Inform Target Product Profile for Rapid Tests to Identify High Viral HBV Infected Pregnant Women: An Individually Selected Study in African Health Workers | BMC Medical

Sub Levels


A total of 576 HCWs responded to the online survey. After excluding 21 non-Africans, the analysis included 555 health workers from 41 African countries. Table 2 shows its characteristics. The majority (70.6%) were aged 30-50 years and 44.5% were female. The breakdown was physicians (62.9%), public health workers (10.3%), laboratory staff (9.4%), midwives (5.6%) and nurses (4.1%). Employed at national hospitals (28.5%), state hospitals (12.8%), district hospitals (10.3%), primary care (7.7%), private (11.7%), public health (18.6%), and others (10.4%) there were. %). More than half (66.3%) reported involvement in HBV patient management or program delivery. Overall, 69.2%, 9.5%, 9.4%, 6.7% and 5.2% worked in West, South, East, North and Central Africa respectively. On the dominance test, most participants (85.9%, 477/555) were rational responders. Reasonable responses were observed in more women than men (89.9% vs 82.8%) and more doctors than nurses (89.4% vs 73.9%) (Table 2). In half of the 12 ‘nondominant’ choice tasks, the pattern of choice between two options differed significantly between rational and irrational respondents (Additional File 4: Table S1).

Table 2 Relations between participant characteristics and being a rational responder

Alternate Unique Multinomial (ASM) Probit Regression Analysis

ASM probit regression model including all respondents (n = 555) had statistically significant coefficients for all attribute levels except time to results of 60 minutes (Additional File 4: Table S2) . Increased cost was significantly associated with decreased coefficient, and increased sensitivity and specificity were associated with increased coefficient. The size of the coefficient (β) of the highest attribute level to the reference level was in order of sensitivity (1.461), cost (-1.118), specificity (0.522) and time to result (-0.031). . When only rational responders were included in the model (n = 477), all coefficients at the eight attribute levels remained statistically significant, with absolute values ​​lower than those for the analysis including all responders. increased (Additional File 4: Table S2). This suggested that there was indiscriminate misclassification among irrational responders. We therefore excluded them from subsequent analyses. We considered a high degree of linear correlation between attribute levels and coefficient values ​​(Additional File 4: Figure S1), after which we considered cost, sensitivity and specificity as continuous variables. Additional File 4: Table S3 shows her ASM probit results with continuous variables.

Mixed Multinomial Logit (MIXL) Model

All attribute levels had statistically significant coefficients in the main analysis using a continuous variable of 477 rational respondents (Table 3). Unit increases in sensitivity and specificity produced significant positive utility (β = 0.269 and 0.132, respectively), whereas unit increases in cost and time to result caused a significant loss of utility. (β = -0.154 and -0.174, respectively). We added her ASC to test A to assess propensity to choose between test A and test B regardless of attribute level (Table 3). In the MIXL model, which included all respondents (n = 555), there was evidence supporting a systematic tendency to choose between test A and test B (βtest A = 0.305, p< 0.001). In contrast, among rational responders (n = 477), there was no systematic tendency to choose test A (βtest A = -0.026, p = 0.758). This indicates that the rational responder was able to effectively trade-off between his two test alternatives, whereas the irrational responder was unable to do so, again suggesting that the latter groups were excluded from the main analysis.

Table 3 Mixed Multinomial Logit (MIXL) Model with Continuous Attribute Levels

Supplemental File 4: Table S4 shows the results of the MIXL model using the categorical variables of the rational responders. As observed in the ASM probit regression analysis, the coefficient size (β) of the highest attribute level relative to the reference level is associated with sensitivity (3.749), cost (-2.550), specificity (1.134), and time to result (-0.284).

“Minimum” and “Optimal” TPP

Figure 2 shows the superiority of RDT over RT-PCR at specificity of 90-95%, time to result of 20-60 minutes, cost of $1-20, and sensitivity of 70-100%. indicates the probability that is prioritized. Assuming a specificity of 95% and a time to result of 20 minutes (Figure 2A), the minimum acceptable thresholds for sensitivity that can be met by more than 70% of African healthcare workers are 82.5%, 85.0 %, 90.5%, and 93.5%. , where the tests cost $1, $5, $15, and $20, respectively. The optimally acceptable thresholds at which more than 90% of HCW could be met were 87.5%, 90.0%, 95.5% and 98.5%, respectively (Additional File 4: Table S5). Assuming a specificity of 90% and a time to result of 60 minutes (Fig. 2D), the minimum acceptable thresholds for diagnostic sensitivity were 85.5%, 88.0%, 93.5%, and 96.5%, respectively, which are optimally acceptable. The thresholds were 85.5%, 88.0%, 93.5% and 96.5% respectively. The probabilities were 91.5%, 93.0%, 98.5% and 100% for test costs of US$1, US$5, US$15 and US$20, respectively.

Figure 2
Figure 2

Probability of preferring RDT over RT-PCR. a. Probability of favoring RDT with 95% specificity and 20 min time to result. B.. Probability of preferring RDT with 95% specificity and 60 min time to result. C.. Probability of favoring RDT with 90% specificity and 20 minute time to result. D..Probability of preferring RDT with 90% specificity and 60 min time to result

Alternating current

Table 4 shows the interaction between participant characteristics and test attributes using ASM probit regression analysis. Compared with younger age groups, older participants lost more utility due to price increases and increased utility less due to increased susceptibility. With respect to occupation type, public health workers (β = -0.101) lost more utility due to cost increases than physicians (-0.067) and midwives (-0.043). In contrast, increased usefulness due to increased sensitivity was higher for physicians (0.129) than for nurses (0.074) and public health workers (0.070). When he extended the time to result from 20 to 60 minutes, midwives’ usefulness decreased significantly (-0.505), but not so much for physicians (-0.011). Heterogeneity in preferences was also observed across workplaces and across African subregions.

Table 4 Interactions between unit increases in test attributes and participant characteristics of rational respondents (n= 477)



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *