IMPRO project's presentation at Understanding Present and Future Public Service Delivery Costs, a kick-off workshop hosted by OECD, Brussels, June 28th, 2019.
1. 10.7.2019 1
Geospatial costs of delivering health in Finland
1
Understanding Present and Future Public Service Delivery Costs,
a kick-off workshop hosted by OECD, Brussels, June 28th, 2019.
Markku Tykkyläinen – Aapeli Leminen – Mikko Pyykönen – Teppo Repo –
Maija Toivakka – Tiina Laatikainen
University of Eastern Finland, IMPRO, Joensuu
https://www.stnimpro.fi/
https://www.stnimpro.fi/contact-information/
https://www.uef.fi/web/geospatial-health
2. markku.tykkylainen@uef.fi 2
At the moment, the 311
municipalities are
responsible for organizing
health and social services.
This responsibility will be
transferred to about 17
NUTS3 regions and capital
city when the health and
social services reform will
be implemented in the
2020s.
Health and social services
reform
Pilot area
Pilot
area
3. markku.tykkylainen@uef.fi 3
Geospatial Health as an interdisciplinary research in UEF (Univ of Eastern Finland)
Joint aims
Inventive
results for
more efficient
health care
Interaction
and
discovery-
oriented
research
Geographies,
GIScience
Computer
Science,
Statistics,
Geostatistics
Health Sciences
Machine
Learning
Research
Group
Geospatial
Health
Research
Group
Research
Groups in
Health
Sciences
4. markku.tykkylainen@uef.fi 4
Scientific background of geospatial
health research
Operations
Research
(complex
systems, CP,
LP, netw.)
Spatial
Statistics
(n-dim. space,
regr., hot
spots)
Computer
science
(algorithms,
clust., sim.,
progr., db)
GPS,
Georef. info
(coordinates,
orientation) GIScience/
Geoinformatics
(data produc-
tion, rem.
sens.)
Health
Sciences
Spatial
economics
(space-time
dynamics)
1990s-
1950-60s
1990s
Geospatial
Health
Spatial
analysis
(geography,
regional sc.;
math. space)
1990s-
NEGEarly location
and spatial
economic
theorizations
1900-40s
5. Optimal structure
and locations of
health care
centres
Joint use of patient registers and geospatial databases in health care
5
Toivakka, M., Repo, T., Leminen, A., Pyykönen, M., Laatikainen, T. & Tykkyläinen, M. (2018). Potilastieto ja paikkatieto kohtaavat. Terra 130: 4, 201–205.
Geospatial data
Digiroad
Location of health services
Statistical data
Sociodemographics
Environment
Location information according to EUREF-FIN terrestrial coordinate system
markku.tykkylainen@uef.fi
Patient database
Domicile
Diagnoses
Visits
Lab results
Background variables
Scale and accuracy
Data
Location
Individuals
Postal code area
Municipalities LAU-2
or health care centre areas
Statistical squares
250 m x 250 m, 1 km x 1 km
Health care district
Ad hoc or NUTS, LAU
Prevalence of
diseases where,
how much and
diffusion
Accessibility,
travel modes,
travel time, costs
Management
and operational
use
Indicators
Maps &
visualizations
Computational
models
Causal
models
Predictions
Alternatives
Scenarios
Outcome of care,
where, why, how
well, optimal
treatments
Strategic
planning
Tactical
planning
6. Geospatial Health - expertise in IMPRO (UEF-Geography)
markku.tykkylainen@uef.fi 6
Geospatial
optimization;
minimizing
costs of travel
and time loss
• Less traveling > less costs
• Savings in health care with
using more self-monitoring and
remote health consultacy
• Relocation of services based
on patients’ locational data and
the expected use of care
• Considering current & future
prevalence and care in plans
Market area
analysis and
minimization
of the costs
of medication
and care
• Determining least-cost
medication alternatives by
locations of patients
• Making least-cost clinical
guidelines geospatially
based on the total costs
• Societal costs and/or out-of-
pocket costs
• Patient data and areal data combined
• Predicting prevalence and the need of
care by statistical square, postal code
area data and health centre area
• Follow-up of the outcomes of care to
improve care by area
• Considering the impacts of SES,
environment etc. on prevalence and care
• Multifaceted data for planning
Prevalence
and hot spots;
allocating care
accordingly
Small area
analyses for
prevalence
and disease
management
• Detecting - hot spots -
the highest
prevalences and needs
of care and research
• Showing the best
practises
• Minimum maximal
distances (or time)
Patient records, geocoded locations
statistical square data, postal code
areas, municipalities, satellite data,
individual statistical data
Patient records,
geocoded locations,
Digiroad, small-area
data
More efficient health care system
”From where to nowhere” ”Where and why”
”What, where
and
whereabouts”
7. Observing needs and
modeling costs by health
centre area, postal code areas and
statistical squares
maija.toivakka@uef.fi; markku.tykkylainen@uef.fi 7
Patients outside the
ordinary commuter
distance (here 25 km)
358 (3.7 %) non-
shaded
If ≥ 10 km
1958 (20.4 %)
If ≥ 5 km
2923 (30.4 %)
Society would
save money if
there were less
visits and less
costs of travel
and lost time
especially in
remote areas.
Localizations randomized
8. Increasing self-monitoring
and increasing remote
consultancy to lower the
costs of care and
traveling - a model and results
• Assumptions in the
model:
- Travels: home->HC
lab and back
- Patients always go
to the same HC
- Travels done at a
”normal day time”;
no rush times
considered in HC
(and traffic)
aapeli.leminen@uef.fi; markku.tykkylainen@uef.fi
Leminen, A. 2016. Potilaiden liikkumiskustannukset ja omaseurannan kehittämisen kustannusvaikutukset
tyypin 2 diabeteksen seurannassa Pohjois-Karjalassa. http://urn.fi/urn:nbn:fi:uef-20160917
Flow chart and data flows:
HC= Health Centre
9. Mean cost of
one trip in a
zip area
Mean cost
of one trip
by HC
T2D patient: mean travel costs of one follow-up trip to HC
Mean cost
of one trip
by HC
Mean cost of
one trip in a
2kmx2km
square
Mean costs of
one trip in a
2kmx2km
square: 1.70 -
340.00€
Mean costs
one trip in
a zip area:
4.80 -
114.60€
One trip
by HC:
6.40 -
19.60€
BASELINE
Calculations
by individual
is possible
HC= Health Centre
10. aapeli.leminen@uef.fi; markku.tykkylainen@uef.fi
Cost savings:
-Health care and travels from 2,5 mill. € to 1,1 mill. € ( = 56.3 %)
of which:
- Health care costs from 2 128 412 € to 942 587 € (= 55.7%)
-Travelling from 406 967 € to 166 767 € (= 59.0 %)
Results of simulation: the cost savings of type
2 diabetes (T2D) by increasing self-monitoring
and increasing remote consultancy
The highest
travel costs
Lakes
Health care centre
Self-monitoring reduces cumulative travel
costs (but not the costs of one trip).
25 x higher
compared with
cities (or if
using own car)
A. Leminen, M. Tykkyläinen & T. Laatikainen, Self-monitoring induced savings on type 2 diabetes patients’ travel and healthcare costs,
International Journal of Medical Informatics, 115 (2018), 125.
11. Atrial fibrillation (AF) medications
Drugs:
Warfarin +
lab visits
(up 20
p.a.)
Or new
DOAC
The locations of health
care centres (labs) are
on the map on the right.
Warfarin should be
used near the health
care centres.
Now the consumption
patterns of alternative
drugs are NOT cost-
efficient at all (on the
map on the left).
Warfarin should not be
used in the peripheries.
Derivation of the market areas of two alternative drugs – warfarin vs. DOAC (Direct-Acting Anticoagulants)
mikko.pyykonen@uef.fi; markku.tykkylainen@uef.fi 11
12. • Drugs: Warfarin &
DOAC
• The number of
annual lab visits up
to 20 if Warfarin is
used
• Out-of-pocket costs
Fixed costs
= DOAC
W is
cheaper
W+lab
Market areas - a bird's eye
view
DOAC
W 2
W 1
Least-cost optimization of AF treatment
- derivation of market areas
TCwcv1
Areawcv1Areawcv2
cv1
cv2
Pw
TCwcv2
Distance d
Gradients of
the total costs
of Warfarin to
the patient
A
€
AreaDOAC
The sizes of market areas are influenced by drug prices and additionally from one to
two laboratory visits if the patient uses Warfarin.
TCwcv2
TCwcv1
cDOAC
TCDOAC
12mikko.pyykonen@uef.fi; markku.tykkylainen@uef.fi
LAB
LAB
13. The least-cost market areas
of anticoagulation therapies
for working persons by 4
travel modes.
Three market area classes
(coloured by the shades of
grey) having 10, 14 and 18
annual INR monitoring visits
indicate where warfarin is
most affordable.
DOAC therapy brings about
the lowest costs outside of the
respective warfarin market
areas for the patient.
Warfarin is most affordable to use in centres (= in
grey areas) and DOAC in peripheries.
13mikko.pyykonen@uef.fi; markku.tykkylainen@uef.fi
A patient’s out-of-pocket cost includes
the cost of both travels and lost time.
M. Pyykönen, A. Leminen, J. Tynkkynen, M. Tykkyläinen
& T. Laatikainen, A geospatial model to determine the
spatial cost-effectiveness of anticoagulation drug therapy;
patients’ perspective. Forthcoming article
DOAC
W
DOAC
W
14. Coronary heart disease
Population in
2kmx2km
statistical square
Coronary Heart
Disease
Hot Spots
- Cost-efficiency
of emergency
services
- Minimize max
distance (time,
transport mode)
to save lives
- Adjust follow- up
and treatment
according to
distance (time,
transport mode)
Central
Hospital
teppo.repo@uef.fi; markku.tykkylainen@uef.fi
Repo, T., Tykkyläinen, M., Mustonen, J., Rissanen, T. T., Ketonen, M., Toivakka, M. & Laatikainen, T. (2018). Outcomes of Secondary Prevention
among Coronary Heart Disease Patients in a High-Risk Region in Finland. Int. J. Environ. Res. Public Health 2018, 15(4), 724
Non-adjusted cumulative incidence hot-spots of CHD
15. Rural-urban classification is based on 250m x 250m statistical square data. Classification
is based on population density, area density rate, land use, industrial diversity and
accessibility.
15
7-class classification of areas
Numbers of
patients and
their areal
percentage
distribution
Proportions
of HbA1c
measured
patientsa
to
the diagnosed
Proportions
of HbA1c
< 7%
patientsa
to
the measured
Patients’ mean
driving
distances and
the rangesb
in
km
Inner urban area 849 (8.8%) 82.8% 74.8% 2.0 (0 – 4.0)
Outer urban area 1433 (14.9%) 80.5% 75.6% 2.1 (0 – 9.5)
Peri-urban area 644 (6.7%) 85.6% 74.8% 5.0 (0.1 – 27.1)
Local centers in rural areas 1414 (14.7%) 79.9% 69.2% 1.8 (0 – 5.7)
Rural areas close to urban areas 725 (7.5%) 84.6% 71.8% 7.8 (0 – 27.9)
Rural heartland areas 2376 (24.7%) 84.9% 73.1% 6.0 (0 – 36.0)
Sparsely populated rural areas 2165 (22.5%) 83.5% 66.7% 12.1 (0 – 91.8)
Total 9606 (100%) 5.9 (0 – 91.8)
a χ² p-value < 0.05
b Minimum and maximum values in brackets
Toivakka, M., Laatikainen, T., Kumpula, T. & Tykkyläinen, M. (2015). Do the classification of areas and
distance matter to the assessment results of achieving the treatment targets among type 2 diabetes
patients? International Journal of Health Geographics 14: 27. doi: 10.1186/s12942-015-0020-x.
An example of a tailored area classification – Rural-urban
Follow-up %
In therapeutic
equilibrium %
maija.toivakka@uef.fi; markku.tykkylainen@uef.fi
local centre
local centres
Type 2 diabetes
16. Targeting care where needed
Patient characteristics (individual-level variables)
• Female gender is associated with higher HbA1c follow-up rates and a higher proportion of achieving the recommended
HbA1c level.
• The probability of HbA1c measurements increases with ageing. However, younger age increases the probability of
achieving the recommended HbA1c level.
• Distance as such is not a barrier to good control or to achieve treatment targets.
Socioeconomic variables (by postal code area)
• Both of the patient’s education level and the area-level predictor of educational status
explains the outcomes of care.
7-class classification of urban and rural areas
• Best follow-up rates: peri-urban area (suburbs), 2nd in rural heartland areas (farming areas),
3nd rural areas close to urban areas (commuter belt of Joensuu).
• Worst follow-up rates: in local centres, 2nd in outer urban areas (detached urban housing).
• Best outcomes: outer urban area (detached urban housing), 2nd in peri-urban area
(suburbs), and inner urban area (city)
• Worst outcomes: in sparsely pop. rural areas and 2nd local centres
maija.toivakka@uef.fi; markku.tykkylainen@uef.fi 16
This information is useful in targeting care and the selection of the most
cost-effective ways of care.
17. 17
Wikström, K., Toivakka, M., Rautianen, P.,
Tirkkonen, H., Repo, T. & Laatikainen, T. (2019).
Electronic health records as valuable data sources
in health care quality improvement process. Health
Services Research & Managerial Epidemiology.
Health care professionals were
notified of the prevalence of T2D
by municipality (the influence
areas of health centres) and
counselled.
Early detection of T2D improved
and differences between
municipalities decreased (2012 ->
2017).
Best-practise health centre
(Outokumpu) was a milestone.
T2D was more thoroughly
diagnosed.
How do follow-up information and training improve the quality of care locally?
maija.toivakka@uef.fi; markku.tykkylainen@uef.fi
18. IMPRO consortium
18
Geospatial Health Research
Group2019-20, 2021-23
https://www.uef.fi/web/geospatial-health
markku.tykkylainen@uef.fi
IMPRO-consortium Deputy PI heading WP4
aapeli.leminen@uef.fi
maija.toivakka@uef.fi
teppo.repo@uef.fi (Stockholm)
mikko.pyykonen@uef.fi
tiina.laatikainen@uef.fi
IMPRO-consortium PI heading WP3
https://www.stnimpro.fi/
tiina.laatikainen@uef.fi; markku.tykkylainen@uef.fi
19. Publications:
• Pyykönen M., Leminen A., Tynkkynen J., Tykkyläinen, M & Laatikainen T. (2019). A geospatial model to determine the spatial
cost-effectiveness of -anticoagulation drug therapy; patients’ perspective. Submitted.
• Leminen, A., Pyykönen, M., Tynkkynen, J., Tykkyläinen, M. & Laatikainen, T. (2019). Modeling patients’ time, travel, and
monitoring costs in anticoagulation management: societal savings achievable with the shift from warfarin to direct oral
anticoagulants. Submitted.
• Toivakka, M., Pihlapuro, A., Tykkyläinen, M., Mehtätalo, L., & Laatikainen, T. (2018). The usefulness of small-area-based
socioeconomic characteristics in assessing the treatment outcomes of type 2 diabetes patients: a register-based mixed-effect.
BMC Public Health 18:1258. https://doi.org/10.1186/s12889-018-6165-3.
• Repo, T., Tykkyläinen, M., Mustonen, J., Rissanen, T. T., Ketonen, M., Toivakka, M. & Laatikainen, T. (2018). Outcomes of
Secondary Prevention among Coronary Heart Disease Patients in a High-Risk Region in Finland. Int. J. Environ. Res. Public
Health 2018, 15(4), 724; https://doi.org/10.3390/ijerph15040724
• Leminen, A., Tykkyläinen, M. & Laatikainen, T. (2018). Self-monitoring induced savings on type 2 diabetes patients’ travel and
health care costs. International Journal of Medical Informatics 115, 120-127.
• Toivakka, M., Laatikainen, T., Kumpula, T. & Tykkyläinen, M. (2015), Do the classification of areas and distance matter to the
assessment results of achieving the treatment targets among type 2 diabetes patients? International Journal of Health
Geographics, 14:27.
• Sikiö, M., Tykkyläinen M., Tirkkonen H., Kekäläinen P., Dunbar J., Laatikainen T. (2014),Type 2 diabetes care in North Karelia
Finland: Do area-level socio-economic factors affect processes and outcomes? Diabetes Research and Clinical Practice 106,
296-503.
10.7.2019 Esityksen nimi / Tekijä 19