The automatic generation of comprehensive RHCPs/PHC was developed in steps. First, we designed the core structure of the RHCP/PHC report, which comprises 35 indicators covering five health care and health service domains (Table 1) selected in a process involving feedback from physicians and other health care providers. The selection was based on five criteria: (1) legal requirements (e.g. care mandates laid down in the Primary Care Act); (2) availability of regional data (data sources are representative and valid down to the level of “health service regions” (Versorgungsregionen), an Austrian territorial unit level between counties and NUTS-3 regions); (3) the relative stability over time of both indicator magnitude and regional rank, even in a small-scale regional analysis; (4) relevance for the tasks of PHCUs; (5) relevance in the eyes of physicians and other care providers in PHCUs.
Table 1
List of indicators (N = 35) used in the RHCPs/PHC
Demography, Socio-Economics
|
1. Population of the catchment area
|
2. Proportion of children under 14 years (< 15a)
|
3. Percentage of population aged 65 and over
|
4. Percentage of population aged 75 and over
|
5. Percentage of population aged 65 and over in one-person households
|
6. Average income per income-receiver
|
Disease Prevention and Risk Factors
|
1. Percentage with "very bad" or "bad" self-reported health, inhabitants ≥ 15a
|
2. Percentage of persons who smoke daily or occasionally, inhabitants > = 15a
|
3. Percentage of population ≥ 15a with too little exercise
|
4. Percentage of population ≥ 15a with obesity
|
Epidemiology and Mortality
|
1. Life expectancy at birth (men)
|
2. Life expectancy at birth (women)
|
3. Prevalence of diabetes mellitus Type 2
|
4. Prevalence of mental disorders
|
5. Prevalence of disorders of the musculoskeletal system
|
6. Prevalence of chronic head-, neck- and backaches in population ≥ 15a
|
7. Percentage of long-term care benefit recipients / level 1–3
|
8. Percentage of long-term care benefit recipients / level 4–7
|
9. Rate of inpatients with heart disease within 2 years
|
10. Rate of inpatients aged 65 and above with femoral neck fracture within 2 years
|
11. Rate of inpatients with cerebrovascular disease within 2 years
|
12. Rate of inpatients with cancer within 2 years
|
Health Care Service Supply
|
1. Inhabitants per general practitioner
|
2. Percentage of statutory health insurance general practitioners aged 55 +
|
3. Inhabitants per private general practitioner
|
4. Children per paediatrician
|
5. Inhabitants per statutory health insurance internist
|
6. Distance to the nearest acute care hospital (incl. branch hospitals), minutes by car
|
7. Number of pharmacies in the catchment area (excluding hospital pharmacies)
|
8. Distance to the nearest nursing home (minutes by car)
|
9. Inhabitants aged 65 and over per retirement home or nursing home in the catchment area
|
Outpatient Care Utilisation
|
1. Percentage of population visiting statutory health insurance general practitioners
|
2. General practitioner equivalents per 100,000 population (incl. hospital outpatient departments)
|
3. Paediatrician equivalents per 100,000 children (< 15a; incl. hospital outpatient departments)
|
4. Internist equivalents per 100,000 population (incl. hospital outpatient departments)
|
The second step was to define how the information provided by the indicators is presented. Figure 1 shows the visualisation design used in the RHCPs/PCH.
Our visualisation technique is based on the design used in the EU project "I2SARE" (see [4], for example). Currently, the I2SARE presentation technique is routinely used by several EU Member States but has been redesigned in more detail for our report generator. What is new in our approach is, firstly, the freedom it offers to select a municipality for a PHCU anywhere in the country and, secondly, the strategy of using catchment areas that are not bound by district (Bezirk) borders and therefore reflect the expected clientele of the PHCU. Additionally, unlike the representation of distributions common in scientific contexts, we used more intuitive aggregate measures: the total range, the 25 and 75 percentage quantiles and the median to depict the indicator value distribution over Austria’s districts. The visualization of 35 indicators is grouped under five domains and is implemented by box plots and overlaid circles/squares (Fig. 1).
Step one (selection of indicators) and step two (design of the output report) were accompanied by a series of feedback rounds with primary health care providers and experts in the field of primary health care with the aim of optimising the indicator set and making its visualisation intuitively understandable. The application test was carried out together with physicians and non-physician primary health care experts in the initial phase of the project. In order to address new requirements for the RHCPs/PHC (as these can change dynamically), an annual update based on user feedback is foreseen.
The third step was the implementation of an interactive report generator to produce location-specific RHCPs/PHC. The report generator is based on the data and functionalities of an integrative geographical information system already available in the Austrian Health Information System [4] (ÖGIS) but is implemented as a stand-alone application. ÖGIS provides ready-prepared, quality-assured and regionally integrated data sources that can be fed into the report generator after pre-processing. For some indicators new external data sources had to be integrated into the system and linked to those already existing. To provide the report generator with appropriate input the functionalities of ÖGIS had to be extended with SQL-bots generating datapoints for indicators for all possible location municipalities for predefined car travel-time isochrones (= catchment areas). This allows an automated output of the approximately 6,500 datapoints needed as input for each indicator in one single output process. In order to concentrate on the indicators’ key data, a function for the automatic cleaning of outliers has also been implemented. Regional health care profiles combine data at a high level of spatial resolution (i.e. a total of 2,122 Austrian municipalities) and comprise 35 indicators. Indicators for each of the 2,122 municipalities are calculated as averages of all municipalities within a 10-, 15- and 20-minute car travel-time isochrone of the PHCU location municipality. We used mainly SQL scripting to pre-process the large amounts of data and a Microsoft Excel file as the automatic report generator. Record linkage at the level of anonymised individual personal identifiers played a role for four indicators (Rate of inpatients with heart disease within 2 years, Rate of inpatients aged 65 and above with femoral neck fracture within 2 years, Rate of inpatients with cerebrovascular disease within 2 years, and Rate of inpatients with cancer within 2 years).
To generate a report the user must specify a freely selectable location municipality and a maximum car-travel time to define the radius of its catchment area. The report thus generated is a RHCP/PHC – in the format of a comprehensive PDF report.