Abstracts from the 7th International Jerusalem Conference on Health Policy

Background: Repurposing of existing drugs for new indications can save years of research and tremendous resources.

Study Question: Can we test a new repurposing model and create an algorithm by utilizing machine learning of big data to identify the potential role of concomitant drugs taken by hypertensive and diabetic Type 2 patients.

Methods: We applied machine learning techniques to identify concomitant drugs not taken for hypertension or diabetes which may contribute to lowering blood pressure or improving glucose control.

Results: Among numerous concomitant drugs taken by hypertensive patients, statins and protein pump inhibitors (PPI) significantly improved blood pressure. For Type 2 diabetic patients, alpha 1 blocking drugs were the only group of medications to significantly improve the success rate of glucose control. These effects of statins, PPI and alpha 1 blockers shown by us have been recently documented in animal models and small human studies, validating this new innovative method.

Conclusions: Machine learning of big data is a novel method to identify effective repurposing of medications already on the market for new indications. Our model has been validated by contrasting these discoveries with recent animal and human studies.

Health Policy Implications: In the case of alpha 1 blockers, because this class of medications is widely used in men with benign prostate hyperplasia at age groups with increased rates of type 2 diabetes, this finding is of potential clinical and health policy significance.

Background: Studies emphasize the importance of quality data for effective planning, monitoring, and evaluation. The 2015 National Immunization Coverage Survey (NICS) in Nigeria revealed data quality issues across all 36 states, (worse in 18 states) of the Country with National Penta3 Coverage at 33% as against Administrative (Admin) Coverage of 98%, hence the importance of RI LQAS to identify reasons for disparities.

Study Question: Does the RI data from health facilities (HFs), aggregated at districts (LGAs) reflect true performance? What reasons explain disparities? What are the primary sources of information that guide vaccination amongst caregivers?

Methods: 18 priority States were selected. Using a probability proportionate to population size, all LGAs in a State, 6 HFs and 6 settlements were selected. Within each settlement, 10 households (HH) were selected from which 1 child 0-11 months per HH was selected for assessment by reviewing the immunization card and/or mother’s recall. Reasons for partially/not immunized for age and sources of information about vaccination were ascertained.

Results: A total of 2,292 sites, 22,920 households, 22,920 children 0-11months sampled in 382 districts. Only 3% of the districts passed, with 56% of the States having < 80% coverage. At the individual level, only 1 of 3 children in the Country was fully immunized for age.

Conclusions: The RI LQAS confirmed findings of the NICS that admin data did not reflect RI performance in Nigeria. Subsequently, a State of Emergency on RI with RI Coordinating Centers set up across States with RI LQAS conducted quarterly to review performance.

Health Policy Implications: Instituting frequent low budget surveys as RI LQAS is effective in ensuring reliable data for planning to improve vaccination of children and enshrine accountability amongst erring HWs that falsify data.

Background: Maccabi Healthcare Services (MHS) has set for itself the target of creating the first population-based biobank in Israel, thus providing an ideal platform for Precision Medicine research. MHS is the second largest healthcare provider in Israel. MHS serves 2.3 million members which constitutes a representative quarter of the Israeli population. Electronic health records have been implemented in MHS for over a quarter of a century. With less than 2% annual member turnover, MHS records have a longitudinal history of many patients throughout their entire lives. In 2017 the Tipa Biobank (Tipa) was launched, a nationwide biobanking program to link samples and EHR data for broad research use.

Study Question: We describe the initial phases of implementation of a large-scale population-based biobank initiative within the setting of a public healthcare fund.

Methods: Patients are approached by trained concenters and offered to participate. Samples are collected in a standardized manner and transported daily using the existing nationwide infrastructure of MHS.

Results: As of March 2019, over 70,000 MHS patients consented to participate in Tipa and over 100,000 vials have been stored. Of the patients approached, over 50% consent to participate. Tipa participants are slightly older (mean age of 49.7) and are more likely to be female (60%). 23% of participants are diagnosed with hypertension and 18% of them suffer from diabetes of pre-diabetes.

Conclusions: With virtually no funding MHS succeeded in creating the first population-based biobank in Israel. Tipa is a resource that enables a new model for translational research that is faster, more flexible, and more cost effective than traditional clinical research approaches. The model is scalable, and will increase in value as resources grow.

Health Policy Implications: Findings from using this resource will impact the development of personalized medicine tools and which will not only impact Tipa participants but will change the way patients are treated.

Background: Interventions requiring a restructuring of health care systems are rarely replicated and difficult to extend beyond select early adopting health care organizations. However, since most program innovations require adoption by functioning health care systems, the classic NIH research phase models ignore this crucial implementation and dissemination phase which also needs to be rigorously tested.

Study Question: What are the necessary conditions under which a health systems intervention can be successfully implemented in the real world of functioning health care systems?

Methods: We examined multiple pragmatic, cluster randomized clinical trials implemented in functioning health care systems in order to identify the characteristics of interventions that are associated with successful replication and implementation of the intervention. A total of 5 different types of interventions were examined with successively more complicated interventions.

Results: Interventions that merely substitute one treatment for another are readily implemented with very successful adherence to the intended design. As interventions require changes in assignments, allocation of resources and shifts in expectations of supervisors and staff, the degree of adherence to the research model of the intervention breaks down considerably. The more the breakdown in intervention adherence, the greater the effect of the intervention must be in order for it to demonstrate the intended effect on patients’ outcomes.

Conclusions: Understanding the potential that innovations in health care practice can achieve the intended effects requires careful consideration of the level of organizational disruption that each entail since health care organizations have considerable entropy which can reduce the likelihood that the desired outcomes are achieved.

Health Policy Implications: Before interventions, shown to be effective in leading organizations, are widely promulgated across the country, it is essential that researchers and policymakers understand how the necessary organizational changes in daily practice can be achieved in the average health care organization.

Background: Under a national endeavor to establish the role of Diabetes Educators (DE) in the healthcare system, during 2015-2018 over 200 healthcare professionals were trained under the auspice of the Ministry-of-Health to serve as educators. The training comprised of theoretical knowledge and hands-on practice, with an internationally-unique accreditation exam of OSCS carried out by standardized patients in clinic-like settings. Up until now, evaluation of the simulated therapeutic meetings was conducted by trained evaluators, but in a subjective manner. In order to assess the suitability of the training to the DE role and establish the accreditation system, there is a need for a reliable and valid quantitative evaluation.

Study Question: (1) To examine the use of OSCS to modulate a real-life therapeutic-meeting. ((2 To develop and validate an instrument for quantitative evaluation of the OSCS.

Methods: All DE were asked to provide feedback through a self-report questionnaire on the suitability of the OSCS modality. In addition, an evaluation tool for the accreditation exam was constructed on the basis of well-established relevant scales and consultation with experts. The validity and reliability of the instrument were tested by several independent rounds of evaluations, having corrections to the inventory after each round. Cost components of OSCS per participant were collected.

Results: Over 90% of the participants expressed satisfaction from the compatibility of OSCS with their everyday work and recommended to continue with this method. The main parts of the final version of the instrument include: general impression, the structure of the meeting, obligatory communication skills, deeper level skills and summary score. The validity and reliability scores of the OSCE evaluation were high. The cost per participant was calculated to be USD 150.

Conclusions: OSCS is a suitable methodology for testing the quality of a therapeutic meeting in a clinical setting. The suggested instrument was proved to allow credible evaluation.

Health Policy Implications: Implement the evaluation tool for accreditation tool as well as for the routine assessment of DE in practice.

Background: A diagnosing of rare diseases takes an unusually long time. During that period, the patient gets 2-3 misdiagnoses. Educational web presentations of patients with rare diseases show that among the participants, at least one has the right answer. An off-line discussion about a patient with an unusual clinical picture may be a promising opportunity for a physician experiencing diagnosing difficulties.

Methods: The social network “Doctor na rabote” (DnR) brings together more than half a million Russian-speaking physicians. Almost every day, its members turn to their colleagues with questions about diagnostic difficulties. We conducted an observational study of difficult-to-solve cases, presented on DnR in the period 1.09.18 - 02.28.19. In order to analyze the practice of a web Concilium regarding patients with possible rare diseases, we chose cases in which at least 20 doctors participated. We investigated the time between publication and the receipt of the last diagnostic assumption, the number of independent diagnostic assumptions and the total number of participants in the discussion.

Results: During the study period, there were 25 discussions that met the inclusion criteria. Two cases of them were presented twice. The average number of participants who made significant comments was 27.32 (20-41). The count of independent diagnostic assumptions per case varied from 6 to 22 and averaged 12.6. The time between publication and receipt of the last of the diagnostic version did not exceed two weeks.

Conclusions: An off-line Concilium regarding difficult-to-diagnose cases yields a significant number of diagnostic assumptions in a very short time.

Health Policy Implications: An adding to off-line web Concilium a tool for choosing between assumptions can significantly speed up the diagnosis of such patients in everyday medical practice.

Background: Seasonal influenza vaccination coverage remains suboptimal in most developed countries, despite longstanding recommendations of public health organizations. The individual’s decision regarding vaccination is located at the core of non-adherence.

Study Question: Can we predict whether a patient will become vaccinated against influenza in the next season?

Methods: A retrospective longitudinal cohort study, utilizing data from the electronic medical records of 250,000 members of Maccabi Healthcare Services.

Results: The vaccination decision of an individual can be explained in two dimensions – personal and social. The personal dimension is strongly shaped by a “default” behavior, such as vaccination timing in previous seasons and general health consumption, but can also be affected by temporal factors such as respiratory illness in the prior year.

Conclusions: The decision of an individual to become vaccinated is highly predictable.

Health Policy Implications: Our study sets a major step toward personalized influenza vaccination campaigns.

Background: Osteoporotic (OP) fractures cause major morbidity and mortality. The clinical importance of fracture risk prediction models such as FRAX is well established, but these models require many inputs and are underutilized in practice.

Study Question: How does the predictive ability of OP-fracture risk scores, derived automatically from routine CTs using an artificial-intelligence based algorithm, compared to that of FRAX?

Methods: Members of Clalit Health Services aged 50-90, who underwent routine chest or abdomen CTs prior to July 2012, were included.

Results: A total of 48,227 individuals were analyzed. Of these, 5,106 (10.6%) and 1,901 (3.9%) suffered major OP and hip fractures during follow-up, respectively.

Conclusions: Fully automatic screening for OP fracture risk using routinely acquired CT scans achieves discrimination which is at least as good as the well-established FRAX model. If data for all required inputs is available, combining the CT scores with FRAX further improves predictive ability.

Health Policy Implications: OP fracture prediction scores can be added automatically to CT reports, to help identify people at risk for fractures who are currently missed.

Background: The act of diagnosing a mental illness amongst one of the children in the family changes the family balance, burdens it with a heavy stigma and makes parental care more difficult. The influence of the disease on the family and parents was researched in relation to the subjective burden concept, a burden which describes the negative psychological consequences of the disease on those treating it as feelings of loss, depression, fright and embarrassment.

Study Question: The research examines how Arabic, Islamic mothers describe and experience the subjective burden that is cast upon them in light of treating the harsh mental disability of their children. Furthermore, it examines the cultural and influences of cultural motives in the construction of the motherly burden experience

Methods: The research included 30 married Arabic mothers who live in the state of Israel. They were part of a ‘purposeful’ sample. The research interview included 19 qualitative questions, in addition to a socio-demographic questioner containing 16 questions. The findings analysis was performed according to the qualitative methodology, and according to the Grounded Theory of Strauss and Corbin.

Results: The research found six areas in that affected the motherly burden: the religious factor; the types of motherly emotions; the perception of the illness; contact with professionals’ stigma’ and changes in the Arabic family. Socio-demographic factors that are unique to the Arabic family such as: low socio-economic status, mother's education, high number of children, lack of gender equality, low accessibility to social services, and families with multi-generational distress, all put an even greater motherly burden.

Conclusions: Mothers of daughters reported a subjective load and burden that were higher than mothers to boys with mental disabilities. A heavy subjective burden was reported by mothers of families that are suffering from multi-generational distress, such as: unemployment; poverty; sickness; drug dealing; and crime.

Health Policy Implications: These findings helped the understanding of the unique motherly needs, and thus helped develop specific intervention programs focused at giving an answer to the consequences of the motherly burden, in addition to helping form a macro level policy to the Arab society in Israel.

Background: Human resources for health is key to achieving the UNAIDS 90-90-90 targets by 2020. Retention in care is a step towards achieving viral load suppression among clients in HIV/AIDS care and treatment. The “Care&Treatment” program is implemented in 137 Ministry of Health facilities in the Rift Valley Province of Kenya.

Study Question: What is the impact of healthcare worker numbers on retention in care among People living with HIV (PLHIV)?

Methods: We performed retrospective analysis on routinely collected program data on staff and retention between October-December, 2018. These were staff numbers, adherence counselors, case managers/mentor mothers, nurses, and registered clinical officer (RCO). We calculated retention as a percentage of adults and children known to be on treatment 12 months after initiation of antiretroviral therapy and used Poisson Regression for count data controlling for facility-level clustering and using the total number of staffs at the facility level as an offset variable to assess the association between staff numbers and retention.

Results: There were 63/548(11.5%) -adherence counsellors, 296/548(54%)-case-managers, 97/548(17.7)-nurses, and 92/548(16.8%) registered clinical officers in 137 facilities. We observed statistically significant association between availability of overall staff numbers and retention, availability of 1-5 staff was IRR [0.35 95%CI (0.33-0.37)], 6-10 staff [IRR [0.14 95%CI (0.13-0.15)], and availability of above 10 staff numbers, IRR [0.055, 95% CI (0.052-0.060)]. Availability of adherence counsellor was protective [IRR 0.46 95%CI (0.44-48)]; 2-4 adherence counsellors [IRR 0.25 95%CI (0.23-0.27)]. availability of between 1-2 mentor-mother(s)/case-manager(s) was protective [IRR 0.40 95%CI (0.37-43)]; 3-4 mentor-mothers/case-managers [IRR 0.19 95%CI(0.18-20)]; above 5 mentor-mothers/case-managers [IRR 0.09 95% CI(0.08-0.10)]. availability of nurses was protective [IRR 0.56 95%CI (0.54-59)]. Availability of 1 RCO was protective [IRR 0.44 95%CI (0.42-46)]; between 2-4 RCOs [IRR 0.27 95%CI (0.25-0.29)].

Conclusions: Availability of healthcare workers improves retention and increased numbers improve retention in care among PLHIV.

Health Policy Implications: Understanding the distribution and allocation of healthcare workers will help the government to allocate scarce healthcare workers to achieve optimal health outcomes.

Background: China has one of the largest numbers of hospitals in the world. Access to good quality health care remains a problem in both urban cities as well as in the rural areas. Patients are rushing to outpatient clinics in well-known hospitals to register (guahao in Chinese) and have to wait to see the experts and specialists. The recent reform of the health services in China has facilitated the emerging giant online health services providers to fill in the gap in the form of ‘Internet + Healthcare’.

Study Question: (1) What are the current government policies and regulations in China towards the practice of ‘Internet + Healthcare’? (2) What are the current types of services and the model of this ‘Internet + Healthcare’? (3) Evaluate the current two major listed ‘Internet + Healthcare’ providers in the China market.

Methods: (1) Content analysis of the government policies and regulations regarding the development of ‘Internet + Healthcare’. (2) Case Study of the two major listed “Internet + Healthcare’ providers in China market in terms of their market, model and their outcomes.

Results: (1) A disruptive model of the health care market in China has been currently dominated by internet or financial giants. (2) Market share is constantly growing in this field. (3) Investment Funding and Initial Public Offering are the major sources of funding for sustaining and developing the services. (4) Changes in the behaviors of local citizens in seeking health advice and consultations.

Conclusions: While the current scale of ‘Internet + Healthcare’ provision in China is stilling developing, it is envisaged that there is a strong growing demand of adopting this new model to seek health services from the citizens in China. While these providers may not acquire their financial profits yet, this new model of economy has attracted much attention from the investors’ point of interest.

Health Policy Implications: (1) Can this model of ‘Internet + Healthcare’ be applied to other countries? (2) How to regulate and assure the quality of services or advices? (3) Urgent need for government to evaluate the effectiveness of the ‘Internet + Healthcare’. (4) Update and refine the current policy regarding ‘Internet + Healthcare’.

Background: Digital-care (DC) models have been criticized and simultaneously praised as contributing to the expansion/reduction of health disparities. Poverty is a multidimensional social problem prominent in all developed countries and an important determinant of health. Are DC models a threat or an opportunity to people living in poverty?

Study Question: To examine the pros and cons of DC models in mitigating the effects of poverty on health and their contribution to people living in poverty's needs.

Methods: A qualitative analysis of 3 approaches focused on the delivery of care to people in poverty (“Pro-poor Health Systems”; “Social Exclusion” and “Poverty-Aware social-work -PPSW) was undertaken. 16 principles were grouped into 3 groups (interpersonal-downstream factors; practical-midstream factors; and social-upstream factors). Both groups and principles underwent examination in order to characterize the challenges and opportunities they insinuate to DC models.

Results: DC models can help mitigate numerous practical-midstream barriers including: economic barriers to care as well as hidden expenses such as expenses on transportation; help expand health care organizations’ ability to expand outreach to people in poverty; and provide better accessibility to knowledge and entitlements. Contrariwise, DC models are challenging mainly for interpersonal-downstream factors such as emotional needs, trust and close relations and “symbolic” capital, important factors identified as crucial in PASW approaches. Social-upstream factors such as Social exclusion are also challenging for DC models since they potentially strengthen individualistic patient-oriented approaches rather than socially-oriented ones.

Conclusions: Consideration of barriers to care for different social groups, stemming from the use of DC models, should expand to include not only the “usual suspects” such as health literacy and technologic accessibility but also to other interpersonal, ethical, social, cultural and political factors.

Health Policy Implications: Systematic examination of DC models for care is needed, especially regarding excluded social groups.

Background: Brazil and Israel are very distinct countries, but both are young multiparty democracies, which adopted universal health systems (UHS) in the late twentieth century.

Study Question: The study presents a comparison between the institutions that constitute the UHS of Israel and Brazil, considering the socio-economic context and the health results obtained by the two countries.

Methods: The study discusses the characteristics of each health system and compares the evolution of health indicators (on mortality, morbidity, risk factors, health actions, health resources, and health expenditures), from the late 1990’s up to 2018. The data were obtained from the World Bank and the Organization for Economic Co-operation and Development.

Conclusions: It is perceived that the Israeli system, which is concerned about achieving greater equity, could benefit from the observation of the successful experiences of the Brazilian system, to avoid the erosion of universality. The Brazilian system, in turn, could benefit from the reflection on strengths observed in the Israeli system, mainly the efficient planning and management of health services.

Health Policy Implications: The results of such studies can be used to support proposals for improvements of UHS, even in countries of varying degrees of economic development.

Background: The Brazilian National Congress (BNC) has been outstanding in approving relevant laws related to health policies. However, the monitoring of legislative proposals is hampered by their large numbers. Previous research has identified more than 20,000 proposals submitted between 1999 and 2006 (about 5,000 related to health). Machine Learning can facilitate the task of identifying and classifying the thousands of health-related legislative proposals.

Study Question: What is the sensitivity and specificity of machine learning models to predict the classification of health-related legislative proposals in the BNC?

Methods: The methodology consists of classifying the health-related legislative proposals that are active in one of the Legislative Houses of the BNC, the Chamber of Deputies (CD); using a typology which has five thematic groups that covers about 60 categories. Between 70 and 80% of those propositions will be used for machine learning, while the rest will be used for the sensitivity and specificity tests of the supervised models, like K-Nearest Neighbors (K-NN), Support Vector Machine (SVM), SVM Kernel, Naive Bayes, Decision Tree and Random Forest.

Results: The project has completed its most time-consuming phase: the classification of health related propositions by human classifiers. The data set has 3,592 health related ordinary bills. The most frequent themes were: prevention (35.2%), rights and responsibilities (28.5%), health care (18%) and management and resources (18%). The tests results will be available for presentation in September, 2019.

Health Policy Implications: The use of machine learning models has the potential to speed the classification of health-related bills, which can be useful in organizing and monitoring the health agenda in Congress, favoring transparency in the public debate and deliberation on health policies.

Background: The wellbeing of the population can be affected if policymakers cannot understand that the health of the population is determined by the physical environment, inhabitants’ living conditions and available services. A lot of data is collected, and a lot of register-based studies are published in scientific papers; however, seldom this data makes sense to policymakers.

Study Question: The aim of this project was to develop a social surveillance instrument suitable for policymakers.

Methods: Sociodemographic and health-related data were collected at the city-level, district-level and subdistrict-level. Maps showing services, housing, roads and public places were produced at subdistrict-level. A policymaker-friendly layout was created in 2014 for one of the ten districts of the city of Gothenburg. Since then the instrument has been updated every year.

Results: Knowledge regarding the geographical differences in social determinants and health in the population has increased among policymakers in the intervention area. Policymakers in the selected district use the created instrument in their official meetings while discussing, planning and following up political decisions. Other organizations also have used the created instrument in discussions regarding social and health-related interventions and regarding the allocation of services, recreational areas and new residential areas.

Conclusions: The gap between biostatisticians and policymakers can be reduced using the adequate channels and languages.

Health Policy Implications: Policymaker-friendly instruments can improve political decisions and improve the health of the population.