Of 68,564 adults and adolescents > 14 years included in the Tanzania and Zambia PHIA studies, 55,340 had a matching interview and biomarker (including HIV testing) datasets. Of these, 39,103 (70.7%), had previously been tested for HIV and 181 (0.3%) had missing previous HIV testing data, and hence excluded from the analysis. Of the remaining 16,056, 15160 (94.4%) were tested for HIV. Excluding those who didn’t have sampling weights, the final study sample was 14,820. (Fig. 1)
Compared to individuals who were eligible for inclusion in this study but had missing HIV testing result, those who were tested for HIV during the survey were likely to be older, male, from urban area. They were also less likely to have multiple sexual partners or sexually transmitted infection in the past 12 months. (Supporting Information Table 1)
Of 14,820 study participants, 57.8% were men, and had a median age of 30 (IQR: 21–24). HIV prevalence was 2.3% (95% confidence interval (CI): 2.0-2.6). HIV prevalence was higher for the age category 25–49, among women, and in urban settings. All HIV risk factors, except for those with Presumptive TB, TB disease, or chronic illness, were found to be statistically significant predictors of HIV infection in individuals who were never tested for HIV. (Table 1)
Figure 2 summarizes HIV prevalence by risk factor. The highest was recorded for people who sold sex (13.5%), followed by spouses of HIV infected adults (11%) and those who were divorced, separated, or widowed (6.1%). The presence of other risk factors had HIV testing yield ranging from 3.1%-5.5%. TB in the past 10 years had a testing yield of 33.3% for Zambia compared to 3.7% for those without TB in the past 10 years, p value < 0.001 (data not shown).
Table 1
Determinants of HIV Infection among adults and adolescents > 14 years who were never tested for HIV before PHIA Surveys conducted in Zambia (2016) and Tanzania (2016–2017)
Variable
|
Response
|
Total, n
|
HIV+, n (%)
|
P value
|
Age
|
15–24
|
8,114
|
49 (0.6%)
|
< 0.001
|
|
25–49
|
3,443
|
196 (5.7%)
|
|
|
50+
|
3,263
|
95 (2.9%)
|
|
Gender
|
Male
|
8,567
|
162 (1.9%)
|
0.002
|
|
Female
|
6,253
|
176 (2.8%)
|
|
Residence
|
Urban
|
4,707
|
130 (2.8%)
|
0.026
|
|
Rural
|
10,113
|
207 (2.1%)
|
|
Education
|
No education
|
2,647
|
82 (3.1%)
|
0.001
|
|
Primary
|
8,182
|
197 (2.4%)
|
|
|
Secondary
|
3,859
|
55 (1.4%)
|
|
|
Tertiary
|
132
|
4 (3.1%)
|
|
Wealth Quintile
|
Lowest
|
3,398
|
73 (2.1%)
|
0.038
|
|
Secondary
|
3,412
|
70 (2.0%)
|
|
|
Middle
|
3,081
|
73 (2.4%)
|
|
|
Fourth
|
2,432
|
79 (3.2%)
|
|
|
Highest
|
2,496
|
43 (1.7%)
|
|
Marital status
|
Single, Married
|
13,045
|
230 (1.8%)
|
< 0.001
|
|
DSW$
|
1,775
|
108 (6.1%)
|
|
Spouse is Known to have HIV
|
No
|
14,777
|
333 (2.3%)
|
0.001
|
|
Yes
|
43
|
5 (11.0%)
|
|
Having Paid Work*
|
No
|
9,692
|
172 (1.8%)
|
< 0.001
|
|
Yes
|
5,128
|
166 (3.2%)
|
|
Slept Away from Home for > 1 month*
|
No
|
12,939
|
273 (2.1%)
|
0.015
|
|
Yes
|
1,881
|
65 (3.5%)
|
|
Multiple Sexual Partners*
|
No
|
12,970
|
282 (2.2%)
|
0.042
|
|
Yes
|
1,850
|
58 (3.1%)
|
|
Ever Sold Sex
|
No
|
14,786
|
333 (2.3%)
|
< 0.001
|
|
Yes
|
34
|
5 (13.5%)
|
|
Paid for Sex*
|
No
|
14,100
|
304 (2.2%)
|
< 0.001
|
|
Yes
|
720
|
36 (4.9%)
|
|
Sexually transmitted infection*
|
No
|
13,385
|
274 (2.0%)
|
< 0.001
|
|
Yes
|
1,435
|
65 (4.5%)
|
|
Has Cervical Cancer
|
No
|
14,755
|
335 (2.3%)
|
0.025
|
|
Yes
|
65
|
4 (5.5%)
|
|
Presumptive TB#*
|
No
|
14,460
|
326 (2.3%)
|
0.215
|
|
Yes
|
360
|
12 (3.3%)
|
|
TB disease, current or past
|
No
|
14,639
|
330 (2.3%)
|
0.055
|
|
Yes
|
181
|
8 (4.4%)
|
|
Sick for the past 3 months*
|
No
|
14,242
|
317 (2.2%)
|
0.078
|
|
Yes
|
578
|
22 (3.8%)
|
|
Total
|
|
14,820
|
338 (2.3%)
|
|
$Divorced, Separated, or Widowed; *within the last 12 months of the survey, |
#Cough, fever, night sweats, or weight loss |
Looking at the performance of Tool 1 at different risk score levels, those individuals having one or more risk factors were found to have an HIV prevalence of 3.2% which increased with increasing cut-off: 4.4%, 5.6%, 7.9% HIV prevalence for two, three, and four cut-off scores respectively. (Table 2) Fig. 3 indicates the ROC curve comparing the different cut-off points for Tool 1. Area under the curve (AUC) can be seen to reduce as the risk assessment cut-off increases. A score of ≥ 1 was found to have the highest sensitivity at 82.3% (95% CI: 78.6%-85.9%) with the next score of ≥ 2 having nearly half the sensitivity at 46.8% (42.0%-51.6%). The specificity was higher for a higher cut-off. Positive predictive value was higher for a higher cut-off point while negative predictive value was comparable between all cut-0ff scores.
Table 2
Association of HIV Risk Scores with HIV Infection using a tool that contains all HIV Risk Factors for Adults and Adolescents > 14 years who were never tested for HIV before PHIA Surveys conducted in Zambia (2016) and Tanzania (2016–2017)
Risk Score
|
Response
|
Total, n
|
HIV+, n (%)
|
P value
|
Score ≥ 1
|
No
|
6,122
|
59 (1.0%)
|
< 0.001
|
|
Yes
|
8,698
|
279 (3.2%)
|
|
Score ≥ 2
|
No
|
11,238
|
179 (1.6%)
|
< 0.001
|
|
Yes
|
3,582
|
159 (4.4%)
|
|
Score ≥ 3
|
No
|
13,556
|
267 (2.0%)
|
< 0.001
|
|
Yes
|
1,264
|
71 (5.6%)
|
|
Score ≥ 4
|
No
|
14,445
|
308 (2.1%)
|
< 0.001
|
|
Yes
|
375
|
30 (7.9%)
|
|
Compared to a cut-off score of ≥ 1, AUC was comparable with a cut-off score of ≥ 2 while it was lower for those with higher cut-off scores (p value < 0.001). (Table 3) The AUC was comparable by age, gender, and residence. (Table 4)
Table 3
Sensitivity, Specificity, Positive Predictive Value (PPV+), and Negative Predictive Value (NPV-) for HIV Risk Screening Tool Containing Various Combinations of All Risk Factors
Screening Tool Score*
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
AUC**
|
P value***
|
Score ≥ 1
|
82.3% (78.6%-85.9%)
|
41.9% (41.1%-42.7%)
|
3.2% (2.8%-3.6%)
|
99.0% (98.8%-99.3%)
|
0.6116
|
|
Score ≥ 2
|
46.8% (42.0%-51.6%)
|
76.4% (75.7%-77.1%)
|
4.4% (3.7%-5.1%)
|
98.4% (98.2%-98.6%)
|
0.6159
|
0.7137
|
Score ≥ 3
|
21.0% (17.1%-24.9%)
|
91.8% (91.3%-92.2%)
|
5.6% (4.3%-7.0%)
|
98.0% (97.8%-98.3%)
|
0.5599
|
< 0.001
|
Score ≥ 4
|
8.9% (6.2%-11.6%)
|
97.6% (97.4%-97.9%)
|
8.0% (5.1%-10.9%)
|
97.9% (97.6%-98.1%)
|
0.5332
|
< 0.001
|
* Score ≥ 1 means an individual who has one or more risk factors for HIV; ** AUC = Area under the Curve of a Receiver Operating Curve; |
*** P value compares AUC for a given score with the reference Score ≥ 1 |
Table 4
Comparison of Receiver Operating Characteristics Curve for HIV Risk Screening Tool (Score ≥ 1) by Age, Gender, and Residence for Adults and Adolescents > 14 years who were never tested for HIV before PHIA Surveys conducted in Zambia (2016) and Tanzania (2016–2017)
Variable
|
Response
|
Observation
|
Area Under ROC Curve
|
P value
|
Age
|
15–24
|
7,868
|
0.5756
|
0.6033
|
|
25–49
|
3,507
|
0.5705
|
|
|
50+
|
3,445
|
0.5478
|
|
Gender
|
Male
|
7,945
|
0.6119
|
0.6448
|
|
Female
|
6,875
|
0.6212
|
|
Residence
|
Urban
|
4,461
|
0.6070
|
0.7531
|
|
Rural
|
10,359
|
0.6137
|
|
Figure 4 summarizes relationship between eligibility, sensitivity, and PPV or HIV testing yield. Eligibility for HIV test decreased with increasing of risk score cut-offs: 56% would be eligible with a cut-off score of ≥ 1 while it was 2% for a cut-off score of ≥ 4. HIV testing positivity (PPV) and sensitivity or eligibility was negatively correlated with both going down with increasing cut-off score while PPV increased.
In the tool that contained only statistically significant risk factors (Tool 2), being DSW (odds ratio (OR): 3.9 (95% CI: 2.9–5.2); p-value < 0.001), being spouse of a known HIV + person (OR: 6.1 (95% CI:2.0-19.1); p-value = 0.003), having history of selling sex for money (OR: 7.7 (95% CI:3.6–16.3); p-value < 0.001), having sexually transmitted infections in the past 12 months (OR: 2.1 (95% CI:1.4-3); p-value < 0.001), and working in the past 12 months (OR: 2.1 (95% CI:1.4-3); p-value < 0.001) were included in the final model. Tool 4 that contained customized risk factors, the combination of risk factors having paid work in the past year and sleeping away from home for more than a month in the past 12 months were combined as predictors in addition to conventional risk factors. Having a paid work and sleeping away from home were statistically significant predictors of undiagnosed HIV infection (OR: 1.8 (95% CI: 1.1-3.0)).
Looking at the different risk assessment tools, all were statistically significant predictors of HIV infection with p-value < 0.001. For all tools, if none of the risk factors were present, HIV prevalence was low at 1.0-1.3%. (Table 5) Sensitivity was better for Tool 1 but the corresponding specificity was the lowest. AUC was better for all other tools as compared to this tool and the difference was much higher for Tools 3 and 4 (p-value < 0.001). (Table 6) PPV or HIV testing yield was highest for Tools 3 and 4 at 4.2% and 4.0%, respectively, if at least one risk factor was present. Tool 3 has the lowest proportion of people eligible for testing at 34%the highest being for tool 1 at 59%. (Table 5) Number needed to test (NNT+) was 24 for Tool 3 while it was 43 if universal testing was used.
Table 5
Association of Risk Scores with HIV Infection using a tool that contains all HIV Risk Factors for Adults and Adolescents > 14 years who were never tested for HIV before PHIA Surveys conducted in Zambia (2016) and Tanzania (2016–2017)
Risk Assessment Tool
|
Response
|
Total, n
|
HIV+, n (%)
|
P value
|
Tool 1: All Risk Factors: Score ≥ 1
|
No
|
6,122
|
59 (1.0%)
|
< 0.001
|
|
Yes
|
8,698
|
279 (3.2%)
|
|
Tool 2: Statistically Significant Risk Factors in final model: Score ≥ 1
|
No
|
7,850
|
90 (1.2%)
|
< 0.001
|
|
Yes
|
6,970
|
247 (3.6%)
|
|
Tool 3: Conventional Risk Factors: Score ≥ 1
|
No
|
9,717
|
123 (1.3%)
|
< 0.001
|
|
Yes
|
5,103
|
215 (4.2%)
|
|
Tool 4: Customized Tool: Score ≥ 1
|
No
|
9,294
|
117 (1.3%)
|
< 0.001
|
|
Yes
|
5,526
|
221 (4.0%)
|
|
Table 6
Sensitivity, Specificity, Positive Predictive Value (PPV+), and Negative Predictive Value (NPV-) for each potential HIV Risk Screening Tool for Adults and Adolescents > 14 years who were never tested for HIV before PHIA Surveys conducted in Zambia (2016) and Tanzania (2016–2017)
Risk Factor Selection Strategy
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
AUC**
|
P value***
|
Tool 1: All risk factors
|
82.3% (78.6%-85.9%)
|
41.9% (41.1%-42.7%)
|
3.2% (2.8%-3.6%)
|
99.0% (98.8%-99.3%)
|
0.6116
|
|
Tool 2: Statistically significant only#
|
73.4% (69.1%-77.6%)
|
53.6% (52.8%-54.4%)
|
3.6% (3.1%-4.0%)
|
98.9% (98.6%-99.1%)
|
0.6267
|
0.035
|
Tool 3: Conventional risk factors
|
63.5% (58.9%-68.1%)
|
66.2% (65.5%-67.0%)
|
4.2% (3.6%-4.8%)
|
98.7% (98.5%-98.9%)
|
0.6469
|
< 0.001
|
Tool 4: Customized tool*
|
65.5% (61.0%-70.1%)
|
63.4% (62.6%-64.2%)
|
4.0% (3.5%-4.5%)
|
98.7% (98.5%-99.0%)
|
0.6412
|
< 0.001
|
#Only those included in the final model were considered; *Customized tool = Conventional risk factors + Working for a payment in the past 12 months and Sleeping away from home for at least 1 month in the past 12 months of the survey. **AUC = Area under the Curve of a Receiver Operating Curve; |
*** P value compares AUC for a given risk assessment tool with the reference tool that contains all risk factors. |