Effectiveness of an mHealth system on access to eye health services in Kenya: a cluster-randomised controlled trial.

BACKGROUND: There is limited access to eye health services in many low-income and middle-income populations. We aimed to assess the effectiveness in increasing service utilisation of the Peek Community Eye Health (Peek CEH) system, a smartphone-based referral system comprising decision support algorithms (Peek Community Screening app), SMS reminders, and real-time reporting. METHODS: In this cluster-randomised controlled trial of eye health in Kenya, community unit clusters were defined as one health centre and its catchment population. Clusters were randomly allocated (1:1) to receive Peek CEH and referral (intervention group) or standard care via periodic health centre-based outreach clinics and onward referral (control group). Individuals in the intervention group were assessed at home by screeners and those referred were asked to present for triage assessment in a central location. They received regular SMS reminders. In both groups, community sensitisation was done followed by a triage clinic at the cluster health centre 4 weeks after sensitisation. During triage, individuals in both groups were assessed and treated and, if necessary, referred to a specific hospital. Individuals in the intervention group received further SMS reminders. The primary outcome was the mean attendance rate (the number of people per 10 000 population) at triage of those with confirmed eye conditions, as assessed at 4 weeks after sensitisation in the intention-to-treat population. We estimated the intervention effect using a Student's t-test on cluster-level rates. This trial is registered with Pan African Clinical Trial Registry, number 201807329096632. FINDINGS: Between Nov 26, 2018, and June 7, 2019, of the 85 community units in Trans Nzoia County, Kenya, 49 were excluded. We randomly allocated 18 community units each to the intervention group (68 348 individuals) and the control group (60 243 individuals). 9387 individuals from the intervention group and 3070 from the control group attended triage assessment. The mean attendance rate at triage by individuals with eye problems was 1429 (92% CI 1228-1629) in the intervention group and 522 (418-625) in the control group (rate difference 906 per 10 000 [95% CI 689-1124; p<0·0001]). INTERPRETATION: The Peek CEH system increased primary care attendance by people with eye problems compared with standard approaches, indicating the potential of this mobile health package to increase service uptake and guide appropriate task sharing. FUNDING: The Queen Elizabeth Diamond Jubilee Trust and Wellcome Trust.

Smartphone-Guided Algorithms for Use by Community Volunteers to Screen and Refer People With Eye Problems in Trans Nzoia County, Kenya: Development and Validation Study.

BACKGROUND: The provision of eye care services is currently insufficient to meet the requirements of eye care. Many people remain unnecessarily visually impaired or at risk of becoming so because of treatable or preventable eye conditions. A lack of access and awareness of services is, in large part, a key barrier to handle this unmet need. OBJECTIVE: This study aimed to assess whether utilizing novel smartphone-based clinical algorithms can task-shift eye screening to community volunteers (CVs) to accurately identify and refer patients to primary eye care services. In particular, we developed the Peek Community Screening app and assessed its validity in making referral decisions for patients with eye problems. METHODS: We developed a smartphone-based clinical algorithm (the Peek Community Screening app) using age, distance vision, near vision, and pain as referral criteria. We then compared CVs' referral decisions using this app with those made by an experienced ophthalmic clinical officer (OCO), which was the reference standard. The same participants were assessed by a trained CV using the app and by an OCO using standard outreach equipment. The outcome was the proportion of all decisions that were correct when compared with that of the OCO. RESULTS: The required sensitivity and specificity for the Peek Community Screening app were achieved after seven iterations. In the seventh iteration, the OCO identified referable eye problems in 65.9% (378/574) of the participants. CVs correctly identified 344 of 378 (sensitivity 91.0%; 95% CI 87.7%-93.7%) of the cases and correctly identified 153 of 196 (specificity 78.1%; 95% CI 71.6%-83.6%) cases as not having a referable eye problem. The positive predictive value was 88.9% (95% CI 85.3%-91.8%), and the negative predictive value was 81.8% (95% CI 75.5%-87.1%). CONCLUSIONS: Development of such an algorithm is feasible; however, it requires considerable effort and resources. CVs can accurately use the Peek Community Screening app to identify and refer people with eye problems. An iterative design process is necessary to ensure validity in the local context.