Applying machine learning and geolocation techniques to social media data (Twitter) to develop a resource for urban planning.

With all the recent attention focused on big data, it is easy to overlook that basic vital statistics remain difficult to obtain in most of the world. What makes this frustrating is that private companies hold potentially useful data, but it is not accessible by the people who can use it to track poverty, reduce disease, or build urban infrastructure. This project set out to test whether we can transform an openly available dataset (Twitter) into a resource for urban planning and development. We test our hypothesis by creating road traffic crash location data, which is scarce in most resource-poor environments but essential for addressing the number one cause of mortality for children over five and young adults. The research project scraped 874,588 traffic related tweets in Nairobi, Kenya, applied a machine learning model to capture the occurrence of a crash, and developed an improved geoparsing algorithm to identify its location. We geolocate 32,991 crash reports in Twitter for 2012-2020 and cluster them into 22,872 unique crashes during this period. For a subset of crashes reported on Twitter, a motorcycle delivery service was dispatched in real-time to verify the crash and its location; the results show 92% accuracy. To our knowledge this is the first geolocated dataset of crashes for the city and allowed us to produce the first crash map for Nairobi. Using a spatial clustering algorithm, we are able to locate portions of the road network (<1%) where 50% of the crashes identified occurred. Even with limitations in the representativeness of the data, the results can provide urban planners with useful information that can be used to target road safety improvements where resources are limited. The work shows how twitter data might be used to create other types of essential data for urban planning in resource poor environments.

Use of standardised patients to assess quality of healthcare in Nairobi, Kenya: a pilot, cross-sectional study with international comparisons.

INTRODUCTION: The quality of clinical care can be reliably measured in multiple settings using standardised patients (SPs), but this methodology has not been extensively used in Sub-Saharan Africa. This study validates the use of SPs for a variety of tracer conditions in Nairobi, Kenya, and provides new results on the quality of care in sampled primary care clinics. METHODS: We deployed 14 SPs in private and public clinics presenting either asthma, child diarrhoea, tuberculosis or unstable angina. Case management guidelines and checklists were jointly developed with the Ministry of Health. We validated the SP method based on the ability of SPs to avoid detection or dangerous situations, without imposing a substantial time burden on providers. We also evaluated the sensitivity of quality measures to SP characteristics. We assessed quality of practice through adherence to guidelines and checklists for the entire sample, stratified by case and stratified by sector, and in comparison with previously published results from urban India, rural India and rural China. RESULTS: Across 166 interactions in 42 facilities, detection rates and exposure to unsafe conditions were both zero. There were no detected outcome correlations with SP characteristics that would bias the results. Across all four conditions, 53% of SPs were correctly managed with wide variation across tracer conditions. SPs paid 76% less in public clinics, but proportions of correct management were similar to private clinics for three conditions and higher for the fourth. Kenyan outcomes compared favourably with India and China in all but the angina case. CONCLUSIONS: The SP method is safe and effective in the urban Kenyan setting for the assessment of clinical practice. The pilot results suggest that public providers in this setting provide similar rates of correct management to private providers at significantly lower out-of-pocket costs for patients. However, comparisons across countries are sensitive to the tracer condition considered.

Observations of infection prevention and control practices in primary health care, Kenya.

OBJECTIVE: To assess compliance with infection prevention and control practices in primary health care in Kenya. METHODS: We used an observational, patient-tracking tool to assess compliance with infection prevention and control practices by 1680 health-care workers during outpatient interactions with 14 328 patients at 935 health-care facilities in 2015. Compliance was assessed in five domains: hand hygiene; protective glove use; injections and blood sampling; disinfection of reusable equipment; and waste segregation. We calculated compliance by dividing the number of correct actions performed by the number of indications and evaluated associations between compliance and the health-care worker's and facility's characteristics. FINDINGS: Across 106 464 observed indications for an infection prevention and control practice, the mean compliance was 0.318 (95% confidence interval, CI: 0.315 to 0.321). The compliance ranged from 0.023 (95% CI: 0.021 to 0.024) for hand hygiene to 0.871 (95% CI: 0.866 to 0.876) for injection and blood sampling safety. Compliance was weakly associated with the facility's characteristics (e.g. public or private, or level of specialization) and the health-care worker's knowledge of, and training in, infection prevention and control practices. CONCLUSION: The observational tool was effective for assessing compliance with infection prevention and control practices across multiple domains in primary health care in a low-income country. Compliance varied widely across infection prevention and control domains. The weak associations observed between compliance and the characteristics of health-care workers and facilities, such as knowledge and the availability of supplies, suggest that a broader focus on behavioural change is required.

Examining the Quality of Medicines at Kenyan Healthcare Facilities: A Validation of an Alternative Post-Market Surveillance Model That Uses Standardized Patients.

BACKGROUND: Promoting access to medicines requires concurrent efforts to strengthen quality assurance for sustained impact. Although problems of substandard and falsified medicines have been documented in low- and middle-income countries, reliable information on quality is rarely available. OBJECTIVE: The aim of this study was to validate an alternative post-market surveillance model to complement existing models. METHODS: The study used standardized patients or mystery clients (people recruited from the local community and trained to pose as real patients) to collect medicine samples after presenting a pre-specified condition. The patients presented four standardized conditions to 42 blinded facilities in Nairobi, Kenya, resulting in 166 patient-clinician interactions and dispensing of 300 medicines at facilities or nearby retail pharmacies. The medicine samples obtained thus resemble those that would be given to real patients. RESULTS: Sixty samples were selected from the 300, and sent for analysis at the Kenya National Quality Control Laboratory. Of these, ten (17%) did not comply with monograph specifications (three ibuprofen, two cetirizine, two amoxicillin/clavulanic acid combinations, and one each for prednisone, salbutamol and zinc). Five of the ten samples that failed had been inappropriately prescribed to patients who had presented symptoms of unstable angina. There was no association between medicine quality and ownership, size or location of the facilities. CONCLUSION: The study shows that the standardized patient model can provide insights into multiple dimensions of care, thus helping to link primary care encounters with medicine quality. Furthermore, it makes it possible to obtain medicines from blinded sellers, thus minimizing the risk of obtaining biased samples.