Statistical Model Research Articles

Veto players and gun violence in drug markets: Analysis based on field observations of eighty drug-selling spots on the Westside of Chicago

PurposeThe study analyzed whether the number of drug-selling spots has an impact on shooting incidents in drug markets on the Westside of Chicago. MethodsA theory of gun violence in drug markets was developed by building on veto players theory, and tested using a series of two-way fixed-effects Poisson regressions. The main variable of interest—drug-selling spots—comes from an original dataset comprised of eighty drug-selling spots identified through covert field observations. Controls included: the average age and average number of drug spot workers at the drug-selling spots, concentrated disadvantage, population, police anti-drug operations, and drug demand. FindingsIn support of the theory, the main statistical models and robustness tests consistently found that the larger the number of drug-selling spots in a given drug market, the higher the number of shooting incidents over time. Only the control for population remained consistently statistically significant across model specifications. ConclusionThe study contributes to the drug market literature by offering a novel theory of drug market violence, as well as a replicable way for researchers to observe drug market activity without disturbing the natural behaviors of market participants. The operationalization of drug-selling spots as sub-factions of organized criminal groups also provides a novel approach to the study of drug market violence.

Patient feasibility as a novel approach for integrating IRT and LCA statistical models into patient-centric qualitative data—a pilot study

IntroductionClinical research increasingly recognizes the role and value of patient-centric data incorporation in trial design, aiming for more relevant, feasible, and engaging studies for participating patients. Despite recognition, research on analytical models regarding qualitative patient data analysis has been insufficient.AimThis pilot study aims to explore and demonstrate the analytical framework of the “patient feasibility” concept—a novel approach for integrating patient-centric data into clinical trial design using psychometric latent class analysis (LCA) and interval response theory (IRT) models.MethodsA qualitative survey was designed to capture the diverse experiences and attitudes of patients in an oncological indication. Results were subjected to content analysis and categorization as a preparatory phase of the study. The analytical phase further employed LCA and hybrid IRT models to discern distinct patient subgroups and characteristics related to patient feasibility.ResultsLCA identified three latent classes each with distinct characteristics pertaining to a latent trait defined as patient feasibility. Covariate analyses further highlighted subgroup behaviors. In addition, IRT analyses using the two-parameter logistic model, generalized partial credit model, and nominal response model highlighted further distinct characteristics of the studied group. The results provided insights into perceived treatment challenges, logistic challenges, and limiting factors regarding the standard of care therapy and clinical trial attitudes.

Zygosity Effects on Human Voice: Fundamental Frequency Analysis of Brazilian Twins' Speech.

Voice production can be influenced by interindividual variations related to genetic, physiological, behavioral, and several environmental factors. Here we examined the effect of zygosity on speaking fundamental frequency (F0) statistical descriptors. Our aims were: (1) to determine whether the genetic similarity between monozygotic (MZ) and dizygotic (DZ) twins affects F0 characteristics, and (2) to quantify the contribution of genetic factors to these characteristics. The study involved 79 same-sex twin pairs of Brazilian Portuguese speakers, comprising 65 MZ and 14 DZ twins, aged 18 to 66 years (31.7 ± 11.6 years), with 21 male and 58 female pairs. Participants were recorded while uttering a greeting phrase and the Brazilian Portuguese version of the 'Happy Birthday to You' song. Speech segments were analyzed using Praat free software, and F0 measures were automatically extracted in both Hertz and semitone scales. Statistical descriptors, including centrality, dispersion, and extreme values of F0 were examined, and the ACE model (i.e., total genetic effects, A; shared environmental influences, C; and nonshared environmental influences, E) was employed to estimate the additive effect;ts of monozygosity. As anticipated, we observed a zygosity effect on several F0 parameters, with more similarity between MZ twins compared to DZ twins. We discuss the genetic influences on F0 parameters and the absence of a monozygosity effect in two of them. Additionally, we briefly address potential biases associated with the selected measurement scale for statistical modeling. Finally, we explore the influence of genetic factors on F0 patterns, as well as environmental, life history and linguistic factors, particularly concerning F0 variation in speech.

Machine Learning and Statistical Models to Predict All-cause Mortality in Type 2 Diabetes: Results from the UK Biobank Study

AimsThis study aims to compare the performance of contemporary machine learning models with statistical models in predicting all-cause mortality in patients with type 2 diabetes mellitus and to develop a user-friendly mortality risk prediction tool. MethodsA prospective cohort study was conducted including 22,579 people with diabetes from the UK Biobank. Models evaluated include Cox proportional hazards, random survival forests (RSF), gradient boosting (GB) survival, DeepSurv, and DeepHit. ResultsOver a median follow-up period of 9 years, 2,665 patients died. Machine learning models outperformed the Cox model in the validation dataset, with C-index values of 0.72–0.73 vs. 0.71 for Cox (p < 0.01). Deep learning models, particularly DeepHit, demonstrated superior calibration and achieved lower Brier scores (0.09 vs. 0.10 for Cox, p < 0.05). An online prediction tool based on the DeepHit was developed for patient care: http://123.57.42.89:6006/. ConclusionsMachine learning models performed better than statistical models, highlighting the potential of machine learning techniques for predicting all-cause mortality risk and facilitating personalized healthcare management for individuals with diabetes.

Interpreting the removal behaviours of aniline and 3,4-diaminotoluene on aromatic carbon-rich magnetic polymer: Statistical modelling, DFT quantitative calculations and new perspectives

The widespread presence of harmful and stubborn amine organic contaminants in aquatic environments motivated the exploration of efficient technologies for alleviating their serious environmental crisis. Herein, the recyclable magnetic Schiff base polymer (Fe3O4-TENE) with satisfactory and promising capacity in treating aniline and 3,4-diaminotoluene as representative amine organic contaminants was fabricated via a simple in-situ polymerization method, which combined the magnetic separability advantage of magnetic oxides with the abundant aromatic carbon content properties of Schiff base polymer. Multiple characterization methods verified the effectiveness of the in-situ polycondensation coating reaction and analyzed the physicochemical properties of Fe3O4-TENE. The Fe3O4-TENE possessed abundant homogeneous reaction sites and excellent regeneration performance, multiple diffusion driving forces and micro-mechanisms co-participated in the removal reaction. The adsorption pathway, underlying reaction mechanism and the most probable bonding system were identified based on the statistical modelling and DFT quantitative calculations. Especially, the visible electrostatic potential molecular and molecular orbitals verified the π-π interaction and electrostatic forces greatly contributed to the removal reaction. The Driving forces – Challenges – Breakthrough – Application framework was proposed for comprehensively assessing the possible environmental application of Fe3O4-TENE. Overall, this work not only provides important insights into the utilization of magnetic polymer composite to alleviate ecological and environmental risks caused by amine organic pollutants, but also contributes to an in-depth excavation of the bonding types and removal mechanisms from both micro and macro perspectives.

Harnessing nano-synergy: A comprehensive study of thermophysical characteristics of silver, beryllium oxide, and silicon carbide in hybrid nanofluid formulations

Nanofluids, promising to improve heat transfer efficiency, encounter stability and durability challenges, hindering their industrial applications. The emerging concept of hybrid nanofluids, alongside surfactant-driven stability research, presents a promising solution to tackle challenges in heat transfer, potentially revolutionizing thermal management systems and advancing nanomaterial science. This comprehensive study investigates the thermophysical properties of simple and hybrid nanofluid formulations composed of silver (Ag), beryllium oxide (BeO), and silicon carbide (SiC) nanoparticles dispersed in water. Hybrid nanofluids were prepared with varying volumetric ratios of 20:80, 40:60, 60:40, and 80:20 and examined across temperatures ranging from 15 to 45 °C. The influence of surfactants on stability was explored to augment thermal characteristics. Results revealed that the surfactants have a significant effect on stability and the specific mixing ratios can lead to more favourable thermal characteristics. Thermal conductivity enhancements, evaluated via the transient hot-wire method, demonstrated improvements of 7.43 %, 7.17 %, and 5.31 % for Ag/SiC (60:40), Ag/BeO (60:40), and SiC/BeO (80:20) hybrid nanofluids, respectively, compared to water. Viscosity measurements revealed Newtonian behaviour for the Ag, SiC, and BeO nanofluids, with a minimum viscosity enhancement of 3.01 % observed for the BeO nanofluid. Hybrid Ag/SiC nanofluid demonstrated a maximum viscosity enhancement of 4.90 % for 20:80 formulation. Density analysis showed maximum augmentation of 0.25 %, 0.097 %, and 0.0775 % for the Ag, SiC, and BeO nanofluids respectively at 0.025 vol% while hybrid Ag/SiC nanofluid exhibited a maximum of 0.23 % density increase for the 80:20 composition. The statistical models were also developed to predict properties against temperature. Furthermore, the cost analysis identified Ag nanofluid as the most expensive option, while the SiC/BeO hybrid was the most economical. However, the Ag-SiC (60:40) hybrid nanofluid offered a balanced trade-off between properties and cost.

A systematic review and guide for using multi-response statistical models in co-infection research.

The simultaneous infection of organisms with two or more co-occurring pathogens, otherwise known as co-infections, concomitant infections or multiple infections, plays a significant role in the dynamics and consequences of infectious diseases in both humans and animals. To understand co-infections, ecologists and epidemiologists rely on models capable of accommodating multiple response variables. However, given the diversity of available approaches, choosing a model that is suitable for drawing meaningful conclusions from observational data is not a straightforward task. To provide clearer guidance for statistical model use in co-infection research, we conducted a systematic review to (i) understand the breadth of study goals and host-pathogen systems being pursued with multi-response models and (ii) determine the degree of crossover of knowledge among disciplines. In total, we identified 69 peer-reviewed primary studies that jointly measured infection patterns with two or more pathogens of humans or animals in natural environments. We found stark divisions in research objectives and methods among different disciplines, suggesting that cross-disciplinary insights into co-infection patterns and processes for different human and animal contexts are currently limited. Citation network analysis also revealed limited knowledge exchange between ecology and epidemiology. These findings collectively highlight the need for greater interdisciplinary collaboration for improving disease management.

Enhancing efficiency and quality control: The impact of Digital Twins in drinking water networks.

This paper showcases the successful development and implementation of two Digital Twin prototypes within the Lab Digital Twins project, designed to enhance the efficiency and quality control of Aigües de Barcelona's drinking water network. The first prototype focuses on asset management, using (near) real-time data and statistical models, and achieving a 70% success rate in predicting pump station failures 137 days in advance. The second prototype addresses water quality monitoring, leveraging machine learning to accurately forecast trihalomethane levels at key points in the distribution system, and enabling proactive water quality management strategies, ensuring compliance with stringent safety standards and safeguarding public health. The paper details the methodology of both prototypes, highlighting their potential to revolutionize water network management. PRACTITIONER POINTS: Digital representation of assets and processes in the drinking water treatment network Early fault detection in assets, and predictions of trihalomethane formation in the drinking water distribution network Reduction on monitoring time and incident response for target assets by means of Digital Twins Improvement in visualization, prediction, and proactive measures for asset management and water quality control Contribution to the growing knowledge on Digital Twins and their potential to revolutionize water network operations.

Analysis of the effectiveness of organic compounds from the amine group in precipitating ions from soda production wastewater

Treating wastewater from the soda industry is a complicated and lengthy process, requiring a great deal of labor and financial resources. No method has yet been developed to eliminate the environmental damage caused by the soda industry entirely. The work focused on the removal of chloride, sulfate, calcium, and magnesium from soda production wastewater by precipitation using organic solvents such as isopropylamine (IPA), diisopropylamine (DIIPA), propylamine (PA) and ethylamine (EA) in various proportions. Statistical modeling through Bayesian beta regression was used to select the amine most effectively removing the tested ions in precipitate form. The effect of precipitating agent dosage on the pH and conductivity of the solution was also investigated. Samples of wastewater obtained from the soda industry were characterized by high values of pH (up to 11.9), specific electrolytic conductivity (up to 128 mS cm-1), and high concentrations of sodium (up to 13 g L-1), chloride (up to 60 g L-1) and calcium (up to 24 g L-1) ions. Solvent-based precipitation showed that organic solvents are effective in precipitating salts from wastewater from the soda industry. Sulfate and chloride removal efficiencies of 85.1 and 34%, respectively, were observed. Statistical analysis showed that isopropylamine was the most effective amine for removal.

Humans vs. large language models: Judgmental forecasting in an era of advanced AI

This study investigates the forecasting accuracy of human experts versus large language models (LLMs) in the retail sector, particularly during standard and promotional sales periods. Utilizing a controlled experimental setup with 123 human forecasters and five LLMs—namely, ChatGPT-4, ChatGPT3.5, Bard, Bing, and Llama2—we evaluated forecasting precision through the absolute percentage error. Our analysis centered on the effect of the following factors on forecasters’ performance: the supporting statistical model (baseline and advanced), whether the product was on promotion, and the nature of external impact. The findings indicate that LLMs do not consistently outperform humans in forecasting accuracy and that advanced statistical forecasting models do not uniformly enhance the performance of either human forecasters or LLMs. Both human and LLM forecasters exhibited increased forecasting errors, particularly during promotional periods. Our findings call for careful consideration when integrating LLMs into practical forecasting processes.

Visual MODFLOW, solute transport modeling, and remote sensing techniques for adapting aquifer potentiality under reclamation and climate change impacts in coastal aquifer

Global environmental changes, such as climate change and reclamation alterations, significantly influence hydrological processes, leading to hydrologic nonstationarity and challenges in managing water availability and distribution. This study introduces a conceptual underpinning for the rational development and sustainability of groundwater resources. As one of the areas intended for the development projects within the Egyptian national plan for the reclamation of one and a half million acres; hundreds of pumping wells were constructed in the Moghra area to fulfill the reclamation demand. This study investigates the long-term impacts of exploiting the drilled pumping wells under climate change. The approach is to monitor the groundwater levels and the salinity values in the Moghra aquifer with various operational strategies and present proposed sustainable development scenarios. The impact of global warming and climate change is estimated for a prediction period of 30 years by using satellite data, time series geographical analysis, and statistical modeling. Using MODFLOW and Solute Transport (MT3DMS) modules of Visual MODFLOW USGS 2005 software, a three-dimensional (3D) finite-difference model is created to simulate groundwater flow and salinity distribution in the Moghra aquifer with the input of forecast downscaling (2020–2050) of main climatic parameters (PPT, ET, and Temp). The optimal adaptation-integrated scenario to cope with long-term groundwater withdrawal and climate change impacts is achieved when the Ministry of Irrigation and Water Resources (MWRI) recommends that the maximum drawdown shouldn’t be more significant than 1.0 m/ year. In this scenario, 1,500 pumping wells are distributed with an equal space of 500 m, a pumping rate of 1,200 m3/day and input the forecast of the most significant climatic parameters after 30 years. The output results of this scenario revealed a drawdown level of 42 m and a groundwater salinity value of 16,000 mg/l. Climate change has an evident impact on groundwater quantity and quality, particularly in the unconfined coastal aquifer, which is vulnerable to saltwater intrusion and pollution of drinking water resources. The relationship between climate change and the hydrologic cycle is crucial for predicting future water availability and addressing water-related issues.

Distribution of radon in large workplaces: an analysis performed on radon levels measured in UK schools

Radon is a radioactive, carcinogenic gas formed by the radioactive decay of uranium and radium that occur naturally in small amounts in all rocks and soils. It is the largest single source of radiation exposure to the UK population, contributing to more than 1 100 lung cancer deaths each year according to an analysis conducted in 2005. Regulations exist to protect employees (and other persons) where radon concentrations exceed the reference level of 300 Bq m-3. Once the reference level is exceeded, annual doses of more than the public dose limit of 1 mSv a-1are considered to be excessive. A radon measurement campaign for schools, which started in 2009, generated a large dataset, including those with high numbers of simultaneous radon measurements. Radon data between buildings (e.g. homes) have been shown to correspond broadly to the lognormal distribution, after the additive contribution of outside air has been removed. However, there are fewer studies of the distribution of radon levels within a single, large property. Radon data collected from 533 UK schools with at least 20 valid, simultaneous results were analysed against several statistical models. In approximately 50% of schools the radon levels could be represented by the lognormal distribution and in 60% by the loglogistic lognormal distribution, the latter being a better fit probably owing to its lower sensitivity to the tails of the distribution. Qualitatively, the lognormal and the loglogistic probability plots appeared to be indistinguishable. These findings indicate that the lognormal and loglogistic might be appropriate models to characterise the distribution of radon in most large workplaces. For each statistical model, the two distribution parameters can be used to provide a better estimate of the average dose to the occupants. However, caution is required when assessing doses, since the average estimator of the radon concentration does not predict the highest value and may significantly underestimate or overestimate the dose in specific areas.

A Statistical Model for Inference of Recent and Incident HIV Infection Using Surveillance Data on Individuals Newly Diagnosed With HIV Infection in Scotland

Background: To inform global ambitions to end AIDS, evaluation of progress toward HIV incidence reduction requires robust methods to measure incidence. Although HIV diagnosis date in routine HIV/AIDS surveillance systems are often used as a surrogate marker for incidence, it can be misleading if acquisition of transmission occurred years before testing. Other information present in data such as antibody testing dates, avidity testing result, and CD4 counts can assist, but the degree of missing data is often prohibitive. Methods: We constructed a Bayesian statistical model to estimate the annual proportion of first ever HIV diagnoses in Scotland (period 2015–2019) that represent recent HIV infection (ie, occurring within the previous 3–4 months) and incident HIV infection (ie, infection within the previous 12 months), by synthesizing avidity testing results and surveillance data on the interval since last negative HIV test. Results: Over the 5-year analysis period, the model-estimated proportion of incident infection was 43.9% (95% CI: 40.9 to 47.0), and the proportion of recent HIV infection was 21.6% (95% CI: 19.1 to 24.1). Among the mode of HIV acquisition categories, the highest proportion of recent infection was estimated for people who inject drugs: 27.4% (95% CI: 20.4 to 34.4). Conclusions: The Bayesian approach is appropriate for the high prevalence of missing data that can occur in routine surveillance data sets. The proposed model will aid countries in improving their understanding of the number of people who have recently acquired their infection, which is needed to progress toward the goal of HIV transmission elimination.

Evaluation, optimization study, and life cycle assessment of novel eco-friendly PVA-based nanocomposite hydrogel adsorbents for methylene blue and paracetamol removal

In this study, an eco-friendly and novel hydrogel based on a crosslinked polyvinyl alcohol (PVA), iota carrageenan (IC) and polyvinylpyrrolidone (PVP) scaffold, containing a large amount (10-50wt. %) of nanoscale palm fronds (NPF) as additives, for water purification was demonstrated. A life cycle assessment (LCA) findings on NPF as biomass waste incorporated into PVA_PVP_IC polymer matrix was presented, and the results highlight the necessity of focused actions to reduce environmental impact and support the palm waste utilization in a sustainable manner. The multicomponent nanocomposite hydrogels were examined as adsorbents in a system work in batches for methylene blue (MB) and paracetamol (PCT) removal. The results show that, the presence of NPF, which dispersed in the hydrogel PVA_PVP_IC scaffolds containing both covalent and non-covalent cross-linking bonds, greatly enhanced the MB and PCT adsorption efficiency. A response surface methodology (RSM) model was used to find the best operating parameters of contaminant adsorption, including time, adsorbent dose, and starting concentration of pollutants. By using this statistical model, it was found that the optimal conditions for the adsorption reaction to achieve the complete removal of MB are 66.7h adsorption time duration, 98.5mgL−1 starting concentration, and an adsorbent dose of 5.9mg, while for the complete removal of PCT, it is 57.6h adsorption time duration, 80mgL−1 starting concentration, and an adsorbent dose of 6mg. The reusability of the nanocomposite hydrogels were tested for 5 cycles, all showed high adsorption capacity, indicating the potential for practical application of this nanocomposite hydrogel system. This study indicates that the prepared nanocomposite hydrogel raises the standard used for treatment of wastewater and also gives a solution to protect the environment and mitigate global warming.

A Novel Strategy Coupling Optimised Sampling with Heterogeneous Ensemble Machine-Learning to Predict Landslide Susceptibility

The accuracy of data-driven landslide susceptibility prediction depends heavily on the quality of non-landslide samples and the selection of machine-learning algorithms. Current methods rely on artificial prior knowledge to obtain negative samples from landslide-free regions or outside the landslide buffer zones randomly and quickly but often ignore the reliability of non-landslide samples, which will pose a serious risk of including potential landslides and lead to erroneous outcomes in training data. Furthermore, diverse machine-learning models exhibit distinct classification capabilities, and applying a single model can readily result in over-fitting of the dataset and introduce potential uncertainties in predictions. To address these problems, taking Chenxi County, a hilly and mountainous area in southern China, as an example, this research proposes a strategy-coupling optimised sampling with heterogeneous ensemble machine learning to enhance the accuracy of landslide susceptibility prediction. Initially, 21 landslide impact factors were derived from five aspects: geology, hydrology, topography, meteorology, human activities, and geographical environment. Then, these factors were screened through a correlation analysis and collinearity diagnosis. Afterwards, an optimised sampling (OS) method was utilised to select negative samples by fusing the reliability of non-landslide samples and certainty factor values on the basis of the environmental similarity and statistical model. Subsequently, the adopted non-landslide samples and historical landslides were combined to create machine-learning datasets. Finally, baseline models (support vector machine, random forest, and back propagation neural network) and the stacking ensemble model were employed to predict susceptibility. The findings indicated that the OS method, considering the reliability of non-landslide samples, achieved higher-quality negative samples than currently widely used sampling methods. The stacking ensemble machine-learning model outperformed those three baseline models. Notably, the accuracy of the hybrid OS–Stacking model is most promising, up to 97.1%. The integrated strategy significantly improves the prediction of landslide susceptibility and makes it reliable and effective for assessing regional geohazard risk.

Computing Euclidean distance and maximum likelihood retraction maps for constrained optimization

Riemannian optimization uses local methods to solve optimization problems whose constraint set is a smooth manifold. A linear step along some descent direction usually leaves the constraints, and hence retraction maps are used to approximate the exponential map and return to the manifold. For many common matrix manifolds, retraction maps are available, with more or less explicit formulas. For implicitly-defined manifolds, suitable retraction maps are difficult to compute. We therefore develop an algorithm which uses homotopy continuation to compute the Euclidean distance retraction for any implicitly-defined submanifold of Rn, and prove convergence results.We also consider statistical models as Riemannian submanifolds of the probability simplex with the Fisher metric. Replacing Euclidean distance with maximum likelihood results in a map which we prove is a retraction. In fact, we prove the retraction is second-order; with the Levi-Civita connection associated to the Fisher metric, it approximates geodesics to second-order accuracy.

Statistical Models and Implant Customization in Hip Arthroplasty: Seeking Patient Satisfaction through Design

Statistical Models and Implant Customization in Hip Arthroplasty: Seeking Patient Satisfaction through Design

The Determinants of International Certification among Manufacturing Firms in Pakistan

This study investigates the factors that influence international certification among Pakistani manufacturing firms. About a third of such firms in Pakistan hold one or more certificates related to management or technical processes. The existing literature suggests that this demand is influenced by factors such as sector of activity, market competition, human capital and information environment, ownership type, and available resources. By analyzing a large, random sample of Pakistani manufacturing firms, we find strong correlations between these factors and international certification. We also discover that owner attitude, an aspect overlooked in previous studies, plays a role in the certification decision. While our statistical model does not fully explain the determinants of certification for small firms, we do find that certification is associated with increased profitability among small enterprises. This implies that small enterprise development agencies can improve the effectiveness of their programs by considering certification status as an additional criterion for allocating support funds. The results of our study have policy implications related to exporting, staff training, diffusion of firm ownership, and enhanced information access facilitated by modern technology tools.

Evaluation of statistical modeling (SM) approaches for landslide susceptibility mapping: geospatial insights for Bhutan

Evaluation of statistical modeling (SM) approaches for landslide susceptibility mapping: geospatial insights for Bhutan

A model of community youth mental health support services in the global South: A description and use of school-based services delivered by a non-governmental organisation in South Africa

ObjectiveYoung people in low- and middle-income countries (LMICs), including South Africa, are disproportionately affected by multiple risk factors for developing mental health problems and there are numerous barriers to accessing mental health care. In such contexts non-governmental organisations (NGOs) play a key role in providing support. We report on the mental health care services provided by NGO Community Keepers in South Africa. MethodCommunity Keepers (CK) offers psychosocial support to young people, parents, and teachers, during regular school hours, and free of charge. At the end of 2019, CK was based at 23 schools in 12 areas in the Western Cape Province, South Africa. In this paper we describe the model of service provision and present descriptive statistics from retrospective observational service data using routinely collected quantitative data from 2019, including details about referrals received, referral sources and reasons, and feedback received from children, parents, and teachers on the psychosocial support CK delivered. ResultsDuring 2019, 34 676 individuals accessed 18169 psychosocial support sessions rendered by CK. The most common referral source of learners to CK was self-referrals (40.6%), and the main referral reasons were emotional (psychological) problems, behaviour problems, or family (community) problems. At termination of therapy 320 (79.2%) learners said they strongly agreed that they would recommend CK to someone at their school. At termination of therapy with learners, 95 (39.3%) teachers rated CK’s services as excellent, 139 (57.4%) teachers gave a rating of good, and 204 (79.7%) parents reported an improvement in their relationship with their children. ConclusionsCK is an example of how psychosocial support can be embedded in schools in a LMIC. CK’s multifaceted model of service delivery could be used by other school-based mental health care service providers in other LMICs to inform and/or refine the services that they wish to offer.