order to avoid aggravation of diseases[30].

The prediction model we constructed in this study indicated that CK-MB, NLR and Mg2+were the independent mortality risk
factors.Creatine kinase isoenzyme(CK-MB) mainly exits in myocardium, and it's a more sensitive indicator in diagnosing some
myocardial damage diseases such as myocardial infarction[31], and as shown in previous studies[32], more than 50 percent hemodialysis
patients died of cardiovascular disease, which was one of the most common cause of death of hemodialysis patients. And moreover,
the incidence of cardiovascular disease of HIV/AIDS patients increased year by year [33], which agreed with this study approximately.
Therefore, we should pay attention to patients with high value of CK-MB due to high rates of cardiovascular accident, and we
shouldcomplete the examination about cardiovascular diseases as much as possible and unite cardiologist consultation, in order to find
out potential cardiac diseases earlier, to diagnose and treat earlier, and add some cardiac drugs properlyto protect myocardium at the
same time.

The body of patients undergoing maintenance hemodialysis was generally in a state of micro inflammation[34], and patients with
chronic kidney disease usually had high inflammatory mediators, including c-reactive protein (CRP), tumor necrosis factor -α(TNF - α),
interleukin − 6 (IL − 6) and so on[35], and more importantly, chronic inflammation was the essential factor for progression of chronic
kidney disease[36]. However, normal patients couldn’t accept those examinations above because of the high price, as a result, neutrophil
to lymphocyte ratio (NLR) became a new inflammatory marker with easy operation and low price, and could predict the mortality rate of
cardiovascular disease[37]. Higher level of NLR was the predictor of cardiovascular mortality of hemodialysis patients, which was found
in a retrospective study[38]. Patients with HIV infected usually had lower immunity and had high rate of different kinds of opportunistic
infections, which lead to high level of all kinds of inflammatory markers including NLR. Meanwhile, NLR was the predictor of HIV/AIDS
patients getting cardiovascular disease[39], and patients with higher level of NLR had higher mortality rate, which was roughly in line with
the present study. However, researches about the relation between NLR and HIV/AIDS patients were still very few currently. In this study,
clinicians should be on the alert of HIV/AIDS patients with higher level of NLR, who had high possibility in having cardiovascular
disease, and complete examinations related as early as possible and review relevant indicators regularly to reduce the incidence of
cardiovascular disease.

Magnesium (Mg2+) was a positive ion ranking fourth after potassium, sodium and calcium in human body, which had important
function in regulating neuromuscular excitability, bone metabolism and cardiovascular function and so on[40]. As so far, studies about
relationship between Mg2+and HIV infection had a very small quantity, and some previous studies showed that higher or lower value of

Page 9/16
Mg2+could both improve mortality risk of hemodialysis patients[41–42]. In this research, the value of Mg2+ had positive correlation with
mortality rate of HIV/AIDS patients undergoing hemodialysis, which probably on account of hypermagnesemia had inhibited effect on
rhythmic cardiac cell, clinical manifestation of which were bradycardia and all kinds of conduction block and arrhythmia, hypotension,
and cause hypoexcitability of respiratory center and respiratory paralysis in serious situation[43]. Therefore, for those hemodialysis
patients with hypermagnesemia, on one hand, we could complete some related examinations to find out if there were systems damaged,
and then get patients symptomatic treatment; on the other hand, we should adjust dialysis mode to improve adequacy of dialysis.

Moreover, this study showed that dialysis age after diagnosis of HIV/AIDS was the independent protection factor. Patients with shorter
dialysis age had higher mortality rate, which was approximately accordance with the result of research conducted by Josefin Santosa,
etc[44]. The reason probably attributed to some patients, especially with higher age, combined with multiple underlying diseases, and
with unstable condition when first starting hemodialysis, and thereby died of some other causes in early stage of hemodialysis, such as
severe infection, heart failure, fatal arrhythmia and so on, and consequently shorten dialysis age. Nowadays, with continuously
strengthen of dialysis technology and widely application of high-flux dialysis, the survival time of dialysis patients gradually get
longer[45].

It is worth mentioning that the research objects this study chose were HIV/AIDS patients, and the nomogram we constructed showed
that HIV-RNA was an independent risk factor. HIV-RNA represents HIV viral load, the higher of the value of which, the more active the
virus replication. Antiretroviral treatment (ART) could reduce HIV-RNA effectively now, and help HIV/AIDS patients rise CD4 count, so that
to reduce incidence of opportunistic infection[46, 47]. Therefore, it is essential for HIV/AIDS patients undergoing maintenance
hemodialysis reduce mortality rate by regular follow-up in infectious diseases department, monitoring lymphocyte subset and HIV-RNA
to evaluate the effect of ART[48].

In recent years, with clinical data increasing continuously, more and more researchers start to construct machine learning models, in
order to conduct different kinds of risk prediction. Mei ZW, et al constructed a nomogram to predict hyperkalemia in patients with
hemodialysis[49], for the purpose of helping clinicians to early identify the risk of patients undergoing hemodialysis occurring
hyperkalemia, and to reduce the rate of hemodialysis patients dying of hyperkalemia. Grace Fet, et al constructed an automated lung
ultrasound image assessment using artifcial intelligence to identify fluid overload in dialysis patients[50], in order to provide a scientific
support for clinicians evaluating the liquid load of hemodialysis patients, so that they can help patients achieve higher dialysis quality.
Ricardo Peralta, et al[51]established a machine learning model predicting arteriovenous fistula failure, which had the ability to help
vascular access doctors predict the rate of hemodialysis patients encountering with arteriovenous fistula failure in early stage, as a
result they could conduct an alternative plan in advance to solve the problem of vascular access of patients. However, there was no
researches predicting the mortality risk of HIV/AIDS patients undergoing hemodialysis at present, and all the studies currently only
focused on the mortality risk of HIV/AIDS patients[24] and hemodialysis patients[25], the clinical application value of which were limited.
As a consequence, we constructed a mortality prediction model of HIV/AIDS patients undergoing hemodialysis in this study. The
indicators in this model not only included hemodialysis related data, such as dialysis age after diagnosis of HIV/AIDS, but also included
HIV/AIDS related data, such as HIV-RNA. Thereby, this prediction model can react the live condition of HIV/AIDS patients undergoing
hemodialysis more real, and had wider applicability. On the second hand, the AUC of the prediction model was 0.921, which represented
a good performance in prediction. In addition, the prediction model not only applied in economically underdeveloped areas, but also in
primary hospitals because of the easily obtained of the clinical data in this prediction model.

To sum up, the results of the present study indicated that education level, dialysis age after diagnosis of HIV/AIDS, neutrophil count to
lymphocyte count ratio, CK-MB, Mg2+and HIV-RNA were independent risk factors of HIV/AIDS patients undergoing hemodialysis. The
mortality prediction model constructed on the basis of the above factors had a good predictive performance, which would help clinicians
identify patients of high risk and intervene earlier, so that it could reduce the mortality rate and lengthen the survival time of these
patients. However, there were insufficiency in this study, first of all, the result of this research came from an only hemodialysis center, so
the population representativeness of the study was limited. Second, the sample capacity of this research was not big enough, and we
should prudently promote the prediction model to other hospitals. We could conduct a multi-center study with bigger sample size
afterwards, to make the research result more reliability. At last but not least, the clinical data we collected in this study were from the
time that patients first started hemodialysis, and we did not explore whether these inspection results changed after treatment and
generated different results, therefore, the result of this study was limited to a certain degree. In the future, we could conduct a research
on different conclusion when some inspection results changed after some treatment measures.

Page 10/16
Declarations
Ethics approval and consent to participate

This study was approved by the Research Ethics Board of Chengdu Public Health Clinical Medical Center, the ethical approval number is
YJ-K2023-73-02, and informed consent was obtained from all study participants.

Consent for publication

Informed consent for publication of identifiable information/ images in an open access journal was obtained from all study participants.

Availability of data and materials

The datasets generated and analyzed during the current study are not publicly available due to privacy and ethical restrictions but are
available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Funding

This study was supported by academy level project of Public Health Clinical Center of Chengdu.

Authors' contributions

Zhu-rui Xian performed the data analysis and writing the main manuscript; Xiao-fei Song prepared tables 1-3; Yong-fu Wang prepared
figures 1-3; Ting-ting Yang performed the validation; Nan Mao ensured quality of the paper. All authors reviewed the manuscript.

Authors’ details

1Chengdu Medical College, Public Health Clinical Center of Chengdu, 610000, China. 2Public Health Clinical Center of Chengdu, 610000,
China. 3Public Health Clinical Center of Chengdu, 610000, China. 4Public Health Clinical Center of Chengdu, 610000, China. 1*Chengdu
Medical College, First Affiliated Hospital of Chengdu Medical College, 610000, China.

References
1. Provenzano R, Kupin W, Santiago GC. Renal Involvement in the Acquired Immunodeficiency Syndrome:Presentation, Clinical Course,
and Therapy. Henry FordHospital Med J, 1987(3):38–41.
2. Simon V, Ho DD, Abdool Karim Q. HIV/AIDS epidemiology, pathogenesis, prevention, and treatment. Lancet. 2006;368(9534):489–
504.
3. Wandeler G, Johnson LF, Egger M. Trends in life expectancy of HIV-positive adults on antiretroviral therapy across the globe. Curr
Opin HIV AIDS. 2016;11(5):492–500.
4. Vigil KJ, Salazar L, Hasbun R. Community-Acquired Meningitis in HIV-Infected Patients in the United States. AIDS Patient Care STDs.
2018;32(2):42–7.
5. Pereira-Díaz E, Moreno-Verdejo F, de la Horra C et al. Changing Trends in the Epidemiology and Risk Factors of Pneumocystis
Pneumonia in Spain. Front Public Health. 2019;7.
6. Méndez-Samperio P. Diagnosis of Tuberculosis in HIV Co‐infected Individuals: Current Status, Challenges and Opportunities for the
Future. Scand J Immunol. 2017;86(2):76–82.
7. Yarchoan R, Longo DL, Uldrick TS. HIV-Associated Cancers and Related Diseases. N Engl J Med. 2018;378(11):1029–41.
8. Samir K, Gupta,Joseph A, Eustace et al. Guidelines for the Management of Chronic Kidney Disease in HIV-Infected Patients:
Recommendations of the HIV Medicine Association of the Infectious Diseases Society of America. Clin Infect Dis, 2005(40): 1559–
85.
9. Naicker S, Rahmanian S, Kopp JB. HIV and chronic kidney disease. Clinical Nephrology. 2015;83 (2015)(S1):32–38.

Page 11/16
10. Heron JE, Bagnis CI, Gracey DM. Contemporary issues and new challenges in chronic kidney disease amongst people living with
HIV. AIDS Res Therapy. 2020;17(1).
11. Abraham AG, Althoff KN, Jing Y, et al. End-Stage Renal Disease Among HIV-Infected Adults in North America. Clin Infect Dis.
2014;60(6):941–9.
12. Ryom L, Kirk O, Lundgren JD, et al. Advanced chronic kidney disease, end-stage renal disease and renal death among HIV‐positive
individuals in Europe. HIV Med. 2013;14(8):503–8.
13. Atta MG, Fine DM, Kirk GD, et al. Survival during Renal Replacement Therapy among African Americans Infected with HIV Type 1 in
Urban Baltimore, Maryland. Clin Infect Dis. 2007;45(12):1625–32.
14. Luo B, Sun J, Cai R et al. Spectrum of Opportunistic Infections and Risk Factors for In-Hospital Mortality of Admitted AIDS Patients
in Shanghai. Medicine. 2016;95(21).
15. Penagos Gaviria S, Zapata N, Villa P, et al. HIV/AIDS infection in critical care: epidemiological profile and risk factors for mortality in
a Colombian cohort. J Infect Developing Ctries. 2023;17(01):102–10.
16. Petoumenos K, Garriga C, García de Olalla P et al. Mortality, Causes of Death and Associated Factors Relate to a Large HIV
Population-Based Cohort. PLoS ONE. 2015;10(12).
17. Kabuba N, Menon JA, Franklin DR, et al. Effect of age and level of education on neurocognitive impairment in HIV positive Zambian
adults. Neuropsychology. 2018;32(5):519–28.
18. Cao Z, Li J, Chen H et al. Effects of HIV-1 genotype on baseline CD4 + cell count and mortality before and after antiretroviral therapy.
Sci Rep. 2020;10(1).
19. Lang J, Scherzer R, Weekley CC, et al. Serum albumin and short-term risk for mortality and cardiovascular disease among HIV-
infected veterans. Aids. 2013;27(8):1339–43.
20. Victoriano AFB, Imai K, Okamoto T. Interaction between Endogenous Bacterial Flora and Latent HIV Infection. Clin Vaccine Immunol.
2013;20(6):773–9.
21. Xie B, Huang W, Hu Y et al. Anemia and opportunistic infections in hospitalized people living with HIV: a retrospective study. BMC
Infect Dis. 2022;22(1).
22. Samir K, Gupta,Joseph A, Eustace et al. Guidelines for the Management of Chronic Kidney Disease in HIV-Infected Patients:
Recommendations of the HIV Medicine Association of the Infectious Diseases Society of America. Clin Infect Dis, 2005(40): 1559–
85.
23. Chen C, Zhang J, Zhou Z, et al. Impact of serum albumin level and variability on short-term cardiovascular-related and all-cause
mortality in patients on maintenance hemodialysis. Medicine. 2021;100:43.
24. Jiang F, Xu Y, Liu L et al. Construction and validation of a prognostic nomogram for predicting the survival of HIV/AIDS adults who
received antiretroviral therapy: a cohort between 2003 and 2019 in Nanjing. BMC Public Health. 2022;22(1).
25. Ouyang H, Shi Q, Zhu J, et al. Nomogram for predicting 1-, 5-, and 10-year survival in hemodialysis (HD) patients: a single center
retrospective study. Ren Fail. 2021;43(1):1508–19.
26. Floria I, Kontele I, Grammatikopoulou MG et al. Quality of Life of Hemodialysis Patients in Greece: Associations with Socio-
Economic, Anthropometric and Nutritional Factors. Int J Environ Res Public Health. 2022;19(22).
27. Melo MCd, Mesquita FC, Barros MBA et al. Sobrevida de pacientes com aids e associação com escolaridade e raça/cor da pele no
Sul e Sudeste do Brasil: estudo de coorte, 1998–1999*. Epidemiologia e Serviços de Saúde. 2019;28(1).
28. Gutiérrez OM. Contextual Poverty, Nutrition, and Chronic Kidney Disease. Adv Chronic Kidney Dis. 2015;22(1):31–8.
29. Amini M, Moradinazar M, Rajati F et al. Socioeconomic inequalities in prevalence, awareness, treatment and control of hypertension:
evidence from the PERSIAN cohort study. BMC Public Health. 2022;22(1).
30. Shahar S, Lau H, Puteh SEW et al. Health, access and nutritional issues among low-income population in Malaysia: introductory
note. BMC Public Health. 2019;19(S4).
31. Serhan OZCAN, Serhat EMEKSIZ, Oktay PERK, et al. Severe coronavirus disease pneumonia in pediatric patients in a referral
hospital. J Trop Pediatr. 2021;37(3):3.
32. Raggi P, Boulay A, Chasan-Taber S, et al. Cardiac calcification in adult hemodialysis patients. J Am Coll Cardiol. 2002;39(4):695–
701.
33. Freiberg MS, Chang C-CH, Kuller LH et al. HIV Infection and the Risk of Acute Myocardial Infarction. JAMA Intern Med. 2013;173(8).

Page 12/16
34. Cobo G, Lindholm B, Stenvinkel P. Chronic inflammation in end-stage renal disease and dialysis. Nephrol Dialysis Transplantation.
2018;33(suppl3):iii35–iii40.
35. Abdulkhaleq LA, Assi MA, Abdullah R, et al. The crucial roles of inflammatory mediators in inflammation: A review. Veterinary World.
2018;11(5):627–35.
36. Graterol Torres F, Molina M, Soler-Majoral J et al. Evolving Concepts on Inflammatory Biomarkers and Malnutrition in Chronic Kidney
Disease. Nutrients. 2022;14(20).
37. Buonacera A, Stancanelli B, Colaci M et al. Neutrophil to Lymphocyte Ratio: An Emerging Marker of the Relationships between the
Immune System and Diseases. Int J Mol Sci. 2022;23(7).
38. Li H, Lu X, Xiong R, et al. High Neutrophil-to-Lymphocyte Ratio Predicts Cardiovascular Mortality in Chronic Hemodialysis Patients.
Mediat Inflamm. 2017;2017:1–8.
39. Lionetti V, Quiros-Roldan E, Raffetti E et al. Neutrophil to Lymphocyte Ratio and Cardiovascular Disease Incidence in HIV-Infected
Patients: A Population-Based Cohort Study. PLoS ONE. 2016;11(5).
40. Piuri G, Zocchi M, Della Porta M et al. Magnesium in Obesity, Metabolic Syndrome, and Type 2 Diabetes. Nutrients. 2021;13(2).
41. Sakaguchi Y, Fujii N, Shoji T, et al. Hypomagnesemia is a significant predictor of cardiovascular and non-cardiovascular mortality in
patients undergoing hemodialysis. Kidney Int. 2014;85(1):174–81.
42. Cirik MÖ, Kilinç M, Doğanay GE, et al. The relationship between magnesium levels and mortality in the respiratory intensive care unit.
Medicine. 2020;99:52.
43. Tan L, Xu Q, Li C et al. High-Normal Serum Magnesium and Hypermagnesemia Are Associated With Increased 30-Day In-Hospital
Mortality: A Retrospective Cohort Study. Front Cardiovasc Med. 2021;8.
44. Santos J, Oliveira P, Malheiro J, et al. Predicting 6-Month Mortality in Incident Elderly Dialysis Patients: A Simple Prognostic Score.
Kidney Blood Press Res. 2020;45(1):38–50.
45. Vilar E, Fry AC, Wellsted D, et al. Long-Term Outcomes in Online Hemodiafiltration and High-Flux Hemodialysis. Clin J Am Soc
Nephrol. 2009;4(12):1944–53.
46. Howard RR, Fasano CS, Frey L, et al. Reference intervals of CD3, CD4, CD8, CD4/CD8, and absolute CD4 values in asian and non-
asian populations. Cytometry. 1996;26(3):231–2.
47. Hadrup SR, Strindhall J, Køllgaard T, et al. Longitudinal Studies of Clonally Expanded CD8 T Cells Reveal a Repertoire Shrinkage
Predicting Mortality and an Increased Number of Dysfunctional Cytomegalovirus-Specific T Cells in the Very Elderly. J Immunol.
2006;176(4):2645–53.
48. Hileman CO, Funderburg NT, Inflammation. Immune Activation, and Antiretroviral Therapy in HIV. Curr HIV/AIDS Rep. 2017;14(3):93–
100.
49. Mei Z, Chen J, Chen P et al. A nomogram to predict hyperkalemia in patients with hemodialysis: a retrospective cohort study. BMC
Nephrol. 2022;23(1).
50. Tan GFL, Du T, Liu JS et al. Automated lung ultrasound image assessment using artificial intelligence to identify fluid overload in
dialysis patients. BMC Nephrol. 2022;23(1).
51. Peralta R, Garbelli M, Bellocchio F et al. Development and Validation of a Machine Learning Model Predicting Arteriovenous Fistula
Failure in a Large Network of Dialysis Clinics. Int J Environ Res Public Health. 2021;18(23).

Figures

Page 13/16
Figure 1

The construction of mortality prediction model of HIV/AIDS patients undergoing hemodialysis in training set. aThe nomogram. b The
ROC curve

Page 14/16
Figure 2

The evaluation of the mortality prediction model of HIV/AIDS patients undergoing hemodialysis in training set. a The calibration curve. b
The decision curve

Page 15/16
Figure 3

The validation of mortality prediction model of HIV/AIDS patients undergoing hemodialysis. a The ROC curve. b The calibration curve. c
The decision curve

Page 16/16