Diagnoses-related procedure bundles in outpatient care –
an approach using secondary data
Pfeffer N.3), Eisl A.1), Endel F.2), Endel G.3), Filzmoser P.2), Scholler C.3), Weisser A.3)
1)
Vienna University of Economics and Business 2) Vienna University of Technology, 3) Main Association of the Austrian Social Insurance Institutions
[email protected]
BACKGROUND / OBJECTIVES
Currently, one focus of discussion concerning health care reform in Austria lies on strengthening ambulatory health care provision. Consequently, legal
changes aim at fostering the development of new structures in health care (group practices). At the same time, payment reform is being discussed in order
to facilitate the adoption of these new concepts under existing financial constraints. Therefore, a research project was set up to assess the feasiblity of the
available administrative health care data, to develop a statistical toolkit in order to identify diagnoses-related procedure bundles in ambulatory care and to
calculate costs for the procedure bundles. The focus of research lies on the feasibility of diagnosis- and patient-oriented methods of payment for episodes
of care in the Austrian outpatient sector.
DATA
We use a data base of pseudonymous claims data for the years 2006-2007, which allows us to analyze health care provision in outpatient care from a patient‘s,
provider‘s and insurer‘s perspective. The data is linked using a unique patient identifier. Currently, no diagnoses for outpatient care are being coded by doctors. Thus, we use the results of another project, where diagnoses were derived from ATC-Codes – this information was attached to the unique ID according
to a person’s personal record of medication. Moreover, we matched hospital data (procedures, DRGs etc.) from the MBDS data set, using statistical methods.
METHODS
Figure 2
1e+05
(V1) 1 Month
Regression line
upper boundary
Significant difference
Regression line
upper boundary
Significant difference
10000
10000
20000
+
20000
Frequency
50000
+
50000
+
(V1) 1 Month
1e+05
Figure 1
Frequency
When calculating procedure bundles, only costly procedures from outpatient care were included (with regard to tariff x frequency). We limited our
research to a number of common chronic diseases (e.g. diabetes, COPD/
asthma, dementia) and used three different approaches to include patients
in the data sample:
(1) patients with no other disease than the disease in question for t0 +/- 6
months,
(2) patients with no other disease than the disease in question for t0 +/- 1
month
(3) all patients having the disease in question at the time t0, regardless of any
other disease.
Then we applied linear regression to identify those procedures that are significantly related to the single diagnoses in the sample (within a time span
of – 90 days/ +180 days from the diagnosis). As an upper boundary, two
times the standard deviation of the residuals was used. We defined procedure bundles as all procedures to the left of the most significant difference
between two adjacent procedures.
0
10
20
30
40
0
10
20
30
40
Index
Index
Figures 1 and 2 examplify the method to derive diagnoses-related procedure bundles. In Figure 1 the difference between two procedures could be shown to be significant (marked by the red cross). Figure 2 shows
an example where no significant difference between two adjacent procedures could be found.
RESULTS (shown for Diabetes mellitus)
Diabetes
+
+
Frequency
5000 10000 20000
Figure 3
++
500
1000
2000
++
+
Cutoff
Limit for inclusion
Significant difference
0
50
Index
100
150
Code
Frequency
rel. Fr.
Procedure (in German)
1
230501
26027
0.1732
Blutzucker quantitiv
2
230525
9019
0.0597
Gamma-Gt
3
230524
8434
0.0558
GPT (ALAT)
4
230503
8254
0.0546
HbA1 oder HbA1c
5
230505
7686
0.0509
Kreatinin quantitiv
6
230513
7495
0.0496
Gesamtcholesterin
7
230523
7395
0.0489
GOT (ASAT)
8
230512
7169
0.0475
Triglyzeride (Neutralfett)
9
020101
7083
0.0469
Ausführliche diagnostisch-therapeutische Aussprache
10
230506
7012
0.0464
Harnsäure quantitiv
11
020201
6955
0.0460
Blutentnahme aus der Vene
12
230201
6894
0.0456
Komplettes Blutbild: Zählung und Beurteilung
13
230220
6452
0.0427
Blutsenkungsgeschwindigkeit (BSG)
14
230514
6159
0.0408
HDL-Cholesterin inklusive eventueller LDL-Cholesterine
15
231302
5358
0.0355
Streifentest im Harn
16
070101
5298
0.0351
EKG in Ruhe
17
030110
4992
0.0330
Untersuchung mit dem Hornhautmikroskop (Spaltlampe)
18
230517
4925
0.0326
Alkalische Phosphate
19
030109
4476
0.0296
Tonometrie
20
230504
3993
0.0264
Harnstoffe oder Reststickstoffe oder BUN quantitiv
Results show that – for most of the diseases we considered- procedure bundles can be identified using the methods described above. To a large extent,
significant procedures for each diagnosis represent technical procedures,
such as the determining laboratory values or ECGs. In addition, expected
non-technical procedures (e.g. eye treatment for diabetes patients) could
be allocated to the relevant diagnoses. However, we couldn’t find feasible
bundles or those diagnoses where “quasi-unique” ATC-Codes do not exist.
The bundles did not vary substantially across the three methods of including
diagnoses in the sample.
CONCLUSIONS
The obtained results can be seen as a first step towards describing outpatient procedure bundles related to a number of chronic diseases. In a next step,
experts have to refine the bundles. These bundles provide a solid ground for the calculation of costs for diagnosis-related procedure bundles in ambulatory
care. The methods used can be implemented in other data sets as well and are therefore not limited to the context of the Austrian health care system.
REFERENCES
Weisser A., Endel. F., Endel G.: Results of the ATC-ICD project, Poster contribution, PCSI 2010, Munich, Germany
Pfeffer N., Eisl A., Endel F., Endel G., Filzmoser P., Scholler C., Weisser A.: Identification of diagnoses-related procedure bundles in outpatient care using statistical methods, Poster contribution, PCSI 2010, Munich, Germany
20110610_poster_iHEA_pf_V2. 1
29.06.11 11:11