This study was carried out in a low-income urban area of Mandalay City, which is densely populated and located in the central part of Myanmar [4-6]. Mandalay City has 7 township health departments providing TB services. Free medical services for TB diagnosis, including CXR and treatment, are available at TB health centers of the township health departments [7]. Routine TB care services are provided with precautions of COVID-19 prevention and control measures. For this study, 1 urban slum area was randomly selected from the list from each of 3 randomly selected townships by township medical officers and the regional TB coordinator.

The “TB-screen” app was developed by AppSheet's no-code app (Seattle, WA, USA) [8]. It was a collaborative effort between the Department of Medical Research, National TB Control Programme, Myanmar, and the Department of Epidemiology, Prince of Songkla University, Hat Yai, Thailand.

The app was built on the principal investigator’s computer directly from Google Sheet. It was designed to be used off-line on a mobile device running an Android operating system (Google, Inc., Mountain View, CA, USA). The app was installed in the mobile phones of health care providers who are responsible for TB screening. They were carefully trained on how to use the mHealth app. They invited a family household with their neighbors to join TB screening. After entering sociodemographic and TB clinical variables of an individual via the app, the TB risk propensity score is computed using the following formula [9]:

where B is a vector of regression coefficients and X is the matrix of sociodemographic and TB clinical covariates of the subject. Details of the variables and their related coefficients are shown in Multimedia Appendix 1. It was noted that the coefficient of the elderly age groups and male groups were higher than all other variables. The oldest age group (55+ years) had an e^2.4=11.0 times higher risk of developing TB than younger ones. The male group had an e^1.0=2.07 times higher risk of developing TB than the female group. These groups had high potential to be identified as presumptive TB cases via the mobile app. The TB risk propensity score is the value that measures the risk of developing TB [2].

With a selected propensity score cut-off level at ≥0.0052 (≥0.5% probability to develop TB), the test was taken as positive. The sensitivity is 80.6%, and the specificity is 63.5% [2]. A person with a positive test result, the presumptive TB case, is recommended to visit a TB health center for CXR examination. If they agree to go to the CXR center, a specific quick response (QR) code with a CXR referral form is scanned to the mHealth app (Figure 1). The presumptive TB case is given the CXR referral form to take to the CRX center. When the app is connected to the internet, the information collected is uploaded to the main server of data storage, with a notification to the corresponding TB health center. The data are updated again by the CXR center by scanning the referral QR code when the person visits and undergoes CXR examination. Figure 1 shows a screenshot of the TB-screen app.

The study was approved by the Institutional Ethics Committee of the Faculty of Medicine, Prince of Songkla University, Hat Yai, Thailand (REC:63-074-18-1), and the Institutional Review Board of the Department of Medical Research, Myanmar (IRB00008835).

The screening process was carried out during November 2020-January 2021. This was after the COVID-19 outbreak substantially subsided and lockdowns and travel restrictions were eased. A trained local health care provider who is responsible for providing both TB and COVID-19 care in the assigned community performed TB screening via the app, following COVID-19 prevention and control measures. For each community, a family household with neighbors was approached for TB screening. Only 4 or 5 family households were approached, and at most 5-6 presumptive TB cases were referred for CXR examination per day to avoid overloading at the CXR center. The health care providers had been vaccinated with the complete dose of the COVID-19 vaccine. The presumptive TB cases were provided with face masks and face shields to be used during the interview and when they went to the CXR center. The presumptive TB cases with suspected COVID-19 were referred to the community fever clinic center at the respective township health department for COVID-19 investigation. Those who tested negative for COVID-19 proceeded to the CXR center for CXR examination.

At the CXR center, there was a separate waiting room for the presumptive TB cases from the same community, with the chairs arranged with social distancing measures. One CXR technician was assigned for the CXR examination during the project. After CXR examination was performed, the presumptive TB cases returned home without waiting for CXR results. They were informed about the CXR results by the health care provider who performed TB screening. The presumptive TB cases needed a single visit for CXR examination, and their remaining diagnostic and treatment procedures were aided by the health care provider and local health volunteers. The presumptive TB cases with abnormal suggestion of TB on CXR were confirmed as active TB cases by performing Gene Xpert Mycobacterium tuberculosis complex and resistance to rifampin (MTB/RIF) examination.

During screening, informed consent for those >18 years old or guardian’s assent for those 15-18 years old was obtained for each participant. The screened participants were informed about survey processes using the app, and they were interviewed about their background, sociodemographic, and TB clinical characteristics; perceived susceptibility to developing TB; perceived benefits of TB screening; and perceived harms of TB screening. After collecting the variables related to the TB risk propensity score via the app, information about usability of the mHealth app was collected. Point coordinate data for location of participants were taken by the app for further use to calculate the distance between their residences and the TB health center for CXR. Accessibility methods to the CXR center were also recorded.

Figure 3 summarizes the overall results of subject classification based on the TB risk propensity score among the screened participants and compliance to undergo CXR examination among the presumptive TB cases. Of 694 participants approached, 631 (90.9%) agreed to be screened via the app. Among those screened, 453 (71.8%) were identified as presumptive TB cases by the app and were suggested to visit a TB health center for a CXR appointment.

As shown in Figure 3, of 453 presumptive TB cases, 322 (71.1%) complied to undergo CXR examination within 1-7 days of TB screening.

Table 1 shows the background characteristics of the 453 presumptive TB cases. Males and the elderly had a high proportion, accounting for 251 (55.4%) and 150 (33.1%) participants, respectively. The mean (SD) age was 46.1 (15.0) years. The median TB risk propensity score was 0.01 (IQR 0.0058-0.022). A high risk score indicated an increase in the risk of developing TB. Over half (n=258, 57.0%) had no TB signs or symptoms. Knowledge of TB was high, with a mean (SD) score of 5.9 (1.4) out of 8. The cases had less perceived susceptibility to developing TB, with a mean (SD) score of 2.7 (1.0) out of 5, but more perceived harms of TB screening, with a mean (SD) score of 3.0 (1.1) out of 5.

Table 2 shows the usability scores of the mHealth app rated by the presumptive TB cases. All items had an average rating score of >3. The overall rating on the usability of the mHealth app was favorable, with a mean (SD) score of 4.1 (1.1) out of 5, indicating a high level of usability rated by the users.

In Table 3, the final CFA showed a good fit after removing 3 items from the usability of the app (U5, U6, and U7) and 2 items from the perceived harms of TB screening (R3 and R4) with factor loadings of <0.3 in the baseline CFA. Each latent variable was rated using a 5-point Likert scale ranging from 1 for “totally disagree” to 5 for “totally agree.”

Table 4 shows that the composite reliability was higher than the average variance extracted and the average variance extracted was >0.5. In addition, the square root of the average variance extracted was higher than the correlations of the latent variables under analysis. Thus, the latent variables had both good consistency and discriminant validity.

As shown in Table 5, the final SEM model was acceptable (TLI>0.95) and showed a better fit than the baseline proposed model (CFI=0.955 vs 0.922).

Figure 4 shows the standardized coefficients (β) of variables in the structural analysis of the causal pathway in the final SEM.

In the final SEM, the usability of the mHealth app and compliance to undergo CXR examination were not associated with marital status, education, perceived susceptibility to developing TB, perceived benefits of TB screening, and variables of accessibility methods to the CXR center. For clarity, these variables are not shown in Figure 4.

The usability of the mHealth app was significantly associated with compliance to undergo CXR examination. As β=.464, 1 standard deviation increase in the usability score of the app would be associated with e^0.464=1.59-fold odds or 59% increase to comply to undergo CXR examination.

The other significant factors associated with compliance were a high TB risk propensity score and high TB knowledge. In contrast, those with a lack of TB signs and symptoms and with perceived harms of TB screening were less likely to comply to undergo CXR examination.

Similarly, having high TB knowledge favored the usability of the app, while a lack of TB signs and symptoms and having perceived harms of TB screening predisposed individuals to be less favorable toward the usability of the app.

Of 453 presumptive TB cases detected by the app, 322 (71.1%) complied to undergo CXR examination. Of them, 56 (17.4%) were identified as showing an abnormal suggestion of TB in the CXR and continued for Gene Xpert MTB/RIF examination. The total active TB case detection was 24 (7.5%) of 322 presumptive TB cases.

This is the first study in Myanmar using a mobile app for screening for TB and assessing its impact on seeking TB health care services. This mHealth app was rated by community members to have a good usability score. A significant proportion of the presumptive TB cases were identified via the app. Nearly three-quarters of the presumptive TB cases identified by the app proceeded to undergo CXR examination, and 7.5% showing CXR compliance were diagnosed as active TB cases. The usability of the mHealth app was the strongest predictor associated with the compliance to undergo CXR examination. Low predicting variables included TB signs and symptoms, perceived harms of TB screening, the TB risk propensity score, and knowledge of TB.

In this study, a high number of presumptive TB cases were identified via the mHealth app and TB health centers were notified. This could be explained by the high percentages (55.4% and 33.1%) of subjects under screening being male and elderly, respectively. The setting of the app gave a high score to these groups. In addition, nearly three-quarters of the presumptive TB cases identified by the app complied to undergo CXR examination. It was relatively high compared to a previous study in Myanmar [18]. Eventually, 24 (7.5%) of 322 users who were suspected by the app as having TB were detected as active TB cases by CXR. This proportion was missing active TB cases in the community that were early detected by the mobile app. However, this achievement would not be possible without the compliance to undergo CXR examination by the presumptive TB cases detected by the app. Completion of this step would lead to early detection of TB cases, achieving the WHO-recommended “universal access to test and treat process” [19].

During implementation of the app, the participants rated the usability of the mHealth app highly and a significant proportion of presumptive TB cases were detected via the app. Many studies have revealed patients’ acceptance of using mHealth apps in TB care to improve health outcomes [20,21]. In addition, the favorable rating for the usability of the mHealth app had a positive association with compliance to undergo CXR examination. In this study, the mobile app for TB screening was integrated into a pathway to TB diagnosis by notifying the CXR center of screened positive cases. This integration might be the reason for the high association of these two variables. A similar improvement in TB referrals by engaging the presumptive TB cases in the health sector via the app was shown in a study in India [22].

In this study, presumptive TB cases without TB signs or symptoms and those who had perceived harms of TB screening gave a less favorable rating to the usability of the app and had poor compliance to undergo CXR examination. A review of national TB prevalence surveys in Asia highlighted that those who screened negative for TB symptoms were less likely to seek TB care until symptoms worsened [23]. Many studies have revealed that having perceived harms of TB screening has a negative impact on TB screening and subsequent diagnostic procedures [24-26]. Having knowledge of TB can increase acceptance of patients with TB and encourage subsequent screening and diagnostic processes [27].

Compared to similar studies, this mHealth app calculated the TB risk propensity score instead of using TB signs and symptoms [22,28-30]. Although the TB risk propensity score has better TB prediction than TB signs and symptoms, there is a limitation in applying it in the real field, probably because of its complexity to calculate [31,32]. However, our mobile app overcomes this gap.

In addition, a review of a national TB prevalence survey in Asia (1990-2012) revealed that 40%-60% of active TB cases are missed by screening for routine TB signs and symptoms, because this proportion of patients are asymptomatic and not identified as presumptive TB cases [23]. In Myanmar, the TB case detection rate by screening for routine TB signs and symptoms was 69%-77% in 2017-2019 [33-35]. This highlighted that nearly one-fourth of active TB cases were missing, either not notified to the national TB program or not diagnosed and treated. Therefore, using a mobile app that calculates the TB risk score is a potential new approach to identifying missing presumptive TB and active TB cases that would not be detected by TB signs and symptoms only.

As strengths, the mobile app was used to calculate the TB risk propensity score so misclassification bias to identify the TB suspects was reduced. Real-time data entry was performed for the CXR examination date so that data error was less likely in assessing compliance. The app is simple to develop, provided that existing TB survey databases can be used to calculate the TB risk propensity score for the population under study. This idea can be adapted to many low-resource countries.

However, a high level of compliance to undergo CXR examination in this study was probably due to worry about long COVID among the population after a serious COVID-19 outbreak recently subsided [36]. In addition to TB detection, they might expect CXR to detect the residual effects of COVID-19 and thus they had high compliance. As a limitation, high compliance to undergo CXR examination may suggest that the TB screening program could be done better immediately after a COVID-19 outbreak or its generalizability should be confirmed when the COVID-19 situation is no more perceived as a serious threat.

The simple mobile app we developed got a high usability score by 453 users. The mobile app usability score successfully predicted compliance to undergo CXR examination. Eventually, 24 of 322 users who were suspected by the mobile app as having TB were detected as active TB cases by CXR. The system should be upscaled for a large trial.