Willingness to Pay for Irrigation Services in the Cold Winter Deserts of Uzbekistan

Irrigation facilities in the cold winter deserts (CWDs) of Uzbekistan are very traditional and poorly managed, resulting in low water use efficiency and low productivity. Improving the irrigation facilities in these deserts is a key priority for the country. This study intended to contribute towards the development of the irrigation systems through identification and quantification of the relative implicit values smallholder farmers confer to the key characteristics of irrigation facilities. We elicited preferences with discrete choice experiments, estimated willingness to pay for these attributes using random parameters logit models, and analyzed heuristics in the choice process using a series of latent class models. Our results show that farmers have clear preferences for higher watering frequency and no interest in sharing irrigation water with downstream users. We also observed that there are distinct groups of farmers with comparable but different levels of preference. The development of irrigation facilities in the water-scarce parts of Uzbekistan would benefit from careful consideration of the preferences of the target communities and targeting of the schemes based on the broad heterogeneities within the communities. This will aid in the maintenance of irrigation systems and, as a result, increase agricultural production and productivity.

Tree-informed Bayesian multi-source domain adaptation: cross-population probabilistic cause-of-death assignment using verbal autopsy

Determining causes of deaths (COD) occurred outside of civil registration and vital statistics systems is challenging. A technique called verbal autopsy (VA) is widely adopted to gather information on deaths in practice. A VA consists of interviewing relatives of a deceased person about symptoms of the deceased in the period leading to the death, often resulting in multivariate binary responses. While statistical methods have been devised for estimating the cause-specific mortality fractions (CSMFs) for a study population, continued expansion of VA to new populations (or "domains") necessitates approaches that recognize between-domain differences while capitalizing on potential similarities. In this paper, we propose such a domain-adaptive method that integrates external between-domain similarity information encoded by a pre-specified rooted weighted tree. Given a cause, we use latent class models to characterize the conditional distributions of the responses that may vary by domain. We specify a logistic stick-breaking Gaussian diffusion process prior along the tree for class mixing weights with node-specific spike-and-slab priors to pool information between the domains in a data-driven way. Posterior inference is conducted via a scalable variational Bayes algorithm. Simulation studies show that the domain adaptation enabled by the proposed method improves CSMF estimation and individual COD assignment. We also illustrate and evaluate the method using a validation data set. The paper concludes with a discussion on limitations and future directions.

OP340 Kidney Patients’ Preferences For A Wearable Digital Health Technology To Support Self-Management Of Chronic Kidney Disease - A Discrete Choice Experiment

IntroductionWearable Digital Health Technologies (WDHTs) can support and enhance self-management by giving individuals with chronic conditions more control over their health, safety and wellbeing. Involving patients early on in the design of these technologies facilitates the development of person-centered products. It may increase the potential uptake of (and adherence to) any intervention they are designed to deliver. This research aims to elicit chronic kidney disease (CKD) patients’ preferences for WDHTs that may help patients manage their conditions.MethodsWe used discrete choice experiments (DCE) to elicit preferences for WDHTs characterized by their generalizable characteristics. The study design was informed by a multi-stage mixed-method approach (MSMMA). This included a review of the published literature, focus group interviews and one-to-one interactions with CKD patients to identify relevant characteristics (that is, attributes and levels) associated with wearable DHTs. We collected the data from 113 patients (age ≥18 years) with stage 3 or above CKD. The analysis started with a conventional multinomial logit model and was extended by investigating heterogeneity in preferences via latent class models.ResultsOur MSMMA yielded ten potential attributes for consideration in a choice task. The final list included five attributes, cross-checked and validated by the research team, and patient representatives. The most preferred attributes of WDHTs were device appearance, format and type of information provided, and mode of engagement with patients. Respondents preferred a discreet device, which offered options that individuals could choose from and provided medical information.ConclusionsWe show how to use MSMMA to elicit user preferences in (and to inform the) early stages of the development of WDHTs. Individuals with CKD preferred specific characteristics that would make them more likely to engage with the self-management support WDHT. Our results provide valuable insights that can be used to inform the development of different WDHTs for different segments of the CKD patients population, moving away from a one-size-fits-all provision and resulting in population health gains.

Latent Classes of First Incident Difficulty in Instrumental Activities of Daily Living

Ability to perform instrumental activities of daily living (IADLs) deteriorates during prodromal Alzheimer’s disease (AD), eventually leading to impaired everyday functioning and dementia. Ordering and timing of IADL difficulty onset may identify individuals at greater risk of cognitive impairment, but most studies only consider total number of difficult tasks. Leveraging longitudinal data from the Advanced Cognitive Training in Independent and Vital Elderly (ACTIVE) Study who entered free of any IADL difficulty (N=1266), we hypothesized that a latent class analysis based on timing of first reported IADL task difficulty would reveal class differences in cognitive functioning . Participants were followed until they self-reported at least one IADL difficulty, study completion (10 years), or loss to follow-up. Discrete-time multiple event process survival mixture (MEPSUM) models were used to simultaneously estimate hazards of incident IADL task difficulty across 7 task groups. Two, 3, 4, and 5 latent class models were fit to the data. Both unadjusted and covariate-adjusted models (adjusted for age, sex, race, education, marital status, and general health rating) were fit. Using the 2-class solution as the most parsimonious model, model entropy was 0.855. The model was able to distinguish a class of participants with lower global cognitive factor scores at baseline (Cohen’s D = 0.23, P = 0.04). We conclude that first incident IADL difficulty may be a useful measure in identifying individuals with worse cognitive functioning.

Returning to work after a sickness absence due to cancer: a cohort study of salaried workers in Catalonia (Spain)

Cancer incidence and survival rates have increased in the last decades and as a result, the number of working age people diagnosed with cancer who return to work. In this study the probability of accumulating days of employment and employment participation trajectories (EPTs) in a sample of salaried workers in Catalonia (Spain) who had a sickness absence (SA) due to cancer were compared to salaried workers with SA due to other diagnoses or without SA. Each individual with SA due to cancer between 2012 and 2015 was matched by age, sex, and onset of time at risk to a worker with SA due to other diagnoses and another worker without SA. Accumulated days of employment were measured, and negative binomial models were applied to assess differences between comparison groups. Latent class models were applied to identify EPTs and multinomial regression models to analyse the probability of belonging to one EPT of each group. Men and women without SA or with SA due to other diagnoses had at least a 9% higher probability of continuing in employment compared to workers who had a SA due to cancer, especially among men without SA (adjusted IRR 1.27, 95% CI 1.06‒1.53). Men without SA had the highest probability of having high stable EPT compared to workers who had a SA due to cancer (adjusted RRR 3.21, 95% CI 1.87‒5.50). Even though workers with SA due to cancer continue working afterwards, they do it less often than matched controls and with a less stable employment trajectory. Health and social protection systems should guaranty cancer survivors the opportunity to continue voluntary participation in the labour market.

Improving the diagnosis of severe malaria in African children using platelet counts and plasma PfHRP2 concentrations

Background Severe falciparum malaria is difficult to diagnose accurately in children in high transmission settings. Platelet counts and plasma concentrations of P. falciparum histidine-rich protein-2 (PfHRP2) are potential biomarkers to increase diagnostic accuracy. Methods We fitted Bayesian latent class models to platelet counts and PfHRP2 concentrations in 2,649 patients enrolled in four studies of severe illness in three countries (Bangladesh, Kenya, and Uganda). We estimated receiver operating characteristic curves and compared parasite densities, haematocrits, total white blood cell counts, blood culture positivity rates, and haemoglobin S genotypes (HbAS and HbSS) across the subgroups defined by the probabilistic models. Findings The platelet count and the plasma PfHRP2 concentration have substantial diagnostic value in severe malaria. In severely ill patients with clinical features consistent with severe malaria, a combined platelet count ≤150,000 per μL and a plasma PfHRP2 concentration ≥1,000 ng/mL had an estimated sensitivity of 74\% and specificity of 93\% in identifying `true' severe falciparum malaria. We estimate one third of African children enrolled in the two clinical studies of severe malaria had another cause of severe illness. Under the model, patients with severe malaria had higher parasite densities, lower haematocrits, lower rates of invasive bacterial disease, and a lower prevalence of both HbAS and HbSS than children misdiagnosed. Mortality in `true' severe malaria was consistent across the African sites at ≈10%. Interpretation Studies of severe falciparum malaria in African children would be improved by including only patients with platelet counts ≤150,000 per μL and plasma PfHRP2 concentrations ≥1,000 ng/mL. Funding Wellcome

Key population size estimation in Nigeria: applying novel Bayesian methods for analysis of three-source capture-recapture data (Preprint)

BACKGROUND Nigeria has the fourth largest burden of HIV globally. Key populations (KP) including female sex workers (FSW), men who have sex with men (MSM), and people who inject drugs (PWID) often have poor social visibility and are more vulnerable to HIV than the general population due to stigma, discrimination, and criminalization of KP-defining behaviors. Reliable, empirical population size estimates (PSE) are needed to guide focused and appropriately scaled HIV epidemic response efforts for KP. We used novel approaches to sampling and analysis to calculate PSE in Nigeria. OBJECTIVE We sampled the population using three-source capture-recapture (3S-CRC) and analyzed results using Bayesian nonparametric latent-class models to generate median PSE with 80% highest density intervals. METHODS During October–December 2018, we used three-source capture-recapture (3S-CRC) to estimate the size of KP in seven United States President’s Emergency Plan for AIDS Relief (PEPFAR) priority states in Nigeria. Hotspots were mapped before 3S-CRC started. We sampled FSW, MSM, and PWID during three independent captures approximately one week apart. During encounters in KP hotspots, distributors offered inexpensive and memorable objects to KP, unique to each capture round and KP type. In subsequent rounds, participants were offered an object and asked to produce or identify objects received during previous rounds (if any); affirmative responses were tallied upon producing or identifying the correct object. Distributors recorded responses on tablets and uploaded to a secure server after each encounter. Data were aggregated by KP and state for analysis. Median PSE were derived using Bayesian nonparametric latent-class models with 80% highest density intervals for precision. RESULTS We sampled approximately 310,000 persons at 9,015 hotspots during three independent captures in all seven states. Overall, FSW PSE ranged from 14,500-64,300; MSM PSE, 3,200-41,400; and PWID PSE, 3,400-30,400. CONCLUSIONS This study represents the first implementation of these 3S-CRC sampling and novel analysis methods for large-scale population size estimation in Nigeria. Overall, our estimates were larger than previously documented for each KP in all states. The current Bayesian models account for factors (i.e., social visibility and stigma) that influence heterogeneous capture probabilities resulting in more reliable PSE. The larger estimates suggest a need for programmatic scale-up to reach these populations at highest risk for HIV.

Latent Class Analysis of Obesogenic Behaviors among Korean Adolescents: Associations with Weight-Related Outcomes

This study aimed to explore sex-specific latent class models of adolescent obesogenic behaviors (OBs), predictors of latent class membership (LCM), and associations between LCM and weight-related outcomes (i.e., weight status and unhealthy weight control behaviors). We analyzed nationally representative data from the 2019 Korea Youth Risk Behavior Survey. To identify latent classes for boys (n = 29,841) and girls (n = 27,462), we conducted a multiple-group latent class analysis using eight OBs (e.g., breakfast skipping, physical activity, and tobacco product use). Moreover, we performed a multinomial logistic regression analysis and a three-step method to examine associations of LCM with predictors and weight-related outcomes. Among both sexes, the 3-class models best fit the data: (a) mostly healthy behavior class, (b) poor dietary habits and high Internet use class, and (c) poor dietary habits and substance use class. School year, residential area, academic performance, and psychological status predicted the LCM for both sexes. In addition, perceived economic status predicted the LCM for girls. The distribution of weight-related outcomes differed across sex-specific classes. Our findings highlight the importance of developing obesity prevention and treatment interventions tailored to each homogeneous pattern of adolescent OBs, considering differences in their associations with predictors and weight-related outcomes.