Figure 1
Panel A – AMG input: Required input variables to compute AMG; Panel B – AMG output: Output variables of the AMG algorithm. *Binary markers (presence/absence) of 15 chronic conditions (from left to right): diabetes mellitus, heart failure, chronic obstructive pulmonary disease, high blood pressure, depression, HIV/AIDS, chronic ischemic heart disease, stroke, chronic kidney disease, cirrhosis, osteoporosis, arthrosis, arthritis, dementia, chronic pain; Panel C– Health Risk Assessment based on AMG: The AMG scoring allows for three key actions: Classification: The population is categorised into specific groups based on their morbidity statuses, such as healthy, pregnancy and labour, acute disease, chronic disease in 1–4 systems, or active neoplasia, which are also divided into five degrees of severity. Stratification: Each individual can be assigned a complexity score that reflects the care needs that people may have based on their health problems. Identification: Individuals with specific major chronic health problems can be identified, which helps track people with more complex care needs.
Table 1
Summary of the pre-implementation process, including the context and trigger, the aims, the baseline situation, and the challenges faced in the Marche region and Estonia.
| MARCHE REGION | ESTONIA | |
|---|---|---|
| Context and trigger |
|
|
| Aims |
|
|
| Baseline situation | The Regional Healthcare Administrative Databases (HADs), used at regional and national levels to monitor healthcare system expenditures and performance, gather information on healthcare services provided to citizens (e.g., hospitalisations, emergency-urgency, homecare, exemption codes, etc.). Common standard models and coding systems, such as DRGs [33] and ICD-9-CM [34], are used across all regions. However, each HAD has its unique structure, unit level, content, and rules for data input. Data linkage across HADs is facilitated using a unique anonymised patient ID code. | The Estonian Health Insurance Fund claims database was used for model data input. In Estonian universal healthcare and single payer model this database entails almost all medical care claims in the country. The database follows a single standard with ICD-10-CM [35] coding, DRGs. Data linkage and access was granted though ethics committee approval and is not easily available as standard. Data was analysed in anonymised format using unique patient ID codes. All AMG analyses were performed by the regional medical authorities under the ethics committee approved application with support from the oGP. |
| Challenges |
|
|
Figure 2
Results of the feasibility analysis: in Catalonia (1) and the adoption regions, Marche Region (2) and Estonia (3). Panel A – AMG risk distribution: itemised by age and gender; Panel B – AMG disease groups: distribution of the seven AMG morbidity groups (bars): healthy, pregnancy, acute disease, 1 or 2–3 or ≥4 chronic diseases, and active neoplasia, and their average Morbidity Burden Score (line).
Table 2
Checklist of key steps for site adoption of Health Risk Assessment (HRA).
| KEY STEPS | DESCRIPTION |
|---|---|
| 1. Scope definition | Identify the purpose, focus/use and ambition of the HRA initiative |
| 2. Source population | Population-health or Population-medicine. For each option, identify specificities of the source population |
| 3. Planned updates | Periodicity of source data update (i.e. yearly basis) |
| 4. Model (Morbidity grouper) | Morbidity grouper selected: AMG, CRG, ACG, others |
| 5. Input variables (Figure 1) | Minimum variables of the morbidity grouper plus additional variables selected for inclusion in the HRA modelling. |
| 6. Data source(s) (data sources and ownership have technical/managerial implications) |
|
| 7. Predictive modelling | Identify statistical methods, Machine Learning, Deep Learning, etc… |
| 8. Output variables(Figure 1) | Variables generated by the predictive modelling approach |
| 9. Feasibility assessment | Includes characterization of site maturity and preliminary testing of the HRA tools to ensure minimum quality before implementation |
| 10. Technological logistics | The complexity of the digital setting is closely related to the characteristics of the data sources and ambition of the HRA strategy. It requires assessment of sustainability over time in terms of technological and human resources. |
| 11. Initial assessment of the core predictive modelling | Quality assessment to be carried out immediately after deployment to define further steps leading to a sustainable HRA program |
| 12. Quality assurance program | Continuous quality assurance checking of the HRA program after sustainable adoption |
| 13. Dashboard preparation | Initial identification of key performance indicators (KPIs) to be monitored after adoption, as well as subsequent enrichment of the dashboard with novel KPIs as required. |
| 14. Stakeholders’ engagement | Highly applicable implementation research tools in place to foster engagement of users. The professional profiles will depend on the focus of HRA: policy makers, managers, clinical professionals, etc… |
| 15. Additional functionalities & Roadmap for further developments | Successful HRA adoption leads to initiatives to expand use and ambition requiring development of additional functionalities (predictive modelling) and definition of a roadmap for further developments either in the health policy area, management, clinical applications and/or research-innovation. |
