Non-disclosure of Known HIV Status among People Living with HIV in the Mozambique Population-Based HIV Impact Assessment (INSIDA 2021)

Abstract

Non-disclosure of known HIV-positive status is a barrier to ending HIV as a global health threat as it leads to biased measurements of HIV-treatment coverage indicators and inaccurate estimates of epidemic progress, resulting in wasted resources. Identifying and understanding factors driving non-disclosure among people living with HIV is necessary for encouraging engagement with HIV services and improving treatment coverage, resource allocation, and monitoring of HIV programs in high HIV-burden areas. This analysis assessed factors associated with non-disclosure among survey respondents who had antiretrovirals (ARVs) detected in blood specimens. HIV-positive blood specimens (n = 2,038) from the 2021 Mozambique Population-based HIV Impact Assessment were tested for the presence of ARVs. Weighted prevalence estimates of non-disclosure and select covariates are reported and factors associated with non-disclosure modeled via multivariate logistic regression. Among 1,358 respondents with ARVs detected, 14.1% did not self-report their HIV-positive status during the interview. Adjusting for socio-demographic and clinical factors, non-disclosure was more likely among younger participants aged 15–24 years (adjusted odds ratio [aOR]: 2.15, 95% Confidence Interval [CI] 1.16–4.01) and among those without knowledge of their recent sexual partner’s HIV-status (aOR: 2.67, 95%CI: 1.38–5.15). Participants with an unsuppressed viral load were over six times (aOR: 6.27, 95%CI: 2.76–14.23) more likely to not disclose. Improving disclosure rates is vital to obtaining accurate HIV-treatment estimates and assessing epidemic progress. Initiatives prioritizing pre- and post-test counseling, stressing treatment literacy, emphasizing undetectable = untransmittable (U = U) campaigns, and encouraging programs that promote social support may encourage disclosure among individuals living with HIV.

Introduction

Mozambique faces a generalized HIV epidemic, marked by 12.5% adult prevalence [1], considered one of the highest in the world [2]. With an estimated 39 million people living with HIV (PLHIV) globally and 2.4 million in Mozambique, controlling the HIV/AIDS epidemic remains a global public health priority [2]. The Joint United Nations Programme on HIV/AIDS (UNAIDS) established the goal of ending the AIDS epidemic as a public health threat by 2030 through a 95–95–95 HIV-treatment cascade, stipulating: 95% of those living with HIV know their status, of those, 95% are on antiretroviral (ARV) therapy (ART), and of those 95% have attained viral load suppression (VLS) [3, 4]. Progress towards achieving these targets is measured through population-based surveys, which utilize a combination of self-reported interview responses and biomarker results [5–7]. Recent population-based HIV impact assessment (PHIA) surveys have incorporated testing for the presence of ARV metabolites in blood samples to verify self-reported responses and adjust estimates to account for discrepancies between self-reported and biomarker-confirmed results [7, 8]. The proportion of survey respondents on ART who do not disclose their HIV-positive status has ranged from 18% in South Africa to 28% in Nigeria [9–12]. Young et al. concluded ARV coverage in Kenya increased nearly 35% when self-reported responses were adjusted by the presence of either undetectable viral load or ARVs in blood [13]. Similarly, after adjusting self-reported responses for the presence of biomarkers, researchers in Nigeria reported the 95-cascade jumped from 30%–90%–83% to 47%–96%–81% [10]. In southern Mozambique, a Health Demographic Surveillance System discovered the proportion of PLHIV aware of their status increased by over 8% after accounting for undisclosed status during HIV-testing [14].

As high-HIV-burden countries approach the 95–95–95 cascade targets, accurate estimates of achievements against the targets become increasingly important. Relying solely on self-report has numerous potential impacts including misallocation of resources between treatment target areas or between countries and misidentification of individuals in need of ART, among others [6, 15]. Absent biomarker adjustments, undisclosed knowledge of HIV-positive status will result in underestimation of both the first and second 95 targets, which has implications for monitoring HIV diagnosis and treatment, assessing program effectiveness, and resource investment [11]. Although there are multiple reasons an underestimation may occur, this paper focuses specifically on non-disclosure.

While discrepancies between self-reported and biomarker confirmed data have been investigated in Southern Africa, to our knowledge such analyses have never been conducted within a representative population of Mozambique. Using data from the 2021 Mozambique Population-based HIV Impact Assessment survey (INSIDA 2021), a population-based serosurvey that combined demographic and HIV-related care and risk behaviors questionnaires with biomarker data, authors seek to describe the cohort of survey respondents who self-reported a negative HIV status but had ARVs detected in their biomarker sample.

Methods

Study Design and Population

INSIDA 2021 was a nationally representative, household survey conducted from April 2021 to February 2022, with a four-stage cluster of 9,015 randomly selected households from 311 enumeration areas. Eligible respondents, aged 15 years and older, who provided written informed consent were interviewed using a standard questionnaire [1]. Of the 17,105 eligible participants interviewed (unweighted response rate 85.9%), 14,488 had blood drawn (unweighted response rate 84.7%), of which 2,038 tested HIV-positive, and of those 1,358 had ARVs detected in blood specimens. After excluding nine participants with incomplete self-reported HIV diagnosis and treatment responses, a total of 1,349 participants remained (Fig. 1).

Fig. 1.

Unweighted number and weighted proportion of adults, 15 years and older, with survey interview, completed blood draw, and completed laboratory HIV testing and ART detection results in Mozambique, INSIDA 2021

Field and Laboratory Procedures for HIV Diagnosis

Trained phlebotomists conducted home-based HIV counseling and testing (HBCT) among eligible participants using a sequential rapid testing algorithm per national guidelines. HBCT included return of results and pre- and post-test counseling, which encompassed risk reduction and HIV prevention, support services, results interpretation, treatment linkage and adherence, and partner testing. Confidential interviews and blood draws were performed at the participant’s preferred location near or in the home.

Eligible participants were initially screened via Determine^™ HIV-1/2 rapid diagnostic test (Abbott Molecular Inc., Des Plaines, Illinois, United States). Non-reactive specimens were classified as HIV-negative and reactive specimens were confirmed with Uni-Gold^™ HIV (Trinity Biotech, plc. Wicklow, Ireland). If the second test was reactive, the specimen was classified as HIV-positive, if non-reactive, the algorithm was repeated. Participants were referred to a health facility for repeat testing after four weeks if the retest was also inconclusive.

Prior to additional testing, blood samples were transported to satellite laboratories where quality assurance procedures were performed. Final HIV status was confirmed for HIV-inconclusive and positive specimens via Geenius HIV 1/2 Supplemental Assay (Bio-Rad, Hercules, California, United States). A positive Geenius result defined HIV positivity for the survey.

Blood samples from participants testing HIV-positive and indeterminate were subjected to CD4-T cell count (cells/microliter [µL]) measurement via PIMA^™ CD4 Analyzer (Abbott Molecular, Inc., Chicago, Illinois, United States, formerly Alere). After which, dried blood spots (DBS) and plasma aliquots were prepared and shipped to the central lab in Maputo for further testing and storage.

Viral Load Testing

HIV-1 viral load (VL) testing (HIV RNA copies per milliliter [mL]), using either DBS or plasma aliquots, was performed on confirmed HIV-positive samples with COBAS AmpliPrep/COBAS TaqMan^® (CAP/CTM), COBAS^® 4800 platforms.

HIV-positive participants received their CD4 and VL results within 8 to 12 weeks at the health facility of their choice.

Antiretroviral Detection

Qualitative screening for detectable ARV concentrations was performed on DBS samples from all HIV-positive participants and those who self-declared HIV-positive but tested HIV-negative, via high-resolution liquid chromatography coupled with tandem mass spectrometry, a modified version of the methodology described by Koal et al. [16]. The qualitative assay is highly specific (as it separates original compound from fragments) and highly sensitive, with a detection limit of 0.02 micrograms/mL (µg/mL) and a signal-to-noise of at least 5:1 for all medicinal products. As detection of all ARVs in use at the time of the interview was economically unfeasible, four ARVs were selected: atazanavir, lopinavir, efavirenz, and dolutegravir, which account for the most prescribed first- and second-line regimens in Mozambique [1]. These ARVs were also selected due to their relatively long half-lives, allowing for a longer period of detection after ingestion, with estimated time to reach a concentration below the detection limit ranging from 1.5 to 28 days [17, 18]. Detection of an ARV indicates recent use at the time of blood collection. ARV detection was performed by University of Cape Town, South Africa.

Measures

Individuals testing HIV-positive were defined as “aware” of their status if they self-reported being HIV-positive or if they had ARVs detected in their blood. Individuals were categorized as “on ART” if they self-reported being on ART or if they had ARVs in their blood. HIV-status was considered undisclosed when persons self-reported their HIV status as HIV-negative or unknown but had detectable ARVs in their blood sample. Viral load suppression was defined as HIV RNA < 1000 copies/ mL in blood.

Data Analysis

Data analysis was restricted to respondents defined as HIV-positive and “on ART” per survey laboratory procedures. Participants missing self-reported HIV diagnosis or treatment responses (n = 9) were excluded from analysis. Analyses were weighted using the jackknife replication method of variance estimation to account for the complex survey design and conducted using the R “survey” package. Nonr-esponse was addressed via inverse-probability weighting and coverage errors through post-stratification [19].

Descriptive analysis was conducted for all covariates (sex, age group, urban residence, province, household wealth quintile, educational attainment, marital status, employment status, knowledge of recent sexual partner’s HIV status, other HIV-positive person in household, VLS, and CD4 category) and proportions and 95% confidence intervals reported. To assess the relationship between covariates and non-disclosure, chi-square tests for association were performed as well as bivariate logistic regression. Variables with a chi-square p-value of 15% or below entered into the multivariate logistic regression model. Variables were then removed in a stepwise fashion, comparing AIC scores until the best model was found. Variables removed were added back to check for confounding; those variables whose addition changed the adjusted odds ratio (aOR) by 10% or more were determined to be confounders and included in the final model. Odds ratios, aORs, and associated 95% confidence intervals and p-values are reported.

For all variables included in the final multivariable logistic regression model, predictive probability models were created and compared to the final model. To evaluate the individual impact of predictor variables on the likelihood of non-disclosure, each predictor variable in the final model was individually removed and the model re-evaluated to assess the effect on non-disclosure. Trend analysis results, assessed using continuous bivariate regression, are reported for variables showing potential trends. We estimated of the number of persons aged 15 years and older with detectable ARVs in their blood that would have self-reported their HIV status as HIV-negative or unknown in Mozambique at the time of the survey.

R version 4.1.3 was used for all analyses [20].

Ethical Considerations

The INSIDA survey protocol was approved by the National Bioethics Committee (CNBS) and the Columbia University Medical Center and the U.S. Centers for Disease Control and Prevention (CDC) IRBs¹. All participants provided written consent prior to interview and blood draw.

Results

One-seventh, 14.1%, of participants with ARVs detected did not self-report their HIV-positive status during the interview (Fig. 1). Respondents who did not disclose their HIV status were more frequently aged 25–34 years (34.6%), resided in Nampula province (17.2%), belonged to the three highest household wealth quintiles, were married or living with a partner (48.0%), were unaware of their recent sexual partner’s HIV status (61.7%), and had a suppressed VL (84.3%) (Table 1). While not significantly associated with nondisclosure, those who did not report a positive HIV-status were more frequently women (63.9%), lived in rural areas (60.2%), had completed primary education (53.1%), were unemployed (62.8%), did not live with another HIV-positive person (66.9%), and had a CD4 count of 500 or above (68.3%).

Table 1.

Socio-demographic and clinical characteristics among HIV-positive adults (15 years and older) who had ARVs detected in Mozambique by self-reported status, INSIDA 2021 (n = 1,349)

Characteristic	Self-Report HIV-negative or unknown status		Self-Report HIV-positive
	%	95% CI	%	95% CI	N	Chi-Square p-value
Sex
Male	36.1	27.8–45.0	32.6	29.6–35.7	395	0.406
Female	63.9	55.0–72.2	67.4	64.3–70.4	954
Age group (years)
15–24	21.2	14.0–29.8	9.1	7.0–11.5	103	< 0.001
25–34	34.6	26.3–43.8	26.1	23.3–29.1	322
35–49	30.8	23.3–39.2	43.7	40.6–46.9	600
50+	13.4	8.4–19.9	21.1	18.6–23.8	324
Residence
Urban	39.8	28.9–51.6	48.8	39.7–57.9	685	0.094
Rural	60.2	48.4–71.1	51.2	42.1–60.3	664
Province
Niassa	10.2	4.7–18.7	2.4	1.5–3.5	61	< 0.001
Cabo Delgado	7.4	2.4–16.5	4.1	2.4–6.6	43
Nampula	17.2	8.4–29.6	9.3	5.8–13.8	98
Zambezia	12.8	6.7–21.5	20.5	16.4–25.1	132
Tete	7.7	3.5–14.3	6.1	4.4–8.3	108
Manica	8.6	4.0–15.5	5.1	3.3–7.3	90
Sofala	7.5	3.4–13.8	8.5	5.3–12.7	95
Inhambane	8.5	4.3–14.7	5.4	3.8–7.4	109
Gaza	6.6	3.5–11.1	12.1	8.8–16.1	212
Maputo Province	7.7	1.9–19.6	17.1	13.1–21.7	215
Maputo City	5.8	2.0–12.8	9.5	6.8–12.8	186
Household wealth quintile
Lowest	13.1	7.3–21.0	8.2	5.3–12.0	98	0.011
Second	18.9	12.2–27.3	10.2	6.9–14.4	141
Middle	22.9	15.9–31.2	20.0	16.0–24.6	247
Fourth	23.9	16.5–32.6	28.0	23.8–32.5	374
Highest	21.2	12.8–32.0	33.6	28.3–39.2	488
Education
No education	22.7	15.8–30.9	21.0	17.4–24.9	270	0.169
Primary	53.1	43.5–62.6	46.8	43.2–50.4	688
Secondary or more than secondary	24.1	17.3–32.2	32.2	27.8–36.8	387
Marital status
Never married	17.1	10.7–25.3	10.5	8.3–13.1	136	< 0.001
Married or living together	48.0	39.0–57.1	62.3	58.9–65.7	775
Divorced or separated	27.1	18.4–37.5	16.1	13.4–19.1	246
Widowed	7.7	4.0–13.1	11.1	9.2–13.1	187
Employed (in past 12 months)
Yes	37.2	28.5–46.6	40.6	36.3–45.0	584	0.508
No	62.8	53.4–71.5	59.4	55.0–63.7	763
Knowledge of most recent sexual partner’s HIV status
No sexual partner in past 12 months	21.3	12.9–31.9	24.2	21.2–27.5	344	< 0.001
Tested together	8.6	5.0–13.7	26.6	23.2–30.2	317
Did not test together	8.4	4.4–14.1	26.0	23.2–29.0	292
Don’t know status	61.7	50.6–72.1	23.2	20.1–26.4	326
Missing	9.6	4.9–16.4	5.9	3.9–8.6	70
HIV+person in household
No	66.9	58.0–74.9	61.1	55.3–66.7	862	0.205
1 or more	33.1	25.1–42.0	38.9	33.3–44.7	487
Viral load category
Suppresseda	84.3	76.0–90.6	96.5	95.1–97.5	1284	< 0.001
Unsuppressed	15.7	9.4–24.0	3.5	2.5–4.9	65
CD4 Category (cells/μL)
< 200	1.8	0.3–5.9	4.6	3.5–5.9	60	0.149
200–349	11.0	6.2–17.8	15.1	12.8–17.6	195
350–499	18.8	12.8–26.1	21.6	18.9–24.6	292
>= 500	68.3	60.0–75.9	58.7	55.9–61.6	798

^aViral load suppression defined as HIV RNA < 1000 copies/mL. Percentages and confidence intervals are weighted and adjusted to account for the survey design. Respondents with missing values were excluded from tabulations (household wealth, 1; education, 4; marital status, 5; employment, 2; CD4, 4)

Predictors of non-disclosure

In bivariate regression, age group, province, household wealth quintile, marital status, knowledge of partner’s status, and VLS were significantly associated with non-disclosure (Table 2). Younger age groups had higher odds of non-disclosure (OR: 3.31, 95% CI: 1.91–5.75) compared to those aged 35–49 years. Residents of Niassa province were more likely not to disclose their status (OR: 9.50, 95% CI: 2.38–37.92) compared to those in Maputo province. Respondents in the lowest and second-lowest household wealth quintiles had nearly three times the odds of reporting a negative status despite having ARVs detected, compared to respondents in the highest household wealth quintile (OR: 2.53, 95% CI: 1.11–5.76; OR: 2.94, 95% CI: 1.46–5.91 for lowest and second-lowest, respectively). Those never married and divorced/separated had twice the odds of not disclosing their status (OR: 2.11, 95% CI: 1.22–3.65; OR: 2.19, 95% CI: 1.37–3.51, respectively) as married respondents. Participants without knowledge of their recent partner’s HIV status had three times the odds of non-disclosure (OR: 3.04, 95% CI: 1.63–5.68) compared to those without a recent sexual partner, and those who knew their partner’s status, irrespective of testing together or not (OR: 0.37, 95% CI 0.18–0.78; OR: 0.37, 95% CI: 0.17–0.78, respectively), were more likely to disclose. People living with HIV with unsuppressed VL were five times more likely to not disclose (OR: 5.10, 95% CI: 2.71–9.59).

Table 2.

Unadjusted and adjusted logistic regression models of non-disclosure of HIV status among HIV-positive adults (15 years and older) who had ARVs detected by select socio-demographic and clinical factors, INSIDA 2021

Characteristic	Self-Report HIV-negative or unknown status
	OR	95% CI	p-value	aOR	95% CI	p-value
Sex
Male	1.17	0.81–1.68	0.410
Female	(ref)
Age group (years)
15–24	3.31	1.91–5.75	< 0.001	2.15	1.16–4.01	0.016
25–34	1.88	1.21–2.92	0.005	1.68	1.03–2.76	0.038
35–49	(ref)			(ref)
50+	0.90	0.51–1.60	0.723	0.94	0.48–1.83	0.850
Residence
Urban	(ref)
Rural	1.44	0.93–2.22	0.101
Province
Niassa	9.50	2.38–37.92	0.002	7.85	1.98–31.13	0.004
Cabo Delgado	3.94	0.77–20.06	0.099	1.63	0.34–7.76	0.538
Nampula	4.09	0.98–17.09	0.053	2.62	0.69–9.91	0.155
Zambezia	1.38	0.37–5.16	0.632	1.03	0.27–3.93	0.969
Tete	2.76	0.61–12.50	0.185	2.35	0.60–9.22	0.220
Manica	3.73	0.87–16.02	0.076	3.98	0.96–16.53	0.057
Sofala	1.94	0.51–7.45	0.332	1.33	0.34–5.27	0.681
Inhambane	3.48	0.92–13.27	0.067	2.76	0.67–11.46	0.160
Gaza	1.21	0.30–4.92	0.792	1.12	0.28–4.42	0.873
Maputo Province	(ref)			(ref)
Maputo City	1.36	0.30–6.16	0.687	1.52	0.32–7.19	0.595
Household wealth quintile
Lowest	2.53	1.11–5.76	0.027	1.28	0.6–2.73	0.528
Second	2.94	1.46–5.91	0.003	1.83	0.86–3.93	0.119
Middle	1.81	0.94–3.48	0.075	1.10	0.55–2.19	0.789
Fourth	1.35	0.70–2.61	0.373	1.07	0.53–2.18	0.841
Highest	(ref)			(ref)
Education
No education	0.95	0.58–1.58	0.853
Primary	(ref)
Secondary or more than secondary	0.66	0.43–1.02	0.062
Marital status
Never married	2.11	1.22–3.65	0.008	1.15	0.55–2.40	0.715
Married or living together	(ref)			(ref)
Divorced or separated	2.19	1.37–3.51	0.001	1.67	0.96–2.90	0.072
Widowed	0.90	0.50–1.65	0.738	0.88	0.45–1.74	0.711
Employed (in past 12 months)
Yes	(ref)
No	1.15	0.75–1.77	0.510
Knowledge of most recent sexual partner’s HIV status
No sexual partner in past 12 months	(ref)			(ref)
Tested together	0.37	0.18–0.78	0.009	0.44	0.19–1.03	0.057
Did not test together	0.37	0.17–0.78	0.010	0.37	0.16–0.87	0.023
Don’t know status	3.04	1.63–5.68	< 0.001	2.67	1.38–5.15	0.004
Missing	1.93	0.80–4.63	0.141	1.74	0.70–4.34	0.234
HIV+person in household
No	(ref)
1 or more	0.78	0.50–1.20	0.256
Viral load category
Suppresseda	(ref)			(ref)
Unsuppressed	5.10	2.71–9.59	<0.001	6.27	2.76–14.23	< 0.001
CD4 Category (cells/μL)
< 200	0.34	0.10–1.15	0.082	0.26	0.07–0.98	0.046
200–349	0.63	0.33–1.19	0.154	0.49	0.23–1.05	0.067
350–499	0.75	0.48–1.17	0.200	0.68	0.39–1.18	0.167
>= 500	(ref)			(ref)

^aViral load suppression defined as HIV RNA < 1000 copies/mL. Percentages and confidence intervals are weighted and adjusted to account for the survey design. OR = odds ratio, CI = confidence interval, aOR = adjusted odds ratio. Ten respondents with missing values for one or more variables excluded from final model

In a separate bivariate regression, an inverse relationship between age and non-disclosure was observed, in which for every 10-year increase in age, non-disclosure was 36% less likely (Supplementary Table 1).

The final multivariable regression model was adjusted for age, province, household wealth quintile, marital status, knowledge of recent sexual partner’s HIV status, VLS, and CD4 (Table 2). The odds of HIV non-disclosure decreased as age increased, where younger participants aged 15–24 years had twice the odds (aOR: 2.15, 95% CI 1.16–4.01) of not disclosing compared to participants 35–49 years-old. Participants residing in Niassa province were significantly more likely to not disclose (aOR: 7.85, 95% CI: 1.98–31.13) compared to residents of Maputo Province. Compared to respondents without a sexual partner in the past year, those who knew their recent partner’s HIV status, though they did not test together, were roughly 40% less likely to not disclose their HIV status (aOR: 0.37, 95% CI: 0.16–0.87); whereas, those without knowledge of their partner’s status were nearly three times more likely to not disclose their positive status (aOR: 2.67, 95% CI: 1.38–5.15). The odds of non-disclosure were over six times greater among respondents with unsuppressed VL (aOR: 6.27, 95% CI: 2.76–14.23) compared to those with suppressed VL. Those with a CD4 count below 200 (aOR: 0.26, 95% CI: 0.07–0.98) were more likely to disclose compared to those with a CD4 greater than 500. In the adjusted model, household wealth was not associated with non-disclosure.

Predictive Probability

When adjusted for the inclusion of other covariates, the probability of non-disclosure among those with ARVs detected that were aged 15–24 years was 20% and 11% in those aged 50 years and above (Table 3). The probability of non-disclosure was greatest among those living in Niassa province, 33%, and the lowest in Maputo Province, 7%. Among those who did not know their recent sexual partner’s status, 25% did not disclose their HIV-positive status. There was a 38% probability of people with antiretrovirals detected in blood with unsuppressed VL not disclosing an HIV-positive status and a 13% probability of non-disclosure among those with a suppressed VL.

Table 3.

Comparison of unadjusted prevalence and adjusted (predicted) prevalence of non-disclosure among HIV-positive adults, INSIDA 2021

Characteristic	Unadjusted prevalence of non-disclosure ± SE	Adjusted (predicted) prevalence of non-disclosure ± SE
Age group
15–24	0.28 ± 0.05	0.20 ± 0.04
25–34	0.18 ± 0.02	0.17 ± 0.02
35–49	0.10 ± 0.01	0.12 ± 0.01
50+	0.09 ± 0.02	0.11± 0.02
Province
Niassa	0.41 ± 0.07	0.33 ± 0.06
Cabo Delgado	0.23 ± 0.10	0.16 ± 0.05
Nampula	0.23 ± 0.06	0.19 ± 0.04
Zambezia	0.09 ± 0.02	0.10 ± 0.02
Tete	0.17 ± 0.05	0.18 ± 0.04
Manica	0.22 ± 0.06	0.21 ± 0.06
Sofala	0.13 ± 0.03	0.11± 0.03
Inhambane	0.21 ± 0.04	0.20 ± 0.04
Gaza	0.08 ± 0.02	0.09 ± 0.02
Maputo Province	0.07 ± 0.03	0.07 ± 0.05
Maputo City	0.09 ± 0.03	0.11± 0.03
Wealth quintile
Lowest	0.21 ± 0.06	0.15 ± 0.03
Second	0.23 ± 0.04	0.19 ± 0.03
Middle	0.16 ± 0.03	0.14 ± 0.02
Fourth	0.12 ± 0.02	0.13 ± 0.02
Highest	0.09 ± 0.02	0.13 ± 0.02
Marital status
Never married	0.21 ± 0.04	0.14 ± 0.03
Married or living together	0.11 ± 0.01	0.13 ± 0.02
Divorced or separated	0.22 ± 0.04	0.18 ± 0.03
Widowed	0.10 ± 0.03	0.12 ± 0.03
Knowledge of most recent
sexual partner’s HIV status
No sexual partner in past 12 months	0.12 ± 0.03	0.12 ± 0.03
Tested together	0.05 ± 0.01	0.06 ± 0.02
Did not test together	0.05 ± 0.01	0.06 ± 0.02
Don’t know status	0.30 ± 0.04	0.25 ± 0.03
Missing	0.21 ± 0.06	0.19 ± 0.06
Viral load category
Suppresseda	0.13 ± 0.01	0.13 ± 0.01
Unsuppressed	0.42 ± 0.08	0.38 ± 0.07
CD4 Category
< 200	0.06 ± 0.04	0.06 ± 0.03
200–349	0.11 ± 0.03	0.10 ± 0.03
350–499	0.13 ± 0.02	0.13 ± 0.02
>= 500	0.16 ± 0.02	0.16 ± 0.02

^aViral load suppression defined as HIV RNA < 1000 copies/mL. Prevalence and confidence intervals are weighted and adjusted to account for the survey design

Population Estimates

Per our estimates, of the 1,298,000 (95% CI: 1,188,000–1,408,000) estimated people with ARVs in their blood in Mozambique, 181,000 (95% CI: 141,000–217,000) would have self-reported their status as HIV-negative or unknown at the time of the survey (Supplementary Table 2).

Discussion

Our study found that one in seven HIV-positive respondents self-reported an HIV-negative status during the survey despite having ARVs detected in their blood, which, per our estimates, represents over 180,000 people in Mozambique not disclosing their HIV-positive status at the time of the survey. Analysis of a Kenya population-based survey similarly revealed that undisclosed HIV status resulted in an underestimation of HIV-positive individuals by 9%, approximately 112,000 people, when based solely on self-reported status [11].

In both surveys and HIV testing campaigns, non-disclosure is more prevalent in younger PLHIV, particularly those under the age of 30 [12, 14, 21, 22]. In Kenya, after adjusting self-reported status for the presence of ARVs or undetectable VL, the youngest age group exhibited the greatest increases in all 95-cascade indicators [13]. Likewise, younger INSIDA survey participants were found significantly more likely to not disclose their HIV-positive status, based on rapid test. As in Nigeria where the sensitivity of self-reported HIV status improved with age [10], our results showed a trend of non-disclosure decreasing with increasing age.

Survey participants may incorrectly report awareness of their HIV-positive status to interviewers for myriad reasons including misunderstanding, stigma, or past negative experiences with the health system. While our study highlights non-disclosure in survey settings and its impact on 95–95–95 cascade target estimation, non-disclosure can also occur in healthcare settings, impacting the quality of care, programmatic planning, and resource allocation. In southern Mozambique, interviews with HIV-positive persons in treatment who did not disclose their serostatus to healthcare providers revealed prior negative encounters with providers, desire to retest under the belief they may have been cured, miscommunication of result, desire to re-engage in care, distrust of initial result, and stigma-related motives as barriers to disclosure [23]. Some study participants disclosed their HIV status to their community but not to a provider, suggesting challenges in the patient-provider relationship, not necessarily stigma/discrimination [23]. Similar findings in South Africa identified misunderstanding of treatment and medication, rather than social desirability bias, as primary drivers in inconsistent self-reporting of HIV status [9]. Nevertheless, in rural South Africa, one-third of individuals initiating or reinitiating ART in ambulatory clinics were found to have undisclosed ART use [24], and interviews with individuals not adhering to ART found that fear of inadvertent disclosure limited interactions with healthcare providers [25], suggesting stigma and discrimination remain barriers. Similarly, in Mozambique and Kenya, a history of missed clinic visits was associated with non-disclosure within home-based testing environments [11, 14]. In 2022, 4% of people testing HIV-positive in Mozambique health facilities and community clinics had previously tested positive [26], indicating a need for improved pre- and post-test counseling to inform PLHIV about the need to test, enhance coping mechanisms, and strengthen linkage to care.

Among persons not adhering to ART, disclosure can be both a benefit and a barrier to treatment adherence. While disclosing to family, friends, or religious groups can provide encouragement to continue treatment, inadequate public support hinders testing and early treatment adherence [25, 27]. INSIDA participants who did not know their partner’s status were three times more likely to not disclose their own status, highlighting the likely role of social support in HIV disclosure. In HIV-treatment clinics in Tanzania, Namibia, and Kenya, individuals who disclosed their serostatus to sexual partners were 24 times more likely to also know their partner’s status [28]. Other studies found 90% of those in stable relationships disclosed their serostatus to their partner [21, 22]. Fear of inadvertent disclosure remains a frequently cited motive for non-disclosure to anyone, emphasizing the need for support networks such as “treatment buddies”, which may encourage disclosure of serostatus or ART use by providing social support [6, 15, 25, 29]. In Mozambique, community adherence and support groups (Grupo de Apoio e Adesão Comunitário [GAAC] in Portuguese) provide PLHIV on treatment with psychological support and allow for less frequent medication pick-ups [30]; however, additional studies are needed to examine the relationship between GAAC members and disclosure of their HIV status.

Findings on the relationship between non-disclosure and wealth vary. In accordance with a South-African study [31], wealth was not significant in the final model; however, other studies concluded higher wealth to be associated with undisclosed HIV status [10, 11]. Conversely, in bivariate analysis, we found non-disclosure greater in lower household wealth quintiles.

Finally, our results suggest those who are currently virally unsuppressed, whether due to poor adherence, drug resistance, or very recent diagnosis and ART initiation, are less likely to accurately report their HIV status. These findings differ from another population-based survey in South Africa which found VLS lower in people who self-reported to be on ART but did not have ARVs detected versus those who self-reported not on ART but had ARVs detected [22]. Alternatively, in bivariate analysis, a Kenyan national survey found VLS associated with undisclosed HIV status while on ART [11]. Considering the differences between disclosure of HIV status and ART use, qualitative studies investigating motives to disclose serostatus versus ART use may be beneficial in increasing linkage and retention in HIV care.

Limitations

Although this analysis offers important insights into the factors associated with non-disclosure of HIV status, limitations must still be considered. By definition, HIV-positive participants who knew their status but did not disclose it and were not on ARVs were excluded from ARV adjustment and, consequently, from this analysis. Discrepancies between biological and self-reported HIV status can arise at various points of the HIV care continuum or during survey implementation, including participant misinterpretation of his/her HIV status, deliberate refusal to disclose, participant misunderstanding of survey questions, or interviewer misreporting the participant’s response. By continuing to emphasize participant comprehension of survey questions, reducing staff misreporting, and improving pre-and post-HIV test counseling, future HIV survey staff trainings can minimize reporting issues. Interview structure (women had greater opportunity than men to correctly report their status given they were asked their HIV status with respect to HIV testing during antenatal care) may also influence patterns of serostatus misreporting [21]. While respondents are advised (during informed consent) that the survey is confidential, it is possible survey respondents refused to respond to questions or to disclose their status for fear of a breach in confidentiality. Unfortunately, it is difficult to directly measure stigma or respondent misunderstanding of diagnosis, treatment, or frustration with the health system in structured interview format. Therefore, we suggest qualitative follow-up with survey participants who failed to disclose their status to better understand the factors leading to their non-disclosure. Finally, some respondents may have been on ART regimens not included in the ARV panel or may have had poor adherence resulting in a false negative ARV screen, which would result in an underestimate of non-disclosure.

Conclusion

A notable proportion (14.1%) of PLHIV on ART did not disclose their known HIV-positive status during the survey. Such non-disclosure would result in underestimation of first and second 95 target achievement, if not adjusted for biomarkers, impacting future funding and resource allocation. PLHIV may choose not to disclose known HIV-positive status for numerous reasons, and future studies should consider that non-disclosure may not be driven solely by fear of stigma and discrimination but could also indicate poor understanding of health status or survey questions. To encourage disclosure and decrease repeat testing, initiatives should prioritize improving pre- and post-test counseling within healthcare settings. This includes accentuating treatment literacy to ensure comprehensive understanding of diagnosis and treatment and empowering PLHIV to manage their care effectively. In ongoing efforts to reduce stigma and discrimination, there is a need to expand and emphasize U = U (Undetectable = Untransmittable) campaigns, which both encourage treatment retention and play a crucial role in combating the spread of misinformation. Finally, strengthening social support programs, such as “treatment buddies” or GAAC in Mozambique, could improve psychological wellbeing and increase disclosure particularly among younger PLHIV.

Supplementary Material

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s10461-025-04699-7.

Data Availability

The 2021 INSIDA public release datasets and files are available at the ICAP PHIA website (https://phia-data.icap.columbia.edu/datasets?country_id=15). Survey data use manuals and documentation are available for download. For access to datasets, please register for an account and submit a data request form.