11.A. Workshop: Leveraging meso-level data to advance population health in Europe: further directions

Abstract   Health services and public health researchers provide timely and critical evidence to answer real-world policy questions and work extensively with policymakers at the macro, meso and micro levels of government. One goal shared by researchers and policymakers is to foster evidence-informed policy and program development to ensure that policy initiatives provide the greatest benefit possible to individuals and society. Among other sources of data, meso-level datasets are usually comprising contextual data aggregated at various geographical areas such as cities, counties, and regions. Although meso-level data are sometimes used as proxies for individual level data, they can also be used to explore complex questions at the population level. This workshop aims to provide a unique, interprofessional, European conversation about how to translate meso-level research evidence into meaningful insights or recommendations. It brings together a group of high-level people from academia, think tanks, and companies who are involved in generating, transferring, or using meso-level evidence to inform public health and health care policy in Germany and France. In the first presentation, Schüttig et al. use district-level data from Germany to suggest that increased spending and improved continuity of care may be effective ways to reduce the rate of potentially avoidable hospitalizations among patients with type 2 diabetes. Then, Mercier et al. analyze department (district)-level data from France to quantify the impact of the population-based prevalence of diabetes and psychiatric conditions, of air pollution, of socio-economic variables, and of meteorological factors on the spread of COVID-19 during the first lockdown. Rodts et al, in a collaboration between a think tank and a small company, use a broad set of district-level variables to classify French ‘departements’ into homogeneous clusters in terms in needs and explore the discrepancies between total health care spending and needs at the population level. Finally, Mâlatre-Lansac et al. build on these studies to discuss how data can be used to inform public health and health care policy making in Europe. In addition, they suggest future directions to improve meso-level data-driven policy at the local, national, and European levels. Beyond methodological points, the discussion will address ethical issues in the use of meso-level data, and how to improve the availability of data, and the ability of local, regional, and national policymakers to use research evidence efficiently. It is designed as a regular workshop with 4 presentations (10 minutes each), ample audience interaction through Q&A after each presentation and one freehand poll in the introduction of each presentation. Key messages • Meso-level data can be efficiently leveraged to inform health care policy. • Further efforts need to be taken to address the information needs of policymakers.


Introduction:
Studies on mental health inequalities are usually based on limited sets of mental health indicators. Objectives: Using a large number of mental health indicators, we explored whether it is possible to identify similar hierarchical rankings regardless of mental health indicators (incl. psychotherapy) among employees representing different socio-demographic statuses, and which groups of employees have the highest mismatch between mental health symptom and treatment.

Methods:
Employees representing different occupational classes and employees from four different areas of Finland were studied and compared. We used national register data to define psychotropic medication (purchases), sickness absence for mood disorders, and the use of psychotherapy between 2017 and 2019 and national survey data from the FinHealth 2017 Study to define the level of psychological symptoms (BDI, GHQ). We assessed the risk of each outcome by population group separately for men and women, and estimated the mismatch between symptoms (BDI/GHQ caseness) and treatment (psychotropic drugs/therapy).

Results:
In all the studied groups, the prevalence of mental health indicators was mostly considerably higher among women than men. The risk of register-based mental health indicators was typically higher among lower non-manual employees. In the case of some mental health indicators, we observed significant interactions between occupation class and region. Some stark mismatches were detected between symptoms and treatment in some populations, whereas at the other end of the spectrum, the correspondence between symptoms and the mobilization of care was rather high.

Conclusions:
Although gender is strongly linked to mental health indicators, occupational class and region influence mental health profiles in the population. There are considerable inequalities between populations in the level of professional care associated with mental health problems.
Health services and public health researchers provide timely and critical evidence to answer real-world policy questions and work extensively with policymakers at the macro, meso and micro levels of government. One goal shared by researchers and policymakers is to foster evidence-informed policy and program development to ensure that policy initiatives provide the greatest benefit possible to individuals and society. Among other sources of data, meso-level datasets are usually comprising contextual data aggregated at various geographical areas such as cities, counties, and regions. Although meso-level data are sometimes used as proxies for individual level data, they can also be used to explore complex questions at the population level. This workshop aims to provide a unique, interprofessional, European conversation about how to translate meso-level research evidence into meaningful insights or recommendations. It brings together a group of high-level people from academia, think tanks, and companies who are involved in generating, transferring, or using meso-level evidence to inform public health and health care policy in Germany and France. In the first presentation, Schüttig et al. use district-level data from Germany to suggest that increased spending and improved continuity of care may be effective ways to reduce the rate of potentially avoidable hospitalizations among patients with type 2 diabetes. Then, Mercier et al. analyze department (district)-level data from France to quantify the impact of the population-based prevalence of diabetes and psychiatric conditions, of air pollution, of socioeconomic variables, and of meteorological factors on the spread of COVID-19 during the first lockdown. Rodts et al, in a collaboration between a think tank and a small company, use a broad set of district-level variables to classify French 'departements' into homogeneous clusters in terms in needs and explore the discrepancies between total health care spending and needs at the population level. Finally, Mâlatre-Lansac et al. build on these studies to discuss how data can be used to inform public health and health care policy making in Europe. In addition, they suggest future directions to improve meso-level data-driven policy at the local, national, and European levels. Beyond methodological points, the discussion will address ethical issues in the use of meso-level data, and how to improve the availability of data, and the ability of local, regional, and national policymakers to use research evidence efficiently. It is designed as a regular workshop with 4 presentations (10 minutes each), ample audience interaction through Q&A after each presentation and one freehand poll in the introduction of each presentation. Key messages: Meso-level data can be efficiently leveraged to inform health care policy. Further efforts need to be taken to address the information needs of policymakers. Ambulatory care sensitive hospitalizations are widely considered as important measures of access to as well as quality and performance of primary care. In our study, we investigate the impact of spending, process quality and continuity of care in the ambulatory care sector on ambulatory care sensitive hospitalizations in patients with type 2 diabetes. We used observational data from Germany's major association of insurance companies from 2012 to 2014 with 55,924 patients, as well as data from additional regional data sources. We conducted negative binomial regression analyses with random effects at the district level. To control for potential endogeneity of spending and physician density in the ambulatory care sector, we used an instrumental variable approach. In doing so, we adjust for a number of known risk factors for hospitalizations among this patient group. We undertook a Shorrocks-Shapley decomposition to investigate the relative contribution of groups of regressors to the pseudo R2. The results of our analysis suggest that spending in the ambulatory care sector has weak negative effects on ambulatory care sensitive hospitalizations. We also found that continuity of care was negatively associated with hospital admissions. Patients with type 2 diabetes are at increased risk of hospitalization resulting from ambulatory care sensitive conditions. The results of the decomposition analysis for groups of variables indicate that ambulatory care characteristics account for 9.8% of the pseudo R2, morbidity of patients (including gender and age groups) for about 85.5%, and system-related factors of health provision for 4.7%. Our study provides some evidence that meso-level factors such as increased spending and improved continuity of care while controlling for process quality in the ambulatory care sector may be effective ways to reduce the rate of potentially avoidable hospitalizations among patients with type 2 diabetes.
The impact of various environmental, socio-economic, and epidemiological factors on COVID-19 transmission and severity is well-known. However, there is little evidence about the respective role of these factors at the populationlevel at a national scale. The objective was to identify the environmental and contextual factors that influenced the spread and the severity of COVID-19 at the French department level during the first national lockdown. We performed a national, population-based, retrospective analysis. The cumulative rate of patients admitted for COVID-19 to any public or private acute care hospital from March 31st, 2020 to May 25, 2020 was modelled at the 'departement' (hereafter county) level. We used spatial regression models to quantify the aggregated effect of population health status, air pollution, meteorological, and socioeconomic factors. 57,356 patients were admitted to an acute care facility for COVID-19 over the period of interest. At the county level, the age and sexstandardized rate of admission ranged from 0.07 to 3.24 admissions per 1,000 people. After adjustment on the prelockdown COVID-19 hospital admission rate, the standardized cumulative rate hospital admission for COVID-19 during the period of interest was significantly and positively associated with the prevalence of diabetes, with the prevalence of mental conditions, and with high cumulative exposure to atmospheric ozone values. It was significantly and negatively associated with high cumulative exposure to ultraviolet radiation. These results suggest that several population-based epidemiological and meteorological factors could have played a role in COVID-19 spread in France. They provide potentially useful insights to design and implement geographically differentiated public health policies. Aligning health care spending with population needs is a goal shared by many public health and health care systems. However, most modelling approaches have proven deceptive and ineffective. We propose a novel data-driven, populationbased approach to help policymakers explore the discrepancies between spending and needs in France. We leveraged several national open data sources covering demographics, social deprivation, epidemiology, environment, health-related behaviors, and all-payer health care spending (hospital inpatient, prescription medicines, ambulatory events, and dental care). We classified the French ''departements'' (hereafter counties) into clusters that are homogeneous in terms of health care needs, based on a multidimensional framework. Then, we calculated all-payer per capita health care spending to analyze its within-and between-cluster variation. Based on these findings, we designed a web-based, interactive mapping tool dedicated to French policymakers and payers. The analysis shows 7 clusters of French counties differing in terms of health care needs and spending. The higher-needs/lower-spending and lower-needs/higher spending clusters suggest considerable room for improvement through a regional distribution of spending at least partially based on needs. Most interestingly, the data we used is publicly available, but policymakers lack expertise and time to undergo such analyses themselves. We plan to develop a dynamic and more granular version of the tool to allow policymakers to accurately design and evaluate health care policies.
Abstract citation ID: ckac129.691 Data-driven policy making: a key step for the healthcare system The work presented in this workshop suggest important room for improvement in data-driven health policy making in France and in Germany. We identify data generated by 15th European Public Health Conference 2022