Figure 1

Figure 1. Study population flowchart for study quantify population burden and effectiveness of decentralized surveillance strategies for skin-presenting neglected tropical diseases, Maryland County, Liberia. Consort diagram shows selection, screening, quality control, and...

We visited 10,007 households across 92 clusters (143 refused, 1.4%) and included 56,825 persons (49.8% female, 47.3% <18 years of age) in the sample population (Figure 1). In total, 34,916 persons were present during CHW household screening visits to observe photographs of skin NTDs. The remaining 38.6% were absent at the time of survey, and referrals among this group were based on proxy responses.

Among the sample population, 3,087 persons (5.4%, 95% CI 5.2–5.6) were referred by CHWs because these persons had possible skin NTD symptoms. Median age of referrals was 27 years (35.7% female, increasing to 48.3% when excluding hydrocoele; 102 missing age or sex data). We observed a linear increase in referral rates by age (p<0.0001), with an approximate threshold at 35 years, over which referrals increased more than 2-fold from 4.1/100 persons screened (95% CI 3.9–4.3) to 8.6/100 persons screened (95% CI 8.1–9.1). CHW referral rates varied substantially by cluster (range 0.5–23.0/100 persons screened; ICC 0.11) and health district (3.1–7.0/100 persons screened; ICC 0.01). Models exploring associations between referral rates and potential operationally relevant variables indicated only older CHW age to be associated with reduced odds of referral (>35 years of age; odds ratio 0.59, 95% CI 0.43–0.81; p = 0.001) (Appendix).

Figure 2

Figure 2. Cluster-level prevalence of all skin-presenting neglected tropical diseases combined, Maryland County, Liberia, June‒October 2018. Inset boxes show major urban areas Pleebo (A) and Harper (B). Black features are buildings (OpenStreetMap...

Mobile verification teams successfully followed up with 2,630 case-patients (81.1% of those referred). This group had minor differences in age compared with those who could not be found for follow-up (27.7 years [95% CI 26.1–29.3 years] vs. 30.3 years [95% CI 29.4–31.1]) but no overt difference in sex (35.0% [95% CI 31.0%–39.1%] female patients followed up vs. 36.7% [95% CI 34.8%–38.6%] female patients not followed up) or implementation district of residence (p = 0.15). We diagnosed 236 cases of skin NTDs (Table 1), a crude prevalence of 41.5 skin NTDs/10,000 persons and a design-adjusted prevalence of 34.0 (95% CI 29.1–38.9) skin NTDs/10,000 persons (Figure 2). The most prevalent condition was LFM, causing 111 lymphedema (17.5 [95% CI 14.1–21.0] cases/10,000 persons) and 58 hydrocoele cases (8.5 [95% CI 4.8–12.3] cases/10,000 persons). We identified 55 cases of suspected Buruli ulcer on the basis of clinical case definitions, although only 4 were confirmed by PCR (0.9 [95% CI 0–1.9] cases/10,000 persons), establishing PCR-confirmed Buruli ulcer as the rarest outcome (Appendix).

Figure 3

Figure 3. Spatial distribution and occurrence of skin-presenting neglected tropical diseases , Maryland County, Liberia, June‒October 2018. A) Buruli ulcer, B) leprosy, C) lymphatic filariasis morbidity; D) yaws. Points represent cluster centroids...

Figure 4

Figure 4. Cluster-level prevalence of skin-presenting neglected tropical diseases, Maryland County, Liberia, June‒October 2018. Colors denote health district of cluster.

Prevalence of any skin NTD was focally distributed within communities (ICC 0.27), with considerable heterogeneity between clusters (range 0–330 case/10,000 persons) (Figures 3,4). Analysis of individual skin NTDs showed a greater degree of spatial heterogeneity, with LFM and yaws exhibiting particularly focal distributions (Table 1). Few clusters were co-endemic for more than 1 skin NTD (22 of 92, 23.9%) and only 1 cluster was co-endemic for >2 diseases. Of potential cases identified in screening, 91.0% (2,394/2,630) were diagnosed with conditions not included within the primary outcome, including superficial fungal infections (471 cases, 17.9% of verified cases), scabies (316 cases, 12.0%), scrotal hernia (279 cases, 10.6%) and skin ulcers of unknown etiology (110 cases, 4.2%) (Appendix).

Figure 5

Figure 5. Comparison of cases of skin-presenting neglected tropical diseases before and after survey Maryland County, Liberia. A) Survey cases previously unknown to the health system; B) annual new case detection rates...

The new case detection rate from existing county-level health records in 2017 was 13.8 cases/10,000 persons compared with our survey point prevalence estimate of 25.4 (95% CI 21.3–29.5) previously unidentified cases/10,000 persons (Figure 5). Overall, there was no evidence of differences in age and sex of case-patients detected through routine reporting. Among leprosy case-patients only, those we detected by using survey methods were older (46.3 vs. 35.2 years; p = 0.02), and there was a greater proportion of paucibacillary leprosy relative to routine data (53.8% vs. 22.9%; p = 0.006).

To assess performance of CHW screening, QC surveys were conducted in 1,382 randomly sampled households (1,379 consented, 99.8%) in 91 clusters before verification of cases took place. Among the QC sample population, 95.1% of households (1,320 of 1,379) reported being visited by the local CHW and shown skin NTD photographs, with no evidence of socioeconomic disparities between households visited or missed (Appendix).

QC teams enumerated 8,021 persons and performed skin examinations on 4,268 household members (53.2%) among 4,409 approached (141 refused, 3.2%). Among persons examined, clinical field officers (trained CHSS cohort) identified 503 cases (11.8 [95% CI 10.8–12.8] cases/100 examined) of skin lesions similar in appearance to photograph-based CHW case definitions. Among the 503 patients who had skin lesions, clinical field officers recaptured patient identification cards from 349 to estimate sensitivity of screening (69.4%; CHSS new case detection rate 3.6 [95% CI 3.1–4.2] cases/100 persons). There was good concordance with CHW referrals for age and proportion of female referrals. We also conducted a sensitivity analysis of the effect of reduced sensitivity on prevalence estimates (Appendix).

We assessed the reliability of clinical diagnoses made by verification teams through separate follow-up QC surveys immediately after case verification activities. We reached 174 of 2,630 verified cases (6.6%) across 16 health facilities and 36 clusters. The crude agreement of all 174 diagnoses as skin NTD was 82.8% (Cohen κ 0.69, 95% CI 0.59–0.79), indicating substantial agreement between raters. Excluding other skin diseases, crude agreement (62.0%) and Cohen κ (0.51, 95% CI 0.39–0.64) were lower for skin NTDs, with a tendency for overdiagnosis among verification teams (Table 2). For individual skin NTDs, we did not estimate Cohen κ because of high prevalence index introduced by our sampling approach (20), but crude agreement between raters showed considerable variation between diseases (Table 2).

This study was a programmatic-scale integrated skin NTD prevalence survey in West Africa and was conducted entirely within the routine health infrastructure of Maryland County, Liberia. Our results show that skin NTDs in this setting are underreported, spatially heterogeneous, and highly focal, imparting a considerable unmet burden on this largely rural and periurban population.

We concurrently provide new evidence on the effectiveness of surveillance strategies that form the basis of skin NTD program delivery outlined in the WHO 2030 NTD roadmap (21). We demonstrate that large-scale screening by CHWs can find unreported cases of stigmatizing diseases while achieving high and equitable coverage among hard-to-reach communities. We also quantified major limitations in sensitivity and specificity from using our chosen approach with this workforce. Integrated clinical training of nonphysician healthcare workers facilitated reliable differentiation between any skin NTD and other skin conditions reported by participants. However, reliability of disease-specific diagnoses of skin NTDs was variable.

The greatest disease burden in Maryland County was attributable to LFM, with both BU and yaws showed markedly lower prevalence. Burden across all skin NTDs was higher than reported through routine surveillance systems for the county, as well as those typically reported in surveillance records nationally and across other co-endemic states in West Africa, although Buruli ulcer remains comparable if limited to PCR-confirmed cases (5,7,22). All diseases appeared spatially heterogeneous in occurrence and prevalence at this implementation scale. The explanatory factors underlying these observations are probably multifaceted, given diverse transmission dynamics, a combination of climatic, ecologic, and sociodemographic (23–26). However, given highly focal distributions, these observations could be attributable to sampling error.

Population-level skin NTD surveys have previously been undertaken in Ethiopia, Rwanda, and Cameroon (12,27–28), demonstrating a similarly high unmet burden. We believe the additional granularity and operational evaluation in our study provides additional strong justification for integrated approaches to skin NTD surveillance. We demonstrated that at the cluster level, most communities did not have individual skin NTDs, resulting in wasted resources if using nonintegrated surveillance strategies. Although findings indicate that disease-specific interventions could be targeted to smaller implementation units, sampling effort required for accurate delineation might outweigh benefits of microplanning.

The use of CHWs for disease-specific surveillance is common in West Africa, particularly for Buruli ulcer, for which increased case numbers or earlier stages of detection have been reported in quasiexperimental studies (13,29,30). Our findings illustrate the feasibility of training a rural community-based workforce with limited smartphone experience to screen for multiple diseases, reliably capture electronic data, and achieve high and equitable population coverage. CHWs identified a large proportion of previously undetected cases, even in a setting with recent previous training of CHWs and formal health workers. We also found no evidence households missing during screening were systematically omitted on the basis of socioeconomic indicators. However, we observed and quantified the probable underestimation of referable skin lesion burden by using our chosen approach. In addition, 91% of persons with verified cases were ultimately diagnosed with non–skin NTD etiology, including a large number of communicable skin diseases (corroborating recent dermatologic surveys in neighboring Côte d’Ivoire [(31)]), debilitating ulcers, and scrotal hernias. Given widespread use of CHWs for skin NTD surveillance, our results quantify major considerations with this approach, including management of a potentially large additional burden of disease.

Sustainable skin NTD programs also depend on decentralized diagnosis and case management by mid-level health workers. The performance of integrated skin NTD training programs has not been formally evaluated, despite recent WHO publication of a manual for frontline healthcare workers (16). Our findings show that a tailored training program reliably identified skin NTDs but that agreement on specific diagnoses could be inconsistent, particularly in the case of hydrocele and leprosy. Furthermore, confirmation rates of clinically suspected Buruli ulcer and yaws highlight the need for laboratory support for diagnosis. Previous studies in West Africa showed success in developing clinical algorithms for common skin diseases (32), and research continues on alternative algorithmic or telemedicine approaches to support decentralized clinical decision-making (33,34). Our findings support the need for further evaluation of integrated training programs to support frontline healthcare workers, especially in settings in which laboratory support is limited.

The first limitation of our study was that we relied on CHWs to conduct screening, a strategy that might have led to us miss the most marginalized households at higher risk for skin NTDs. Nevertheless, our QC survey suggested high coverage, a finding also supported when cross-comparing household global positioning system points with satellite imagery. Second, screening relied upon self-report and proxy-report of stigmatized conditions. We quantified a degree of loss in sensitivity through QC skin examinations, but inclusion might have been further biased downwards if affected persons were less willing to participate. Third, ascertainment of leprosy and LFM was dependent on clinical diagnosis, with variable reliability potentially biasing estimates from true population prevalence. Finally, we observed a notable percentage (≈19%) of patients who did not continue follow-up between screening and verification stages, although we did not see overt differences in the demographics of the censored population. We would expect this aspect to bias final prevalence estimates down, but the magnitude of this effect remains unclear.

With the new WHO 2030 NTD roadmap explicitly mapping out a 10-fold scale-up of skin NTD programs, there is an urgent need to better clarify disease burden and strategies for integrated surveillance to support this global transition (8). Our results provide a multifaceted overview of disease epidemiology and operational evaluation of surveillance strategies that can guide countries who are beginning integrated skin NTD programs.