Development and validation of a nomogram model for predicting negative/indeterminate HIV-1 antibody confirmation: a strategy to streamline the diagnostic pathway

Abstract

Background

The standard HIV diagnostic algorithm comprises sequential serological screening, confirmatory antibody testing, and nucleic acid confirmation when necessary. This multi-step process often causes diagnostic delays, particularly in cases with indeterminate or discordant results. Developing a Nomogram-based predictive tool to identify individuals at high risk of anti-HIV-1 negative or inconclusive antibody confirmation could significantly optimize testing efficiency.

Methods

A total of 2,468 individuals with HIV reactive screening results were randomly divided into a training set (70%) and a validation set (30%). In the training set, the risk factors of anti-HIV-1 negative/inconclusive results were identified by univariate and multivariate logistic regression models to construct a predictive model, and its internal stability was robustly assessed via 1,000-resample bootstrap validation. In the validation set, the Hosmer-Lemeshow test and calibration curves were used to evaluate the consistency of the model, while the receiver operating characteristic (ROC) curve was utilized to assess its discriminative ability.

Results

Multivariate analysis identified Gender, Education, voluntary counseling and testing (VCT), and Retesting results as independent predictors. Internal bootstrap validation confirmed excellent model stability (optimism-corrected C-index = 0.862). Furthermore, the model demonstrated good predictive performance with high discrimination ability (area under the ROC curve [AUC] = 0.874, 95% CI: 0.826–0.922) and calibration accuracy (Hosmer-Lemeshow test, P = 0.098) in the validation set.

Conclusions

The nomogram accurately identifies individuals at high risk of negative/inconclusive antibody confirmation, enabling a streamlined two-step testing pathway that bypasses Western blot(WB) for selected cases. This approach reduces diagnostic delay and resource utilization while maintaining accuracy, offering a practical strategy for optimizing HIV testing services in diverse settings.

Trial registration

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12911-026-03546-6.

Background

Human immunodeficiency virus (HIV) infection remains one of the most pressing global public health challenges [1]. The World Health Organization (WHO) indicated that HIV has caused approximately 113 million infections and 40.4 million deaths worldwide [2]. Global epidemiological data from 2024 estimate 1.3 million new HIV infections and approximately 630,000 AIDS-related deaths, with sub-Saharan Africa accounting for nearly half of all new cases [3]. Beyond the direct health implications, people living with HIV (PLHIV) face a significant economic burden associated with lifelong treatment and monitoring, as well as psychological distress including chronic anxiety, depression, and self-stigmatization resulting from discrimination. Consequently, national health systems allocate substantial resources to address the epidemic, which places a significant strain on public finances [4–6]. Central to addressing these multifaceted challenges are early diagnosis and prompt initiation of ART. Highly active antiretroviral therapy (HAART) has been proven to reduce the risk of viral transmission by more than 96% among infected individuals [7]. However, progress toward the WHO’s “95-95-95” targets for 2030 remains uneven—globally, only 87% of PLHIV know their status, with even greater disparities among key populations [3]. This gap underscores the critical importance of optimizing HIV testing systems to advance global treatment targets and maximize the benefits of early diagnosis and treatment.

The diagnosis of HIV infection follows a multi-step algorithmic approach, with specific strategies varying across countries and regions depending on resource availability and testing infrastructure. The initial screening step employs highly sensitive assays such as enzyme-linked immunosorbent assay (ELISA) or chemiluminescent immunoassays (CLIAs). In high-income settings, fourth-generation antigen/antibody combination tests are commonly used, whereas rapid tests (RTs) are preferred in primary care and resource-limited areas due to their operational simplicity and quick turnaround. Reactive specimens proceed to the confirmatory stage. In the United States, platforms such as Geenius have been adopted to facilitate type-specific differentiation between HIV-1 and HIV-2 [8], while many developing countries—including China—still rely heavily on Western blot (WB) as the primary serological confirmation method [9]. However, WB is associated with a considerable rate of indeterminate results and cannot exclude infections during the serological window period [9, 10]. In such cases, nucleic acid testing (NAT) is necessary to resolve ambiguous serological findings. NAT serves as an essential complementary assay, capable of clarifying inconclusive WB results or negative WB results in high-risk individuals (those with recent exposure or symptoms of acute retroviral syndrome) [11]. It also enables viral load quantification, which is critical for monitoring ART [12, 13]. Despite its high accuracy, NAT remains technically complex, costly, and difficult to scale in resource-constrained settings, limiting its broad implementation in public health programs.

While existing HIV testing algorithms play a crucial role in clinical practice, their effectiveness is constrained by the inherent limitations of confirmatory assays, particularly indeterminate results and false negatives [14]. These uncertain outcomes frequently occur during the seroconversion window, due to nonspecific reactions, underlying autoimmune disorders, or prior HIV exposure without development of typical antibody responses [15, 16]. Current guidelines recommend follow-up HIV-1 nucleic acid testing for definitive clarification in such cases [13]. However, this supplementary step prolongs the time to diagnosis, delays the initiation of ART, and exacerbates both the economic burden on healthcare systems and the psychological distress experienced by patients.

The ability to early identify screening-positive individuals who are likely to yield negative or indeterminate confirmatory results would significantly optimize the current diagnostic pathway. In response, this study aims to develop and validate a nomogram model based on readily available clinical and epidemiological variables to identify individuals at high risk of negative or indeterminate confirmatory outcomes. By applying this model, such patients could bypass the antibody confirmation step and proceed directly to nucleic acid testing, thereby streamlining the diagnostic process from a three-step to a two-step algorithm. This optimized approach enhances operational efficiency while maintaining diagnostic accuracy. Furthermore, this study provides a theoretical foundation for the continued refinement of HIV testing strategies.

Materials and methods

Study setting and population

This retrospective study was conducted at the HIV Confirmatory Laboratory of the Second Affiliated Hospital of Chengdu Medical College (Nuclear Industry 416 Hospital), a comprehensive tertiary care center in Chengdu, Southwestern China. Serving as the regional confirmatory hub for nine medical institutions, our laboratory enforced strict, standardized data reporting protocols. Importantly, all demographic and sequential testing records were exclusively captured utilizing a dedicated regional HIV case management system. Because complete data entry was a structural prerequisite for processing any screening or confirmatory order within this specialized system, no missing data were observed for the variables included in the final analysis. This inherent data integrity allowed us to naturally obtain a complete cohort without post-hoc exclusions for missing values, thereby eliminating the selection bias typically associated with complete-case analyses.

The study cohort consecutively included all individuals with reactive initial HIV screening results whose samples were processed between June 2019 and April 2024. To establish a rigorous analytical cohort, two pre-defined exclusion criteria were applied: (1) patients under 18 years of age, due to ethical considerations and distinct pediatric diagnostic algorithms; and (2) patients with repeated screening records within a 4-week period who exhibited inconsistent final confirmatory outcomes, an exclusion implemented to prevent repeated-measures bias and avoid diagnostic ambiguity. Ultimately, 2,468 eligible adult participants were enrolled and randomly allocated into a training cohort (70%, n = 1,727) and an internal validation cohort (30%, n = 741).

Laboratory procedures and quality assurance

Initial HIV screening was conducted in state-certified HIV screening laboratories using routine methodologies authorized by national guidelines, including but not limited to RTs, ELISA, and CLIA. In strict accordance with the Chinese National Guidelines for Detection of HIV/AIDS, a rigorous duplicate retesting protocol was implemented. Immediately upon the detection of an initial reactive result, the original blood sample was retested in duplicate (Figure S1). A specimen progressed to Western blot (WB) confirmation only if at least one of the duplicate tests yielded a reactive result.

Stringent quality assurance protocols were universally enforced. All nine referring institutions and the central confirmatory laboratory were state-certified and operated in strict compliance with the Chinese National Guidelines for Detection of HIV/AIDS. Routine internal quality control (IQC) measures, including standard positive and negative controls, were executed across all platforms for every test batch. Furthermore, all participating laboratories underwent mandatory proficiency testing and continuous External Quality Assessment (EQA) managed by the provincial Center for Disease Control and Prevention (CDC).

Data collection and variable definitions

Demographic parameters (e.g., age, gender, ethnicity, education level, and VCT status) and laboratory screening data were extracted from standardized records maintained by the initial screening institutions. The primary predictor variables were rigorously defined prior to analysis (Supplementary Table S1). Specifically, the variable “Retesting results” was classified into two categories based on the duplicate retesting outcomes: “Both reactive” (both tests yielded reactive signals) and “Single reactive” (only one test yielded a reactive signal, while the other remained non-reactive). Per the national guidelines, specimens with two non-reactive retesting results were directly reported as HIV-negative, did not proceed to confirmation, and were therefore excluded from this cohort. The ultimate dependent variable was the outcome of the WB confirmation. According to the manufacturer’s instructions (MP Biomedicals Asia Pacific Pte. Ltd., Singapore), WB results were initially interpreted as HIV-1 antibody positive, HIV antibody negative, or HIV antibody indeterminate. For the purpose of model development and to address the clinical challenge of delayed diagnosis, these outcomes were further dichotomized into two groups: the anti-HIV-1 positive group and the anti-HIV-1 negative/indeterminate group.

Statistical analysis

Data processing and visualization were conducted utilizing SPSS version 24.0 and R version 4.2.3. The dependent variable was the result of HIV-1 antibody confirmation (coded as 0 for anti-HIV-1 positive and 1 for anti-HIV-1 negative/inconclusive). The independent variables used in the analysis are detailed in Supplementary Table S1. A two-sided P-value < 0.05 was considered statistically significant. Categorical variables were compared using the Chi-square test or Fisher’s exact test. To identify independent predictors for the model, univariate and multivariate logistic regression analyses were performed, and a predictive nomogram was subsequently formulated.

Model validation was conducted systematically. First, to robustly quantify internal stability and address the lack of an external validation cohort, bootstrap resampling validation (1,000 resamples) was performed on the training cohort using R software. This calculated the optimism-corrected C-index (AUC) and generated a bootstrap-based calibration curve, ensuring the predictive performance was stable and not overfitted. Second, the model was evaluated in the internal validation cohort. The Hosmer-Lemeshow test and calibration curves were employed to assess goodness-of-fit, while receiver operating characteristic (ROC) curves were generated to calculate the AUC and determine optimal diagnostic cutoff values.

Results

Demographic profiles and HIV antibody confirmation outcomes

Among the 2468 participants, a total of 2099 individuals tested positive for HIV-1 antibodies, resulting in a positive rate of 85.05% (Table S2). Conversely, there were 369 cases with negative/ inconclusive results, accounting for 14.95% (Table S2).

The training set consisted of 1727 cases, while the validation set comprised 741 cases. There were no significant group differences in age, gender, occupation, ethnicity, marital status, education, residential area, VCT, testing methods and types used for initial screening, and retesting methods and types. However, a statistically significant difference was found in the submitting institutions (Table 1).

Table 1.

The comparison of disparities between the training set and validation set

Variables	Total (n = 2468)	Validation set (n = 741)	Training set (n = 1727)	P-value
Gender, n (%)				0.77
Male	2020 (81.85)	609 (82.19)	1411 (81.70)
Female	448 (18.15)	132 (17.81)	316 (18.30)
Age (years old), n (%)				0.9
≤ 50	1704 (69.04)	513 (69.23)	1191 (68.96)
> 50	764 (30.96)	228 (30.77)	536 (31.04)
Occupation, n (%)				0.2
Employed	783 (31.73)	254 (34.28)	529 (30.63)
Unemployed	859 (34.81)	249 (33.60)	610 (35.32)
Other	826 (33.47)	238 (32.12)	588 (34.05)
Ethnicity, n (%)				0.73
Minority ethnic	95 (3.85)	27 (3.64)	68 (3.94)
Han ethnic	2373 (96.15)	714 (96.36)	1659 (96.06)
Marital status, n (%)				0.74
Unmarried	1637 (66.33)	495 (66.80)	1142 (66.13)
Married	831 (33.67)	246 (33.20)	585 (33.87)
Education, n (%)				0.64
Above high school	1015 (41.13)	310 (41.84)	705 (40.82)
High school or below	1453 (58.87)	431 (58.16)	1022 (59.18)
Residential area, n (%)				0.87
Local	2192 (88.82)	657 (88.66)	1535 (88.88)
Non-local	276 (11.18)	84 (11.34)	192 (11.12)
VCT, n (%)				0.29
Yes	1349 (54.66)	393 (53.04)	956 (55.36)
No	1119 (45.34)	348 (46.96)	771 (44.64)
Submitting institution, n (%)				0.02
Medical institutions	1662 (67.34)	474 (63.97)	1188 (68.79)
CDC	806 (32.66)	267 (36.03)	539 (31.21)
Screening method, n (%)				0.75
ELISA	46 (1.86)	12 (1.62)	34 (1.97)
CL	734 (29.74)	216 (29.15)	518 (29.99)
RT	1688 (68.40)	513 (69.23)	1175 (68.04)
Screening type, n (%)				0.95
Ag/Ab	674 (27.31)	203 (27.40)	471 (27.27)
Ab	1794 (72.69)	538 (72.60)	1256 (72.73)
Retesting type, n (%)				0.68
Not contain Ag/Ab	2445 (99.07)	735 (99.19)	1710 (99.02)
Including Ag/Ab	23 (0.93)	6 (0.81)	17 (0.98)
Retesting results, n (%)				0.19
Both reactive	2250 (91.17)	684 (92.31)	1566 (90.68)
Single reactive	218 (8.83)	57 (7.69)	161 (9.32)

Risk factors associated with HIV-1 antibody negative/inconclusive

The univariate analysis revealed the risk factors of anti-HIV-1 negative/inconclusive results were Gender, Age, Occupation, Marital status, Education, VCT, Submitting institution, Screening type, and methodology and Retesting results (P < 0.05). Incorporating these variables in a multiple logistic regression model further demonstrated that Gender, Education, VCT, and Retesting results were significant determinants influencing anti-HIV-1 negativity or inconclusive (P < 0.05) (Table 2).

Table 2.

Univariate and multivariate logistic regression analysis of risk factors

Variables	Univariable model		Multivariable model
	OR (95%CI)	P-value	OR (95%CI)	P-value
Gender
Male	1.00 (Reference)		1.00 (Reference)
Female	6.67 (5.03 ~ 8.84)	< 0.001	4.36 (2.76 ~ 6.90)	< 0.001
Age (years old)
≤50	1.00 (Reference)		1.00 (Reference)
>50	1.83 (1.41 ~ 2.39)	< 0.001	0.74 (0.45 ~ 1.23)	0.247
Occupation
Employed	1.00 (Reference)		1.00 (Reference)
Unemployed	0.96 (0.71 ~ 1.28)	0.762	0.88 (0.54 ~ 1.45)	0.621
Other	0.37 (0.26 ~ 0.54)	< 0.001	0.94 (0.54 ~ 1.62)	0.817
Ethnicity
minority ethnic	1.00 (Reference)
Han ethnic	1.26 (0.62 ~ 2.58)	0.521	/	/
Marital status
Unmarried	1.00 (Reference)		1.00 (Reference)
Married	2.96 (2.28 ~ 3.85)	< 0.001	1.62 (0.99 ~ 2.64)	0.053
Education
Above high school	1.00 (Reference)		1.00 (Reference)
High school or below	1.53 (1.17 ~ 2.01)	0.002	0.49 (0.29 ~ 0.80)	0.005
Residential area
Local	1.00 (Reference)
Non-local	0.76 (0.49 ~ 1.19)	0.228	/	/
VCT
Yes	1.00 (Reference)		1.00 (Reference)
No	5.20 (3.86 ~ 7.00)	< 0.001	2.13 (1.33 ~ 3.42)	0.002
Submitting institution
Medical institutions	1.00 (Reference)		1.00 (Reference)
CDC	3.73 (2.86 ~ 4.86)	< 0.001	0.84 (0.49 ~ 1.45)	0.539
Screening method
ELISA	1.00 (Reference)
CL	2.18 (0.51 ~ 9.30)	0.294	/	/
RT	3.54 (0.84 ~ 14.90)	0.084	/	/
Screening type
Ag/Ab	1.00 (Reference)		1.00 (Reference)
Ab	2.08 (1.49 ~ 2.91)	< 0.001	0.70 (0.44 ~ 1.10)	0.123
Retesting type
Not contain Ag/Ab	1.00 (Reference)
Including Ag/Ab	2.20 (0.77 ~ 6.30)	0.141	/	/
Retesting results
Both reactive	1.00 (Reference)		1.00 (Reference)
Single reactive	1981.54 (275.06 ~ 14274.79)	< 0.001	1674.82 (226.64 ~ 12376.71)	< 0.001

Construction of a predictive model for anti-HIV-1 negative/inconclusive results

The regression equation was derived by incorporating the four predictors that showed statistical significance in the multivariate logistic regression model. Specifically showing follow: Logit (P) = -2.54 + 1.47 × Gender (Male = 0, Female = 1) − 0.72 × Education (Above high school = 0, High school or below = 1) + 0.76 × VCT (Yes = 0, No = 1) + 7.42 × Retesting results (Both reactive = 0, Single reactive = 1). The complete set of regression coefficients with their associated statistical parameters is tabulated in Supplementary Table S3. The nomogram was utilized to visualize the predictive model (Fig. 1).

Fig. 1.

The nomogram for predicting anti-HIV-1 negative/ inconclusive results

Internal validation and efficacy assessment of the predictive model for anti-HIV-1 negative or inconclusive results

Internal bootstrap validation (B = 1,000) on the training cohort confirmed the model’s stability with negligible overfitting. The optimism-corrected C-index was 0.862 (apparent C-index: 0.865; optimism: 0.004). Additionally, the calibration curve demonstrated excellent agreement with a mean absolute error of 0.013 (Figure S2).

Subsequently, the model was validated using the validation set data. The Hosmer-Lemeshow test indicated a strong goodness of fit for the model (χ² = 12.094, P = 0.098). The calibration curve demonstrated a close alignment between the prediction and calibration curves with the ideal curve, indicating a high level of consistency between predicted and actual probabilities (Fig. 2). This suggests that the model exhibits excellent calibration. The area under the ROC curve was 0.874 (95% CI: 0.826–0.922), with a diagnostic cut-off value of 0.285, sensitivity of 0.99, and specificity of 0.63, indicating excellent discriminative ability for the model (Fig. 3).

Fig. 2.

Calibration curves for the validation set of the prediction model

Fig. 3.

The ROC curve of the validated set for the prediction model

Discussion

The primary objective of this study was to address the inefficiencies of the current HIV diagnostic algorithm by identifying predictors of anti-HIV-1 negative or inconclusive confirmatory outcomes among screen-reactive individuals and developing a clinically actionable Nomogram model. Through univariate and multivariate logistic regression analyses, we identified four independent predictors—gender, education level, VCT status, and retesting results. Among these, discordant retesting outcomes emerged as the most influential variable, carrying significantly greater predictive weight than the other factors. To rigorously address potential over-fitting, the internal stability of the nomogram was confirmed via 1,000-resample bootstrap validation, yielding an excellent optimism-corrected C-index of 0.862. Furthermore, the nomogram demonstrated strong performance during internal validation, with the receiver operating characteristic curve (AUC) of 0.874, a diagnostic cutoff value of 0.285, sensitivity of 0.99, and specificity of 0.63. The calibration curve indicated a strong agreement between predicted and observed probabilities, which was supported by a Hosmer-Lemeshow test result (χ² = 12.094, P = 0.098). These findings not only validate the model’s utility for risk stratification but also provide critical insights into the underlying drivers of discordant screening-confirmatory results, offering targeted directions for optimizing HIV testing workflows.

Among the 2,468 HIV screen-reactive individuals in this study, 2,099 (85.05%) tested positive for anti-HIV-1 via WB, while 369 (14.95%) exhibited discordant WB outcomes (either anti-HIV-1 negative or inconclusive). This rate of screening-confirmatory discrepancy aligns with global reports: WB indeterminate rates for HIV antibody confirmation in China range from 4% to 20% [17–20], consistent with international observations [15, 21], and screen-reactive but WB-negative results are commonly documented in clinical practice [22]. Notably, a negative HIV antibody result does not definitively rule out infection, as this discrepancy can stem from specific disease conditions [14, 23, 24], acute HIV infection [25], or cross-reactivity with other viruses [16].

A key finding of this study is the significant association between discordant initial screening-retesting results and the risk of anti-HIV-1 negative or inconclusive confirmatory outcomes. This factor outweighs all other predictors in the model—underscoring that methodologic factors are the primary contributors to ambiguous confirmatory results [26, 27]. This association is grounded in the inherent variability of HIV screening technologies. Different reagents vary substantially in their detection targets—such as antibody-only versus antigen-antibody (Ag/Ab) combinations—as well as in sensitivity and specificity [28]. Furthermore, methodological differences across major platforms, including enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay (CLIA), and rapid diagnostic tests (RTs), contribute significantly to this variability [27, 29, 30]. For instance, while high-sensitivity reagents are essential for minimizing false negatives in population screening, they are also more susceptible to interference from non-specific antibodies induced by conditions like autoimmune diseases, viral co-infections, or pregnancy [31, 32]. Such interference can produce false-positive initial results, which may later resolve as negative or inconclusive upon retesting with an alternative reagent or platform. Additionally, acute HIV infection represents a critical window where p24 antigen is detectable before seroconversion [33]. This can result in discordant retesting outcomes: antibody-only assays may yield negative results, while Ag/Ab combination tests detect p24 antigen—creating inconsistency that frequently translates to Western blot (WB) inconclusiveness. Moreover, variations in reagent manufacturing, batch-to-batch differences, and laboratory operational procedures further exacerbate retesting discrepancies [34–36]. Collectively, these data confirm that methodologic limitations—including suboptimal reagent performance and incompatible detection algorithms—are the primary drivers of anti-HIV-1 negative or inconclusive confirmatory results. This highlights that the development of detection reagents with both high sensitivity and specificity, alongside the exploration of optimized testing algorithms, is critical to resolving this clinical challenge.

Beyond methodological influences, the model also identified demographic and behavioral variables as significant predictors. Female gender was associated with an increased risk of discordance, consistent with reports suggesting that non-specific antibodies induced during pregnancy or hormonal fluctuations may contribute to false-reactive screening results [37, 38], though the lack of association in some studies [39] indicates that population-specific factors may modulate this relationship. Higher education level served as a protective factor, likely reflecting greater health literacy, earlier testing behavior, and more timely clinical presentation [40]. Non-VCT status was also associated with elevated risk, which can be attributed to the lower prior probability of HIV infection in populations such as general clinical patients or routine health check attendees, compared to VCT participants who often report defined high-risk exposures [41].

Beyond the statistical significance of individual predictors, the clinical value of our model must be contextualized within real-world public health challenges. Conducted in Chengdu, Southwestern China, this study reflects a region that, despite an overall low-prevalence trend, faces a substantial HIV/AIDS burden due to its massive population. Driven by the national “Expanded Free HIV Screening” initiative, the immense testing volume necessitates the widespread use of 3rd-generation antibody-based rapid tests (RTs)—which accounted for 68.4% of initial screenings in our cohort—owing to their cost-effectiveness and rapid turnaround. However, applying high-sensitivity screening (including 4th-generation CLIA or ELISA) across a broad, low-prevalence population inherently diminishes the positive predictive value (PPV). Consequently, widespread screening inevitably generates a significant absolute number of false-reactive initial results, which fundamentally underscores the rationale for the current WB-centric confirmatory algorithm to mitigate misdiagnoses. The proposed nomogram addresses this specific clinical tension by serving as an epidemiologically adaptable triage tool prior to WB.

In low-prevalence frameworks like our current cohort, the model acts primarily as an efficiency-enhancer. By accurately quantifying the risk of a non-confirmatory WB outcome, it prioritizes high-probability false-positives for concurrent nucleic acid testing (NAT), effectively streamlining the diagnostic pathway. Conversely, in high-prevalence settings (e.g., specialized clinics for key populations) where the PPV is inherently elevated, the model’s utility shifts toward diagnostic resolution. In such high-yield scenarios, it identifies the rare subset of reactive screenings likely to be non-confirmatory—flagging potential acute infections or distinct cross-reactivities—and expedites NAT to shorten diagnostic delays without compromising algorithmic rigor.

Together, these findings emphasize that while methodological factors are the dominant source of screening-confirmatory discordance, demographic and behavioral variables also play important roles. The integration of retesting consistency, gender, education, and testing context into a single predictive framework not only aligns with existing literature but also provides a quantifiable, evidence-based approach to identifying individuals at high risk of indeterminate outcomes.

However, translating this streamlined strategy into a routine clinical workflow requires balancing efficiency gains against operational trade-offs. Bypassing WB entirely in favor of NAT carries a theoretical false-negative risk, particularly for elite controllers or individuals on undisclosed antiretroviral therapy (ART) who may exhibit undetectable viral loads. Furthermore, shifting the diagnostic burden to molecular laboratories could introduce cost and capacity challenges in high-volume centers. Therefore, we advocate operationalizing this nomogram not as a rigid replacement, but as an adjunctive decision-support tool. Ultimately, it introduces a rational, evidence-based triage step to optimize resource allocation, advancing the pursuit of personalized and resource-conscious HIV testing strategies across diverse epidemiological landscapes.

Several limitations of this study should be noted. First, the retrospective, single-center design relying on referral cohorts introduces potential selection bias, which may limit generalizability to unselected, community-based populations. Second, extrapolating this model—which was calibrated in a low-prevalence context—to high-prevalence settings requires future recalibration. Finally, while internally stable, prospective multi-center external validation is necessary to independently verify the model’s long-term predictive robustness and cost-effectiveness across diverse epidemiological landscapes. Future research should prioritize multi-center prospective validation across diverse populations and conduct cost-effectiveness analyses to support the model’s clinical translation.

Conclusion

This study develops and validates a Nomogram model incorporating gender, education level, VCT status, and retesting results to predict anti-HIV-1 negative or inconclusive confirmatory outcomes among screen-reactive individuals, with excellent discriminative ability and clinical applicability. By enabling targeted risk stratification, the model streamlines HIV diagnostic workflows, reduces unnecessary Western blot testing and diagnostic delays, and optimizes healthcare resource utilization—particularly valuable for resource-limited settings. This evidence-based tool provides a practical solution for enhancing HIV testing efficiency while maintaining accuracy, supporting progress toward global HIV prevention and control targets.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

HIV

Human Immunodeficiency Virus

AIDS

Acquired Immunodeficiency Syndrome

ART

Antiretroviral Therapy

WB

Western Blot

NAT

Nucleic Acid Testing

ELISA

Enzyme-Linked Immunosorbent Assay

CLIA

Chemiluminescent Immunoassay

RDTs

Rapid Diagnostic Tests

Ag/Ab

Antigen/Antibody

PLHIV

People Living with HIV

VCT

Voluntary Counseling and Testing

WHO

World Health Organization

UNAIDS

The Joint United Nations Programme on HIV/AIDS

ROC

Receiver Operating Characteristic

AUC

Area Under the Curve

PPV

Positive Predictive Value

Author contributions

G.L.: Conceptualization, Writing – original draft, Project administration, Validation, Software；S.Z.: Writing – review & editing, Resources, Project administration, Funding acquisition；Y.W.: Methodology, Formal analysis；Y.J.: Supervision；X.L.: Investigation；H.Q.: Project administration, Resources；H.W.: Writing – review & editing, Funding acquisition.All authors read and approved the final manuscript.

Funding

This work was supported by the Health Commission of Sichuan Province Medical Science and Technology Program, China (25QNMP002), the Clinical Science Research Project of Chengdu Medical College, China (24LHBBYY1-03), the Special Scientific Research Project of Sichuan Medical Association, China (2024HR21), and the Chengdu Medical Research Project of Chengdu Health Commission, China (2025588).

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The present study was approved by the Medical Ethics Committee of the Second Affiliated Hospital of Chengdu Medical College (China National Nuclear Corporation 416 Hospital), China (YJ-2024-030), and an exemption was granted for informed consent. Moreover, this study has been registered in the National Health Security Information Platform-Medical Research Registration and Filing Information System, China (MR-51-24-010247), and all procedures were in accordance with the principles of Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.