CODING VARIATIONS in the Discharge Abstract Database (DAD) Data

CODING VARIATIONS in the Discharge Abstract Database (DAD) Data
CODING
VARIATIONS
in the Discharge Abstract Database (DAD) Data
Coding Variations in the Discharge
Abstract Database (DAD) Data
FY 1996–1997 to 2000–2001
May 2003
Contents of this publication may be reproduced in whole
or in part provided the intended use is for non-commercial
purposes and full acknowledgement is given to the Canadian
Institute for Health Information.
Canadian Institute for Health Information
377 Dalhousie Street
Suite 200
Ottawa, Ontario
K1N 9N8
Telephone: (613) 241-7860
Fax: (613) 241-8120
www.cihi.ca
ISBN 1-55392-224-7 (PDF)
© 2003 Canadian Institute for Health Information
TM
Registered Trade-mark of the Canadian Institute for Health Information
Cette publication est disponible en français sous le titre : « Variations de codification dans la Base de données
sur les congés des patients (DAD), exercices de 1996-1997 à 2000-2001 » ISBN 1-55392-225-5 (PDF)
Table of Contents
Acknowledgements .............................................................................................................. i
Introduction and Background ............................................................................................. 1
Introduction to the Project ............................................................................................. 1
Coding Variation in the CIHI DAD .............................................................................. 2
Diagnosis Typing and the Complexity Overlay.............................................................. 3
Baseline Comparisons 1996–1997....................................................................................... 8
Choice of 1996–1997 as Baseline Year........................................................................... 8
Changes from 1996–1997 to 2000–2001 .......................................................................... 18
Comparisons for 2000–2001.............................................................................................. 27
Understanding Changes in Ontario .................................................................................. 34
Focus on Ontario .......................................................................................................... 34
Ontario Diagnosis Typing ............................................................................................. 38
Ontario Grade List Diagnosis Analyses........................................................................ 41
Ontario Analysis Conclusions ...................................................................................... 47
Sensitivity to Complexity Methodology Changes............................................................. 50
Grade List Revisions...................................................................................................... 50
Impact on 2001–2002 Ontario Data ............................................................................ 53
Use of Revised Grade List Grouper to Assess All DAD Data ..................................... 55
Conclusions from Analysis Results.................................................................................... 62
Review of Results .......................................................................................................... 62
Lessons Learned ............................................................................................................ 64
Monitoring and Reporting Variation in the DAD........................................................... 65
Potential Indicators....................................................................................................... 66
i
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Acknowledgements
Coding Variations in the Discharge Abstract Database (DAD) was made possible by
the project leadership of Chris Helyar, Hay Health Care Consulting Group and analysis
by Bill Croson, associate consultant with the HayGroup.
Project support was provided from the Canadian Institute for Health Information (CIHI)
under the direction of Nizar Ladak, and a project team comprising Frank Ivis, Sukanya
Gopinath, Micheline Mistruzzi, Nicole DeGuia, Holly Bartoli, Craig Homan, Brenda Antliff,
Andre Lalonde, Jason Sutherland, Greg Webster, Sheril Perry and Michael Cohen.
In February 2003, a draft of this report was circulated as background reading material for
meetings that took place in February and March 2003 in selected provinces. This final
version benefited from external reviewer comments and those discussions. In particular,
CIHI acknowledges Andrée Martin Desjardins, Shirley Groenen, Gordon Kramer and
Helen Whittome along with their respective teams, and the Executive staff of the Canadian
Health Records Association (CHRA) for their detailed and thoughtful comments.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
1
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Introduction and Background
Introduction to the Project
Recent Research Identified Large Changes in CIHI Acute Care Data
Recent research conducted by CIHI, the Ontario Ministry of Health and Long-Term Care
and the Ontario Joint Policy and Planning Committee (JPPC), identified unusual changes
in reported acute care patient discharge abstract data for individual hospitals. The
magnitude of some of these changes, particularly with respect to weighted cases, and the
increasing use of Resource Intensity Weights or RIWTM weighted cases for hospital
funding, has raised the question as to whether these changes reflect:
•
changes in the clinical complexity1 of patients seen in hospital, or
•
changes in comprehensiveness and quality of clinical documentation, or
•
changes in hospital health records coding and abstracting practices.
Report Presents Results of Broader Analysis of DAD Data
The Ontario experience has prompted CIHI to closely examine its coding standards and
grouping methodologies and to conduct a broader investigation of the potential variations
in the comprehensiveness and comparability of the data in the Discharge Abstract
Database (DAD). The quality assurance processes applied to the DAD are described in a
recent CIHI publication2 and the results of the 2-year CIHI DAD re-abstraction study
were recently published.3,4 This report presents the results of the parallel investigation of
variations in the DAD data.
1
Complexity refers to diagnoses other than the most responsible that prolong length of stay and where most costly
treatment is reasonably expected.
2
CIHI, Quality Assurance Processes Applied to the Discharge Abstract and Hospital Morbidity Databases, August 2002.
3
CIHI, Discharge Abstract Database Data Quality Re-Abstraction Study—Combined Findings for Fiscal Years 1999/2000
and 2000/2001, December 2002.
4
Re-abstraction studies can identify whether the process of extraction of information from the medical record was done
accurately and in compliance with CIHI guidelines. What they cannot do is assess whether the contents of the medical
record are accurate. No retrospective analysis can definitively prove or disprove the accuracy of the contents of the
medical record. This could only be done at the time a patient is in hospital and with the same assessment and
investigation conducted by the clinicians that provide the documentation in the medical record.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
2
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Goals of Analyses
The initial goals of the analyses of the DAD, with a focus on the assignment of
complexity, were to:
1. confirm and document overall changes in data submitted to DAD;
2. confirm and quantify apparent changes in Ontario data versus rest of Canada;
3. assess contributing factors to the changes in the Ontario DAD data; and
4. provide background data, analyses, and findings to support development and
refinement of CIHI data quality methodologies and initiatives.
During the project the opportunity arose (through parallel JPPC activities) to more
specifically assess data reporting practices of individual Ontario hospitals and to identify
opportunities to create new measures to support this assessment. This activity is also
documented in this report.
All of these analyses are intended to support ongoing CIHI efforts to maintain and
enhance the quality of the data in all patient-specific databases, and to assess the
potential impact of coding variation on data comparability in the DAD.
Coding Variation in the CIHI DAD
There has been increasing concern that variation in abstracting and coding practices
across the CIHI DAD hospitals could compromise the comparability and utility of the
DAD data. Whether coding variation results from initiatives designed to maximize RIW
value or to optimize a hospital’s relative position, is of no direct consequence to CIHI. Of
concern to CIHI is the fact that variation in coding practices compromises comparability
over time and within regions and provinces. Variation in coding practices could
compromise the comparability of morbidity data, which could in turn call into question
the validity of analysis and reporting efforts by CIHI, Statistics Canada, and others.
Definition of Upcoding
For the purposes of this document, we have used the term “upcoding” to refer to the
recently observed changes in coding, since the variation found has typically resulted in an
increase in the use of selected codes in defining patient complexity. Upcoding can be
defined as:
Change in the apparent complexity and/or care requirements of the
patients separated from a hospital, attributable to changes in the
comprehensiveness or categorization of the data reported on the
discharge abstract, and not attributable to actual changes in patient
characteristics and/or care requirements.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
3
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
It should be noted here that “undercoding”, referring to a lack of comprehensiveness in
coding, is also of concern in maintaining comparability of data. However, our analysis
found that it was increases in the reporting of codes that indicate patient complexity
which were increasing variation and therefore warranted our attention. It should also be
noted that there is no intended connotation of good or bad in the use of the term
upcoding. Instead, upcoding simply refers to the uni-directional increase in the reporting
of codes that help define patient complexity.
The analysis performed by CIHI focused on coding variation that compromised
comparability of data. There are other techniques such as clinical audits involving chart reabstraction that can be used to assess the reasons underlying change in coding practices.
However, these techniques were not employed in this analysis and therefore any conclusions
about the nature or intent that resulted in the coding variation could not be made.
Diagnosis Typing and the Complexity Overlay
To understand the purpose and results of the analyses presented in this report, it is
necessary to understand the process of “diagnosis typing” and how the CIHI complexity
overlay is used to identify patients with additional diagnoses that may lead to unusually
long lengths of stay or high cost.
Diagnosis Typing
All diagnoses recorded on a DAD in-patient record must have a diagnosis type. The
main diagnosis that can be described as having been most responsible for the patient’s
stay in the hospital is the Most Responsible Diagnosis (MRDx). The MRDx determines
the assignment of a record to a Major Clinical Category (MCC) and in conjunction with
the principle procedure (for surgical cases) determines the assignment to a Case Mix
Groups or CMGTM.
A Type 3 diagnosis (secondary) is a diagnosis for which a patient may or may not have
received treatment, but does not satisfy the requirements to be considered a comorbidity.
The two comorbidity types are defined as follows:5
•
Pre-Admit Comorbidity—Type 1 Dx. A co-existing condition presents prior to
admission that has a significant influence on the patient’s length of stay or
significantly influences the management/treatment of the patient while in hospital.
•
Post-Admit Comorbidity—Type 2 Dx. A condition arising during the hospital
observation or treatment that has a significant influence on the patient’s length
of stay or significantly influences the management/treatment of the patient while
in hospital.
5
The 2002 diagnosis typing standard was clarified in a CIHI bulletin published in January 2003. The definitions here reflect
the guidelines in place prior to fiscal year 2002–2003.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
4
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Type 1 comorbidities are often used to support risk adjustment methodologies used with
CIHI data grouped into CMG. Type 2 comorbidities have been used to measure
complication rates for hospital quality measurement.
Complexity Methodology
The CIHI Complexity Methodology or PlxTM was introduced in response to the concern
that CMG were not always sufficiently clinically and/or statistically homogeneous.
Categories based on the original DRG concept (such as CMG), driven by single diagnosis,
may not be sensitive to differences in the burden of illness, patient age, or severity of
illness, a specific patient may have.
The CIHI Plx methodology (including segregation of cases into three age bands) was
introduced for fiscal year 1997–1998 after a 3-year project to identify mechanisms to
improve CMG clinical and statistical performance. The Plx methodology assigns a
complexity level to each in-patient case:
•
1 – No complexity
•
2 – Complexity related to chronic condition(s)
•
3 – Complexity related to serious/important condition(s)
•
4 – Complexity related to potentially life-threatening condition(s)
•
9 – Complexity not assigned
Some types of patients do not have complexity assigned (e.g. 112 out of 472 CMG,
in Obstetrics, Neonates, Mental Health, Trauma) and are given complexity level 9
by default.
The assignment of a complexity level to a patient record is dependent on the diagnoses
(beyond the MRDx) recorded. Only Type 1 and 2 diagnoses are used to assign complexity
levels. Any medical or surgical patient record with no Type 1 or 2 diagnoses will be
assigned to complexity level 1.
Not all Type 1 and 2 diagnoses will impact the complexity assignment. During the
development of the complexity methodology CIHI identified a list of selected diagnoses
(the 440 “Grade List” diagnoses) with significant impact on the length of stay of a patient.
A patient must have at least one Type 1 and/or 2 grade list diagnosis to be assigned a
complexity level higher than 1.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
5
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Highest Volume Grade List Diagnoses in CIHI DAD (2000)
ICD-9
Code
Dx Name
Volume
4019
Essential Hypertension Unspec
76,264
4280
Congestive Heart Failure
53,390
4273
Atrial Fibrillation And Flutter
50,774
2859
Anemia Unspecified
45,395
5990
Urinary Tract Infect Site Nos
44,076
486
Pneumonia Organism Unspecified
32,419
411
Oth Ac/Subac Ischemic Heart Dis
23,846
2851
Acute Posthemorrhagic Anemia
22,901
410
Acute Myocardial Infarction
21,340
5119
Unspecified Pleural Effusion
20,682
2768
Hypopotassemia
19,535
7806
Pyrexia Of Unknown Origin
18,148
The impact of grade list diagnoses on the complexity assignment for an individual
in-patient depends on:
•
the “class” of diagnosis :
− A [life threatening]
− B [important LOS impact]
− C [chronic disease]
− D [debilitating condition]
− P [psych dx];
•
the diagnosis type (Type 1 versus Type 2);
•
the number and mixture of class of diagnoses; and
•
whether the grade list diagnoses are in same MCC as the MRDx.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
6
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Why Was Complexity Introduced?
The complexity overlay was introduced to respond to the concern that CMG were not
sufficiently sensitive to differences in the burden of illness, patient age, or patient
complexity. The grade list diagnoses were selected on the basis of their impact on the
patient length of stay (and resource use). The following chart shows the average length of
stay for Typical6 patients discharged in fiscal year 1996–1997 by complexity level. As
would be expected, the average in-patient length of stay increases as the assigned case
complexity increases.
1996 Average Length of Stay for Typical Patients in the CIHI DAD by Assigned
Case Complexity Level
22.7
10.9
8.2
3.8
1
3.9
2
3
4
9
It would also be expected that in-hospital mortality would be higher for patients with
higher complexity levels (particularly for complexity level 4, patients with life-threatening
illness). The following chart (which is based on 1996–1997 CIHI DAD data) shows that
the percent of acute care in-patients that die in the hospital increases from 2.1% for
complexity level 1 patients to 31.3% for complexity level 4 patients.
6
A Typical patient is defined by CIHI as being an inpatient in an acute care hospital who has a full course of acute
treatment and who is not a long-stay outlier, a transfer to or from another acute care hospital, and who does not die in the
acute care hospital or sign themselves out against medical advice.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
7
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
1996 Average Percent In-hospital Mortality in the CIHI DAD by Assigned Case
Complexity Level
31.3%
16.5%
8.3%
2.1%
0.3%
1
2
3
4
9
The longer length of stay for higher complexity patients is reflected in the RIW values
assigned to these patients. The chart below shows the RIW values assigned to Typical
patients, aged 18 to 69, in the Craniotomy CMG.
CMG/RIW 2000 Typical RIW for CMG 1 Craniotomy (age 18 to 69)
by Complexity Level
Typical RIW
9.02
2.28
1
3.23
4.12
2
3
4
Complexity Level
Sensitivity of RIW Assignment to Complexity
A Typical craniotomy patient with no grade list comorbidities will be assigned an RIW of
2.28. Depending on the number and mixture of comorbid grade list diagnoses, the
assigned weight can increase almost four-fold, to 9.02 weighted cases. As will be shown
later in this report, the sensitivity of weighted case assignment to complexity level has
generated a focus on understanding variation in rates of recording of grade list diagnoses
and the impact of this variation on comparability of DAD data.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
8
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Baseline Comparisons 1996–1997
Choice of 1996–1997 as Baseline Year
The baseline for the review of changes in the CIHI DAD is the data for the 1996–1997
fiscal year.7 The 1996–1997 fiscal year has been used as the baseline because:
•
1996–1997 is the earliest year of data available that has been re-grouped using the
CMG 2000 grouper and the RIW 2000 case weighting methodology.8
•
1996–1997 was the last year of data submitted to the CIHI DAD before full
introduction of the Complexity or PlxTM methodology. It has been argued that the
introduction of Plx resulted in a change to coding practices as concepts of mandatory
verses optional reporting of comorbidities emerged. However, the 1996–1997 data
should reflect the state of the DAD prior to the introduction of these practice
changes. The selection of this baseline is also consistent with the aim of this analysis
which is to examine variations and the consistency of coding practices across hospitals
and provinces.
The table on the following page shows summary statistics for the 1996–1997 DAD data
by province.
Records Assigned to Province Based on Hospital Location
For purposes of all of the analyses in this report, patient records are assigned to a province
or territory on the basis of the location of the acute care hospital. For example, the
hospital record for a resident of Saskatchewan who was hospitalized in British Columbia is
included in the British Columbia data. In most cases the vast majority of hospital care for
the residents of a region is provided by a hospital located in that region. However, for the
residents of the territories9 most of their tertiary and quaternary hospital care is provided
in southern hospitals. The shorter length of stay (LOS) and lower weighted cases for the
territories will reflect only the hospital care provided within the territories.
7
All references to a year in this report refer to the fiscal year beginning on April 1st. Where a single year is identified it refers
to the fiscal year that began in that calendar year (e.g. 1996 refers to fiscal year 1996–1997 and 2000 refers to fiscal year
2000–2001.
8
Unless otherwise stated, all data shown as been consistently grouped using CMG 2000.
9
Data for the Yukon, the Northwest Territories, and Nunavut have been combined because of low volumes in each
individual territory.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
9
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
1996 CIHI DAD Data by Province
Province
IP Cases
IP Days
ALOS
% ALC
RIW per RIW
SDS as % RIW per
SDS Cases
Case
per Day
of Total SDS Case
N.L.
72,827
539,431
7.41
4.3%
1.34
0.181
41,199
36.1%
0.192
P.E.I.
16,382
119,756
7.31
2.0%
1.24
0.169
9,147
35.8%
0.172
N.S.
120,147
874,700
7.28
3.2%
1.40
0.192
77,433
39.2%
0.177
N.B.
118,941
753,047
6.33
0.6%
1.15
0.182
46,446
28.1%
0.192
Ont.
1,222,354
7,459,992
6.10
9.6%
1.29
0.211
960,622
44.0%
0.178
Man.
87,659
834,072
9.51
14.5%
1.82
0.191
69,077
44.1%
0.189
Sask.
144,827
869,883
6.01
2.7%
1.18
0.197
67,148
31.7%
0.191
Alta.
332,165
1,860,276
5.60
3.9%
1.23
0.220
N/A
N/A
N/A
B.C.
453,165
2,747,185
6.06
6.9%
1.25
0.206
262,518
36.7%
0.198
11,832
46,599
3.94
1.8%
0.76
0.193
4,146
25.9%
0.192
2,580,299 16,104,941
6.24
7.4%
1.28
0.206
1,537,736
37.3%
0.183
Territories
Total
Case Volumes
In 1996, there were 2.6 million in-patient cases in the CIHI DAD and 1.5 million
ambulatory procedure (SDS) cases. Almost half (47%) of the in-patient cases were from
Ontario hospitals and 62% of the ambulatory procedure cases were from Ontario. Because
ambulatory procedure cases were not consistently reported to CIHI by Alberta hospitals,
all of the Alberta SDS records have been excluded.
For most provinces the DAD data reflects all of the acute care in-patient cases.
Exceptions are Quebec (where hospital separation data is reported to MED-ÉCHO
rather than CIHI) and Manitoba, where only the largest hospitals reported their data to
CIHI. Ambulatory procedure activity is also less comprehensive for provinces such as
Alberta and New Brunswick.
Length of Stay
In 1996, the longest average acute care hospital length of stay was in Manitoba
(9.51 days) and the shortest was in the territories (3.94 days). However, both of these
length of stay values do not reflect the average length of stay for hospitalizations of most
of the residents of the region (since the Manitoba data is for the hospitals treating the
most complex cases and the territory data excludes complex patients treated elsewhere).
For the provinces where the data reflects patterns of care for most of the residents, the
longest LOS was in Newfoundland and Labrador and the shortest LOS was in Alberta.
Average acute care hospital length of stay tended to be shorter in the western provinces
than in the east (a geographic pattern also seen in the United States).
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
10
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
ALC Days
In 1996, there was large variation in the percentage of total in-patient days reported as
Alternate Level of Care (ALC), from a low of less than 1% of days in New Brunswick, to a
high of 14.5% of days in Manitoba. This may reflect provincial variation in the capacity of
post-acute placement options (e.g. home care or long-term care beds) and/or variation in
the comprehensiveness of reporting of discharge placement delays. In other words, there is
national variation in the comprehensiveness of reporting of ALC days and not all facilities
in New Brunswick reported ALC days.
Resource Intensity Weights
The average RIW weighted cases per in-patient separation ranged from a low of 0.76 in
the territories to a high of 1.82 in Manitoba. For the provinces for which the data shows
the hospitalizations of most residents, the lowest average RIW per case was in New
Brunswick (1.15) and the highest in Nova Scotia (1.40). The average RIW per case is a
measure of the relative resource use per case.
A second measure of the relative resource use is the average RIW per day. Because RIW
values for long-stay outliers are partially assigned on a per diem basis, the average RIW
per case may be high because a hospital has long lengths of stay. The average RIW per day
is a better measure of the relative daily intensity of in-patients. In 1996, Prince Edward
Island had the lowest average RIW per day (0.169) and Alberta had the highest average
RIW per day (0.220).
Percent Use of Ambulatory Procedures
A gross measure of the extent to which the hospitals in a province have replaced
in-patient surgery with ambulatory procedures is the percent of total cases in the DAD
that were reported through the ambulatory procedure reporting system.10 The percent use
of ambulatory procedures as it applies to the 1996–1997 baseline data year are being
examined in that it may also reflect differences in data collection and reporting and
therefore contribute to the variation found in this analysis. This is particularly relevant for
provinces like New Brunswick and Ontario where some ambulatory procedures do not get
reported to CIHI. In 1996, (for those provinces where the data reflect hospitalizations of
most residents) the lowest ambulatory procedure rate was 28.1% of all (in-patient and
ambulatory procedure) cases for the hospitals in New Brunswick. The highest ambulatory
procedure rate was 44.0% for the hospitals in Ontario.
The average RIW value per ambulatory procedure (a measure of the relative cost intensity
of the ambulatory procedures) ranged from a low of 0.172 in Prince Edward Island
hospitals to a high of 0.198 in British Columbia hospitals.
10
Not all activity reported to CIHI as an ambulatory procedure is surgical in nature.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
11
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The differences in the measures shown above, derived from the CIHI DAD data, likely
reflect a combination of differences in the structure, capacity, and role of acute care
hospitals and the broader health system between provinces. However, as already stated,
they may also reflect differences in data collection and reporting.
Diagnoses per In-patient Case
The following chart shows the variation across provinces in the average number of
diagnoses, by type, reported to the CIHI DAD for in-patient separations. Every case must
have a Most Responsible Diagnosis (MRDx) and the numbers in the chart exclude the
MRDx. The diagnosis types shown are Type 1 (pre-admit comorbid condition), Type 2
(post-admit comorbid condition), and Other (mainly Type 3 or secondary diagnoses).
1996 Average Diagnoses per In-patient Case by Province
Total
Territories
B.C.
Alta.
Sask.
Man.
Ont.
N.B.
T1 Dx per Case
N.S.
T2 Dx per Case
P.E.I.
Other Dx per Case
N.L.
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
In 1996, Alberta hospitals reported almost three diagnoses per in-patient case, and more
comorbid diagnoses per case than the hospitals in any other province. The actual values
by diagnosis type by province are shown in the following table.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
12
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
1996 Average Diagnoses per Case by Diagnosis Type by Province
Province
T1 Dx
T2 Dx Other Dx Total Dx
per Case per Case per Case per Case
N.L.
0.44
0.09
1.21
1.75
P.E.I.
0.55
0.07
0.97
1.59
N.S.
0.65
0.15
1.60
2.40
N.B.
0.68
0.07
0.92
1.67
Ont.
0.87
0.11
0.93
1.91
Man.
0.78
0.15
1.33
2.26
Sask.
0.75
0.11
1.05
1.91
Alta.
1.08
0.16
1.66
2.89
B.C.
0.79
0.10
0.89
1.79
Territories
0.46
0.05
0.67
1.18
Total
0.84
0.12
1.07
2.03
In 1996, Newfoundland and Labrador hospitals reported the lowest number of
pre-existing comorbidities (Type 1 diagnoses) per in-patient case and one of the lowest
number of post-admit comorbidities (Type 2 diagnoses) per case. However, these same
Newfoundland and Labrador hospitals had a relatively high reporting rate for other
diagnoses per case.
Factors Influencing Diagnosis Reporting
The factors that might influence the reported diagnoses per in-patient case in the CIHI
DAD include:
•
Patient health and burden of disease. For those provinces where the data reflects
the majority of the hospitalizations of the residents, we would not expect to see
great variation in the population health status contributing to the variation in
reported diagnoses.
•
Incentive to comprehensively capture and report diagnostic information. Hospitals in
jurisdictions where hospital funding uses discharge separation abstract data or where
there is history of performance measurement using such data may have a greater
incentive to report more diagnoses.
•
Availability of health records resources. There is a cost to identifying and coding
diagnoses and if health records resources are limited, diagnoses may be less
comprehensively reported.
•
Provincial variation in data capture and reporting guidelines (or variation in
awareness and understanding of CIHI standards).
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
13
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
•
Documentation practices that are either conducive to or detract from comprehensive
data capture and reporting of diagnoses.
The variation in reporting of Type 1 and Type 2 diagnoses has the greatest potential
impact on comparability of DAD data, since the comorbidity diagnosis data may be used
to assign case weights, risk adjust for quality measurement, and to measure the incidence
of complications.
Grade List Diagnoses
In 1996, only a small subset of reported diagnoses were grade list diagnoses that would
lead to the assignment of higher complexity levels.11 The average overall DAD ratio of
non-grade list diagnoses to grade list diagnoses was 12.5.
1996 Ratio of Non-Grade List Diagnoses to Grade List Diagnoses*
Grade List Dx
per Case
Non-Grade
List Dx per
Case
Ratio of NonGrade List to
Grade List
N.L.
0.10
1.65
17.2
P.E.I.
0.09
1.50
16.6
N.S.
0.14
2.26
16.4
N.B.
0.10
1.57
15.5
Ont.
0.16
1.76
11.3
Man.
0.18
2.09
11.8
Sask.
0.12
1.79
14.8
Alta.
0.20
2.70
13.7
B.C.
0.14
1.65
11.8
Territories
0.04
1.14
31.0
Total
0.15
1.88
12.5
Province
* Please note that numbers have been rounded-off for presentation
purposes. Calculations were performed with 16 decimal places.
11
The complexity methodology was not routinely applied to all DAD data until fiscal year 1997–1998, although the
methodology was widely publicized and applied on an ad hoc basis prior to that year.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
14
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
There was a two to one difference between Alberta, Newfoundland and Labrador,
Prince Edward Island, and New Brunswick in grade list diagnoses per case. It is unlikely
that this truly reflects differences in the incidence of the grade list conditions for hospital
in-patients between the provinces. This is more likely the result of differences in
comprehensiveness of reporting of grade list diagnoses. In 1996–1997, Alberta hospitals
had already had years of experience with hospital funding based on discharge data,12 and
were developing discharge data based hospital system performance measures.
A primary use of the CIHI DAD by acute care hospitals is the establishment of length of
stay targets and the monitoring of length of stay performance. CIHI uses the DAD to
calculate an expected LOS for Typical patients, and CIHI reports routinely compare the
actual LOS for a patient with the expected LOS (ELOS). The following table shows, for
Typical patients, the 1996–1997 ratio of the actual LOS (ALOS) to the ELOS, by
province. The ELOS is calculated after removal of the reported ALC days, so that ALOS
used for comparison here also excludes the reported ALC days.
1996 Typical Case LOS Performance by Province
Province
IP Cases
IP Days
ALOS
(incl.
ALC)
ALOS
ALOS as
(excl. % ALC % of
ALC)
ELOS
N.L.
58,662
296,727
5.06
5.04
0.4%
123%
P.E.I.
13,272
69,838
5.26
5.23
0.6%
124%
N.S.
97,978
500,198
5.11
5.08
0.6%
116%
N.B.
99,230
479,914
4.84
4.83
0.1%
116%
Ont.
1,055,062
4,734,047
4.49
4.38
2.4%
100%
Man.
70,951
353,102
4.98
4.91
1.4%
111%
Sask.
117,826
519,424
4.41
4.39
0.5%
108%
Alta.
281,797
1,188,510
4.22
4.17
1.1%
96%
B.C.
375,948
1,672,660
4.45
4.39
1.4%
101%
10,415
33,354
3.20
3.18
0.5%
99%
2,181,141
9,847,774
4.51
4.44
1.7%
102%
Territories
Total
In 1996–1997, the overall actual length of stay for Typical in-patient cases in the DAD
was 102% of the ELOS.13 Only Alberta and territorial hospitals had actual Typical lengths
of stay shorter than the DAD ELOS. Typical lengths of stay in hospitals in the Atlantic
Provinces were all at least 16% longer than the expected LOS. The Ontario Typical LOS
was equal to the ELOS.
12
The Alberta Acute Care Funding system in the early 1990’s used the U.S. refined DRG (RDRG) grouper and weights,
rather than the CIHI (or HMRI, the predecessor to CIHI) CMG grouper and weights.
13
The ELOS is based on CMG 2000.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
15
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
When the LOS performance is broken down by broad program14 (medicine, surgery,
obstetrics/neonates, and psychiatry) in some cases it differs from the provincial average,
particularly for Psychiatry.
1996 Typical Case LOS Performance (as % of Expected LOS)
by Program by Province
Province
Medicine
Surgery
Obs/Neo
Psych
N.L.
119%
127%
127%
124%
P.E.I.
123%
129%
138%
109%
N.S.
119%
116%
115%
109%
N.B.
116%
117%
127%
108%
Ont.
99%
101%
101%
97%
Man.
108%
109%
112%
129%
Sask.
105%
108%
124%
102%
Alta.
98%
93%
92%
105%
B.C.
97%
104%
107%
100%
Territories
95%
107%
115%
83%
102%
103%
105%
101%
Total
While the Alberta LOS performance was below 100% for Medicine, Surgery, and
Obstetrics/Neonates, it was 105% for Psychiatry. The Prince Edward Island LOS
performance was 109% for psychiatry but above 120% for the other programs. These
differences may reflect provincial difference in the structure of the mental health system
(e.g. availability of tertiary mental health services, categorization of mental health beds as
acute care, and community service capacity).
In 1996, 30% of in-patient cases submitted to the CIHI DAD were in CMG to which the
complexity methodology did not apply. The distribution of the remaining cases, by
complexity level, and by province, is shown in the following table.
14
Cases were assigned to programs as follows: MCC 19 cases were assigned to Psychiatry; MCC 14/15 cases were assigned
to Obstetrics/Neonates. Any of the remaining cases with an operative procedure were assigned to Surgery, and all other
cases were assigned to Medicine.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
16
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
1996 Distribution of In-patient Cases by Complexity Level
N.L.
23.5%
Distribution of Cases by Plx
(excl. Plx 9)
Plx 1
Plx 2 Plx 3 Plx 4
89.0%
6.3% 2.8% 1.8%
P.E.I.
28.9%
87.5%
8.0%
2.9%
1.6%
N.S.
26.4%
84.0%
8.9%
4.1%
2.9%
N.B.
21.6%
87.8%
7.6%
3.1%
1.5%
Ont.
31.4%
81.1%
10.7%
5.0%
3.2%
Man.
35.3%
78.6%
11.1%
5.7%
4.5%
Sask.
24.8%
85.7%
8.4%
3.6%
2.2%
Alta.
31.8%
79.5%
11.0%
5.5%
4.0%
B.C.
29.6%
83.6%
9.3%
4.3%
2.7%
Territories
40.4%
93.1%
5.1%
1.3%
0.4%
Total
30.0%
82.4%
9.9%
4.7%
3.0%
Province
% of Total
in Plx 9
Overall in 1996, 82.4% of medical/surgical cases in the DAD were assigned a complexity
level 1 (no complexity). Manitoba had the highest percent of medical/surgical cases in
complexity level 4 (life threatening condition) at 4.5%, followed by Alberta at 4.0%.
We would expect that there would be some correlation between the percent of
medical/surgical cases that are assigned to complexity level 4 and the actual percent of
medical/surgical cases that die in hospital. The following graph compares the complexity 4
and in-hospital mortality rates. Generally, for those provinces where the percent of
patients in complexity level is higher, the actual percent of in-hospital deaths is also
higher. The exception is Alberta, which has the 2nd highest percent of medical/surgical
patients in complexity level 4, but the 2nd lowest actual percent of in-hospital deaths.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
17
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Percent Complexity Level 4
1996 Percent of Medical/Surgical Cases in Plx 4 Versus Actual In-hospital Mortality
for Medical/Surgical Cases
5.0%
4.5%
4.0%
3.5%
3.0%
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
Man.
Alta.
Ont.
Sask.
N.B.
B.C.
N.S.
N.L.
P.E.I.
Territories
0%
1%
2%
3%
4%
5%
6%
Percent In-hospital Deaths
1996 Data Conclusions
For fiscal year 1996–1997, prior to the implementation of the CIHI complexity overlay,
there were apparent differences between provinces in use of acute care facilities and in
reporting of patient separation data to CIHI. Some of these differences were caused by
differences in the role of acute care hospitals and the availability of other types of hospital
beds and health services. Other differences were likely caused by measurement difference
(e.g. comprehensiveness of diagnostic data, interpretation of coding guidelines). Some of
this variation is inevitable. The goal of this analysis and assessment of the DAD data for
the subsequent years (fiscal year 1997–1998 to 2000–2001) is not to determine whether
there is any data reporting variation between provinces, but rather to determine whether
the variation is significant enough to compromise the comparability of the data.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
18
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Changes from 1996–1997 to 2000–2001
In addition to the fiscal year 1996–1997 DAD data, DAD data were analyzed for the
fiscal years from 1997–1998 through to 2000–2001. All of these data were grouped using
CMG 2000 (the same grouper used with the 1996–1997 data). When the 1997–1998
through 2000–2001 data were collected, the complexity methodology was routinely
applied to all acute care in-patient records, and all CIHI clients received reports showing
their length of stay performance and RIW weighted cases based on complexity. The
following table shows the percent change in key acute care activity measures from
1996–1997 to 2000–2001.
Percent Change from 1996 to 2000 in Overall DAD Activity Measures by Province
Province
IP Cases IP Days
ALOS
% ALC
RIW per RIW per
Case
Day
SDS as RIW per
% of
SDS
Total
Case
12%
16%
-7%
SDS
Cases
N.L.
-12%
-6%
7%
-7%
6%
-1%
P.E.I.
13%
20%
6%
114%
7%
0%
6%
-4%
-2%
N.S.
-11%
0%
12%
157%
14%
2%
25%
21%
8%
N.B.
-8%
1%
9%
105%
15%
6%
44%
35%
-8%
Ont.
-7%
-7%
0%
-5%
12%
11%
18%
14%
2%
Man.
-13%
-16%
-3%
-56%
-2%
1%
-7%
4%
6%
Sask.
1%
-2%
-4%
-4%
1%
4%
49%
28%
-5%
Alta.
1%
15%
13%
52%
11%
-2%
N/A
N/A
N/A
B.C.
-8%
2%
10%
118%
11%
1%
10%
11%
5%
-10%
-4%
7%
32%
9%
2%
23%
25%
-1%
-6%
-2%
4%
17%
10%
5%
18%
15%
2%
Territories
Total
From 1996–1997 to 2000–2001, there was a 6% decrease in in-patient cases in the DAD.
All provinces exhibited a decrease except for Alberta and Saskatchewan, which had a 1%
increase, and Prince Edward Island, which had a 13% increase.
LOS Changes
There was a 4% increase in the average length of stay of the patients represented in the
DAD data. Some of this increased length of stay is accounted for by the 17% increase in
the percent of days used by ALC patients. The very large increases in percent ALC days
in some provinces (more than 100%) probably represents more complete reporting of
ALC days.
The average RIW per case increased in all provinces except Manitoba. The average
RIW per in-patient day increased in all provinces except Newfoundland and Labrador
and Alberta.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
19
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Ambulatory procedure volumes increased by 18% overall and in all provinces except
Manitoba. The change in average RIW per ambulatory procedure was mixed, with some
provinces higher and other lower.
The changes in the activity measures from 1996–1997 to 2000–2001 are consistent with
an overall pattern of improved utilization of acute care beds, with fewer in-patient cases,
more ambulatory procedures, and a longer residual LOS (and higher resource intensity)
for the cases not shifted to ambulatory care.
The graph below shows the trend in average RIW per in-patient case.
Average RIW per In-patient Case Trend from 1996 to 2000 by Province
2.00
1.80
1.60
1.40
1.20
1.00
0.80
0.60
1996
1997
1998
1999
2000
N.L.
P.E.I.
N.S.
N.B.
Ont.
Man.
Sask.
Alta.
B.C.
Territories
Most of the provinces show a steady increase in the average RIW per in-patient case over
the 5-year period, except for Manitoba and Saskatchewan.
RIW per In-patient Day
The chart below shows the trend in average RIW per in-patient day. Here Ontario stands
out, with a steady increase such that by 2000–2001 the Ontario RIW per in-patient day is
clearly higher than that of the other provinces. New Brunswick also has a steady increase,
but remains lower than most other provinces in 2000–2001. Hence a perspective resulting
from the graph below is that the changes in New Brunswick reflect more comprehensive
coding practices generally as opposed to practices that focused on improving RIW. This is
discussed more fully later in the discussion section of the document.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
20
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Average RIW per In-patient Day Trend from 1996 to 2000 by Province
0.24
0.23
0.22
0.21
0.20
0.19
0.18
0.17
0.16
0.15
1996
1997
1998
1999
2000
N.L.
P.E.I.
N.S.
N.B.
Ont.
Man.
Sask.
Alta.
B.C.
Territories
Increases in Reported Diagnoses
The changes in RIW per in-patient day over the 5-year period for Ontario and
New Brunswick prompted an analysis of corresponding changes in the reporting
rate for diagnoses.
Over the 5-year period there was a 13% increase in reported diagnoses in the DAD, but
this increase was concentrated in the Type 1 diagnoses (26% increase) and the Type 2
diagnoses (46% increase). Other diagnoses (mainly Type 3 secondary) dropped by 2%.
Change in Reported Diagnoses for In-patient Cases
by Diagnosis Type from 1996 to 2000 by Province
Province
Type 1 Type 2
Other
Total
N.L.
-7%
0%
-38%
-28%
P.E.I.
33%
59%
-17%
3%
N.S.
0%
-7%
-8%
-6%
N.B.
84%
114%
-8%
34%
Ont.
42%
92%
-5%
22%
Man.
-9%
-20%
-5%
-7%
Sask.
11%
-23%
3%
5%
Alta.
9%
11%
15%
13%
B.C.
2%
3%
-2%
0%
Territories
68%
13%
-17%
18%
Total
26%
46%
-2%
13%
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
21
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Variation by Province
The change in reported diagnoses was not consistent across the provinces. Three
provinces had a decrease in reported diagnoses (Newfoundland and Labrador,
Nova Scotia, and Manitoba) while two provinces had increases in excess of 20%
(New Brunswick and Ontario).
Type 1 Diagnoses
The differences are even greater when examined by diagnosis type. New Brunswick had
an 84% increase in Type 1 (pre-existing comorbidity) diagnoses. The corresponding
increase for the territories was 68% and 42% for Ontario. Newfoundland and Labrador
and Manitoba had 7% and 9% respectively decreases in Type 1 diagnoses. The large
increases in Type 1 diagnoses must reflect changes in coding practices since they are too
large to be due to changes in the underlying health status of the acute care in-patients.
Type 2 Diagnoses
The increases in reported Type 2 (post-admit comorbidity) diagnoses vary from reductions
in Nova Scotia, Manitoba, and Saskatchewan, to a 114% increase in New Brunswick and a
92% increase in Ontario. Thus, this variation remains a concern for CIHI.
Alberta is the only province where the change in reporting of diagnoses does not appear
to be related to diagnosis type. In all other provinces and territories a greater proportion of
the reported diagnoses were Type 1 or 2 in 2000–2001. In Alberta, the number of other
(secondary) diagnoses grew at a faster rate than the rate for Type 1 or 2 diagnoses. This
may be evidence of the residual impact in Alberta of prior use of the RDRG grouper,
(rather than the CMG grouper) which did not rely on diagnosis typing to assign
complications and comorbidities.
The change in reported Type 2 diagnoses per in-patient case over the 5-year period is
shown in the graph below.
Change in Reported Type 2 Diagnoses per In-patient Case
from 1996 to 2000 by Province
0.25
0.20
0.15
0.10
0.05
1996
1997
1998
1999
2000
N.L.
P.E.I.
N.S.
N.B.
Ont.
Man.
Sask.
Alta.
B.C.
Territories
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
22
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The graph shows that as a result of the increased reporting of Type 2 diagnoses in
Ontario, the Ontario Type 2 diagnosis rate was substantially higher than the rate in the
other provinces. While New Brunswick had the highest rate of increase, because it started
with a low Type 2 diagnosis reporting rate, it remained lower than both Ontario and
Alberta in 2000–2001.
Grade List Diagnoses
There are also differences between the provinces in the volumes of grade list diagnoses
reported per case since the introduction of the complexity methodology. There has been a
55% increase in the reported grade list diagnoses per in-patient case in the DAD (versus a
17% increase for non-grade list diagnoses) from 1996–1997 to 2000–2001. Every province
had an increase in grade list diagnoses. New Brunswick and the territories had a 108%
increase in grade list diagnoses and Ontario had a 95% increase.
Change in Grade List Diagnoses Reported per In-patient Case
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1996
1997
N.L.
Man.
P.E.I.
Sask.
1998
N.S.
Alta.
1999
N.B.
B.C.
2000
Ont.
Territories
Once again, the increase in reported grade list diagnoses leaves Ontario with a much
higher rate than everywhere else in 2000–2001, and New Brunswick with a rate similar to
that of Alberta. In spite of the over 100% increase in grade list diagnoses in the territories,
the 2000–2001 rate is still less than the rate in the other provinces.
The following table compares the increase in grade list diagnoses with the increase in nongrade list diagnoses. In every province except Alberta, the increase in grade list diagnoses
is greater than the increase in non-grade list diagnoses. Non-grade list diagnosis reporting
dropped in Prince Edward Island and Newfoundland and Labrador.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
23
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Focus of Health Records Professionals on Most Significant Diagnoses
These results support the anecdotal feedback from health records professionals that lack
of availability of sufficient coding staff has caused health records departments to focus on
coding and reporting those diagnoses that are the most significant and that impact RIW
assignment (i.e. Type 1 and 2 grade list diagnoses).
Change in Grade List and Non-Grade List Diagnoses per Case From 1996 to 2000
17%
Total
55%
27%
Territories
8%
B.C.
Alta.
6%
3%
Sask.
Man.
2%
14%
12%
Non-Grade List
Grade List
12%
7%
26%
Ont.
5%
N.S.
-20%
-40%
-20%
95%
41%
N.B.
-10%
108%
108%
16%
P.E.I.
28%
N.L.
7%
0%
20%
40%
60%
80%
100%
120%
Distribution of Cases by Complexity
The emphasis on reporting of grade list diagnoses has had an impact on the distribution of
cases by complexity level.
Percent Change in CIHI DAD In-patient Cases by Complexity from 1996 to 2000
56%
33%
16%
-8%
-12%
Plx 1
Plx 2
Plx 3
Plx 4
Plx 9
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
-6%
All Cases
24
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Across the DAD there was a 6% decrease from 1996–1997 to 2000–2001 in in-patient
cases and an 8% decrease in complexity level 9 cases. Complexity level 1 cases decreased
by 12%, while case volumes in the higher complexity levels increased.
For Ontario and New Brunswick, the overall DAD pattern is replicated but with l
larger decreases in complexity level 1 cases and larger increases in complexity level 2,
3, and 4 cases.
Percent Change in Ontario In-patient Cases by Complexity
from 1996 to 2000
98%
57%
25%
-18%
Plx 1
Plx 2
Plx 3
Plx 4
-9%
-7%
Plx 9
All Cases
Percent Change in New Brunswick In-patient Cases by Complexity
from 1996 to 2000
113%
66%
54%
-17%
Plx 1
Plx 2
Plx 3
Plx 4
-9%
-8%
Plx 9
All Cases
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
25
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
In contrast, Alberta had a slight increase in complexity 1 cases and small increases in
complexity 2, 3, and 4 cases.
Percent Change in Alberta In-patient Cases by Complexity from 1996 to 2000
120%
100%
80%
60%
40%
20%
0%
19%
1%
8%
2%
1%
-1%
-20%
-40%
Plx 1
Plx 2
Plx 3
Plx 4
Plx 9
All Cases
Correlation Between Level 4 and Mortality
If the 56% increase in complexity level 4 (life threatening illness) cases was purely due to
increased patient complexity, we might expect to see a corresponding increase in actual
in-hospital mortality.
Percent Change in In-hospital Deaths and Plx 4 Cases
from 1996 to 2000 by Province
Province
In-hospital
Deaths
Plx 4
Cases
N.L.
3%
-5%
P.E.I.
33%
36%
N.S.
1%
5%
N.B.
4%
113%
Ont.
-2%
98%
Man.
-10%
-4%
Sask.
11%
14%
Alta.
9%
19%
B.C.
1%
8%
58%
110%
1%
56%
Territories
Total
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
26
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
However, the 56% increase in complexity level 4 cases was accompanied by only 1%
increase in-hospital deaths. In Ontario, for the same period where there was a 98%
increase in complexity level 4 cases, there was a reduction of 2% in the actual number of
in in-hospital deaths. In Newfoundland and Labrador, while there was a 5% decrease in
complexity level 4 cases, there was a 3% increase in in-hospital deaths.
Percent Change in In-hospital Deaths and Plx 4 Cases from 1996 to 2000
by Patient Group by Province
Province
N.L.
Medicine
Surgery
Other
Deaths Plx 4 Deaths Plx 4 Deaths
5%
4%
-5%
-10%
-16%
P.E.I.
39%
63%
-3%
7%
9%
N.S.
2%
12%
-1%
-1%
7%
N.B.
5%
154%
1%
69%
-10%
Ont.
-2%
124%
-2%
71%
-18%
Man.
-10%
4%
-11%
-10%
-39%
Sask.
13%
28%
0%
0%
-29%
Alta.
10%
29%
7%
10%
14%
B.C.
0%
12%
2%
4%
-2%
51%
219%
300%
-7%
-60%
1%
76%
0%
37%
-12%
Territories
Total
Particularly for Ontario, the change in complexity level 4 cases is not correlated with the
actual change in in-hospital mortality. This suggests that the changes in the DAD data
contributed by Ontario hospitals are more due to changes in coding and reporting
practices than in changes in the acute care in-patient population in the province.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
27
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Comparisons for 2000–2001
This chapter of the report shows the resulting distribution of DAD data in fiscal
year 2000–2001, and repeats many of the comparisons previously shown for fiscal
year 1996–1997.
2000 CIHI DAD Data by Province
Province
IP Cases
IP Days
ALOS
% ALC
SDS as % RIW per
RIW per RIW per
SDS Cases
Case
Day
of Total SDS Case
N.L.
64,142
509,608
7.94
4.0%
1.42
0.179
46,034
41.8%
0.179
P.E.I.
18,459
143,681
7.78
4.3%
1.32
0.170
9,674
34.4%
0.169
N.S.
107,228
870,763
8.12
8.1%
1.59
0.195
96,948
47.5%
0.191
N.B.
109,963
756,875
6.88
1.2%
1.33
0.193
67,003
37.9%
0.177
Ont.
1,136,183
6,966,650
6.13
9.1%
1.44
0.234
1,135,556
50.0%
0.181
Man.
76,389
702,643
9.20
6.4%
1.78
0.193
64,341
45.7%
0.201
Sask.
146,497
848,390
5.79
2.6%
1.19
0.205
100,345
40.7%
0.182
Alta.
336,916
2,140,956
6.35
6.0%
1.37
0.215
NA
NA
NA
B.C.
419,088
2,792,841
6.66
15.1%
1.39
0.208
287,476
40.7%
0.208
10,647
44,776
4.21
2.4%
0.83
0.198
5,099
32.4%
0.191
2,425,512 15,777,183
6.50
8.6%
1.41
0.217 1,812,476
42.8%
0.186
Territories
Total
In 2000–2001, Ontario hospitals still provided 47% of in-patient cases, and 63% of
ambulatory procedure cases (Alberta ambulatory procedure cases have been excluded).
Average lengths of stay remain longer in the Atlantic Provinces than in the rest of
Canada (although inconsistent reporting in New Brunswick may partly explain this).
Newfoundland and Labrador and Prince Edward Island hospitals report the lowest RIW
per day and Ontario the highest RIW per day.
New Brunswick and Ontario have the highest rates of Type 1 diagnoses per in-patient
case, more than double the rates in Newfoundland and Prince Edward Island.
Ontario has the highest rate of Type 2 diagnoses per in-patient case, and Prince Edward
Island and Saskatchewan the lowest.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
28
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Alberta has the highest rate of reporting of other (secondary) diagnoses, followed by Nova
Scotia. The rates for other diagnoses are below the DAD average in New Brunswick and
Ontario, and are the lowest in Newfoundland and Prince Edward Island. Inconsistent
reporting and the propensity to record diagnosis types more so in one province verses
another is a concern for CIHI. In our Provincial visits that took place in the spring, 2003,
we identified these inconsistencies as an area for future focus as these practices contribute
to variations in the data.
2000 Diagnoses per In-patient Case by Diagnosis Type by Province
Province
T1 Dx T2 Dx per Other Dx
per Case
Case
per Case
Total Dx
per Case
N.L.
0.46
0.11
0.86
1.43
P.E.I.
0.65
0.09
0.71
1.46
N.S.
0.73
0.15
1.65
2.53
N.B.
1.35
0.16
0.91
2.42
Ont.
1.33
0.24
0.95
2.51
Man.
0.82
0.14
1.46
2.41
Sask.
0.82
0.09
1.07
1.98
Alta.
1.16
0.17
1.89
3.22
B.C.
0.88
0.12
0.94
1.93
Territories
0.85
0.07
0.62
1.53
Total
1.12
0.18
1.12
2.43
2000 Average Diagnoses per In-patient Case by Province
Total
Territories
B.C.
Alta.
Sask.
Man.
Ont.
N.B.
T1 Dx per Case
N.S.
T2 Dx per Case
P.E.I.
Other Dx per Case
N.L.
0.00
0.50
1.00
1.50
2.00
2.50
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
3.00
3.50
29
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Grade List Diagnoses
Ontario has the highest rate of grade list diagnoses per in-patient case, three times the
Newfoundland and Labrador rate. The overall ratio of non-grade list diagnoses has
decreased from 12.5 in 1996–1997 to 9.4 in 2000–2001.
2000 Ratio of Non-Grade List Diagnoses to Grade List Diagnoses*
Province
Grade List Dx Non-Grade List
Dx per Case
per Case
Ratio of NonGrade List to
Grade List
N.L.
0.10
1.32
12.9
P.E.I.
0.11
1.34
11.7
N.S.
0.16
2.37
14.7
N.B.
0.21
2.21
10.6
Ont.
0.30
2.21
7.3
Man.
0.18
2.23
12.3
Sask.
0.14
1.84
13.6
Alta.
0.21
3.01
14.3
B.C.
0.16
1.77
11.2
Territories
0.08
1.45
19.0
Total
0.23
2.20
9.4
* Please note numbers have been rounded off for presentation purposes.
In 2000–2001 the actual DAD LOS was 96% of the ELOS.
2000 Typical Case LOS Performance by Province
Province
IP Cases
IP Days
ALOS
(incl.
ALC)
ALOS
ALOS as
(excl. % ALC % of
ALC)
ELOS
N.L.
51,974
275,852
5.31
5.27
0.6%
119%
P.E.I.
14,609
75,569
5.17
5.14
0.6%
118%
N.S.
85,608
434,238
5.07
5.02
1.0%
109%
N.B.
90,830
470,613
5.18
5.17
0.3%
110%
Ont.
985,014
4,625,022
4.70
4.53
3.5%
90%
Man.
63,318
349,968
5.53
5.44
1.6%
112%
Sask.
117,074
508,402
4.34
4.32
0.6%
101%
Alta.
277,741
1,229,088
4.43
4.35
1.6%
98%
B.C.
341,246
1,542,255
4.52
4.37
3.3%
95%
9,107
30,047
3.30
3.27
0.8%
95%
2,036,521
9,541,054
4.68
4.56
2.6%
96%
Territories
Total
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
30
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The actual Ontario Typical LOS went from 100% of the ELOS in 1996–1997 to 90% of
the ELOS in 2000–2001. The total LOS (based on all cases) for Ontario was lower than
the LOS in all other provinces except Saskatchewan. When ALC days are removed, the
average LOS for all cases in Ontario is the lowest (5.57 days), just below Saskatchewan
(5.64 days) and B.C. (5.65 days). This suggests that the exceptional Ontario Typical LOS
performance of 90% of ELOS is not just a result of higher rates of reporting grade list
diagnoses (and correspondingly higher case complexity assignment and higher ELOS
estimate), but truly does reflect shorter acute hospital lengths of stay in Ontario.
The Alberta and British Columbia actual lengths of stay were also below the DAD ELOS.
The Ontario actual lengths of stay for Medicine and Surgery (the programs to which the
complexity methodology is applied) are less than 90% of the ELOS. The provinces of
Newfoundland and Labrador and Prince Edward Island (which have the lowest grade list
diagnosis reporting rates) have the poorest Typical case length of stay performance.
2000 Typical Case LOS Performance (as Percentage of Expected LOS)
by Program by Province
Province
Medicine Surgery Obs/Neo
Psych
N.L.
113%
122%
123%
131%
P.E.I.
120%
119%
132%
100%
N.S.
113%
103%
114%
103%
N.B.
111%
108%
117%
100%
Ont.
89%
87%
99%
92%
Man.
113%
107%
108%
130%
Sask.
98%
104%
115%
95%
Alta.
98%
96%
91%
112%
B.C.
93%
96%
100%
97%
Territories
92%
98%
108%
84%
Total
95%
93%
101%
99%
As a result of the changes in reported rates for Type 1 and 2 grade list diagnoses in
Ontario, Ontario has the lowest percent of medical/surgical in-patient cases in complexity
level 1, and the highest percent in complexity level 4.
Newfoundland and Labrador has the highest percent of medical/surgical cases in
complexity level 1 and the lowest percent in complexity level 4.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
31
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
2000 Distribution of In-patient Cases by Complexity Level
Province
N.L.
% of
Total in
Plx 9
24.0%
Distribution of Cases by Plx
(excl. Plx 9)
Plx 1
Plx 2 Plx 3 Plx 4
87.4%
7.5%
3.1% 2.0%
P.E.I.
24.7%
85.1%
9.4%
3.7% 1.8%
N.S.
24.0%
82.4%
9.7%
4.6% 3.3%
N.B.
21.2%
78.4%
12.5%
5.6% 3.4%
Ont.
30.8%
70.6%
14.2%
8.4% 6.8%
Man.
35.6%
77.3%
11.7%
6.0% 5.0%
Sask.
23.6%
84.2%
9.2%
4.1% 2.4%
Alta.
31.2%
78.6%
11.0%
5.8% 4.6%
B.C.
28.8%
81.8%
10.2%
4.8% 3.2%
Territories
40.5%
86.7%
9.3%
2.9% 1.0%
Total
29.3%
76.3% 12.1%
6.5% 5.0%
In 1996–1997 only the Alberta percent of cases in complexity level 4 was much higher
than the actual percent of in-hospital deaths. In 2000–2001, Ontario also appears to
be an outlier.
2000 Comparison of Percent of Medical/Surgical Cases in Plx 4 Versus Actual
In-hospital Mortality for Medical/Surgical Cases
Percent Complexity Level 4
8%
7%
Ont.
6%
5%
Man.
Alta.
4%
N.B.
3%
Sask.
2%
N.S.
B.C.
P.E.I.
N.L.
Territories
1%
0%
0%
1%
2%
3%
4%
Percent In-hospital Deaths
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
5%
6%
32
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
2000 Data Conclusions
There have been large increases in the volume of grade list diagnoses reported by hospitals
in Ontario and New Brunswick since 1996–1997. However, because the initial reporting
rates in New Brunswick were relatively low, the 2000–2001 New Brunswick data remains
comparable with data from other provinces represented in the DAD. This is an important
point in that, while the nature of this analysis focused on increases over a specified time
period, in the case of New Brunswick, this increase was magnified over that period
highlighting the province in this analysis. While variation remains a concern for CIHI and
must be addressed to maintain comparability, New Brunswick’s increases for reporting
have simply placed it at similar levels to other provinces.
Ontario Has Become Outlier
The same cannot be said for Ontario. The Ontario increases have caused Ontario to be
an outlier, in terms of volumes of grade list diagnoses and the impact on case distribution
by complexity, weighted cases, and LOS performance. The next section of this report
examines the Ontario DAD data. Understanding the changes in the Ontario DAD data,
and the factors that have caused the changes to occur, can help CIHI develop improved
methodologies to ensure comparability of DAD data between provinces and between
individual hospitals.
In spite of the changes in the DAD data, the complexity methodology still effectively
differentiates in-patient cases by length of stay when applied across the database.
However, the average LOS for complexity level 1 cases has increased by 0.1 days,
decreased by 0.6 days for complexity level 2 cases, decreased by 1.0 days for complexity
level 3 cases, and decreased by 2.9 days for complexity level 4 cases.
2000 Average Length of Stay for Typical Patients in the CIHI DAD
by Assigned Case Complexity Level
19.8
9.9
7.6
4.0
3.7
1
2
3
4
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
9
33
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
While the different complexity levels still exhibit different length of stay patterns, the
differences have been slightly compressed.
Similar patterns are seen when the 2000–2001 in-hospital mortality by complexity level is
examined. The actual in-hospital mortality for complexity level 4 cases has decreased
from 31.3% in 1996–1997 to 26.2% in 2000–2001.
2000 Average Percent In-hospital Mortality in the CIHI DAD by Assigned Case
Complexity Level
26.2%
12.8%
6.7%
2.0%
1
0.3%
2
3
4
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
9
34
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Understanding Changes in Ontario
Focus on Ontario
Reasons for Focus on Ontario Data
Over the course of this project, much of the analysis has focused on the Ontario data in
the DAD. Reasons for this Ontario focus include:
•
The initial comparisons of changes in the DAD data showed the greatest changes in
New Brunswick and Ontario. However, even with the changes, New Brunswick data
could not be considered to be an outlier, while on many measures, Ontario had
become an outlier.
•
Ontario contributes almost half of the records to the DAD. Any better understanding
of changes in the Ontario data will help assess the validity and comparability of the
entire DAD.
•
The results of the 2-year CIHI re-abstraction study showed higher than average rates
of diagnosis discrepancies for some of the Ontario hospitals included in the sample.
•
In parallel with this project, the Ontario Ministry of Health and Long-Term Care and
the Ontario Joint Policy and Planning Committee, established a working group to
assess the impact of variation in DAD data coding and reporting practices on
comparability of weighted case measures for funding purposes. The Ontario Ministry
of Health and Long-Term Care has previously expressed their commitment to
increased use of CIHI data to support hospital funding allocation.
Ontario 6-Year Data Set
To support the detailed analysis of the Ontario DAD data we analyzed in-patient acute
care data for individual Ontario hospitals for fiscal years 1996–1997 through 2001–2002
(6 years of data, one more than for the analyses shown in the previous chapters of this
report). All of the data were assigned to a hospital organization on the basis of the site
management and ownership in fiscal year 2001–2002.
The data were also available separated by major program (Medicine, Surgery, Psychiatry,
Maternal, and Neonates).
The following table shows the change, over the 6-year period, for the Ontario DAD data.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
35
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Changes in Ontario DAD Activity Measures from 1996–1997 to
2001–2002 (6 years)
Complexity Levels
Assigned
Complexity Not Applicable
Medical
Surgery
Psych
IP Cases
-6.7%
-6.3%
-3.1%
-8.0%
-4.9%
-6.5%
Average RIW per Case
14.2%
20.6%
0.6%
4.7%
3.7%
14.3%
RIW per Day
13.6%
19.2%
5.7%
4.3%
3.8%
14.1%
Plx 1 Cases
-24.2%
-22.0%
-23.4%
Plx 2 Cases
29.6%
50.9%
35.8%
Plx 3 Cases
84.2%
76.9%
81.7%
Plx 4 Cases
177.5%
102.9%
140.9%
-1.6%
-3.7%
Activity Measure
Deaths
Plx 9 Cases
Total
Maternal Neonate
-8.1%
-50.0%
-4.0%
-2.3%
-3.1%
-8.0%
-4.9%
-6.2%
Total Dx per Case
40.8%
52.6%
29.0%
30.5%
36.5%
41.8%
Type 1 Dx per Case
78.2%
67.9%
54.4%
50.3%
46.4%
69.1%
Type 2 Dx per Case
163.7%
155.7%
89.6%
49.2%
110.6%
150.4%
Type 3 Dx per Case
-10.5%
-1.5%
-18.9%
8.1%
-22.6%
-6.5%
Grade List Dx per Case
66.0%
93.0%
38.5%
47.6%
43.0%
72.6%
Non-Grade List Dx/Case
25.8%
28.2%
26.3%
30.1%
35.7%
27.4%
Greatest Changes Relate to Complexity Methodology
The 6-year comparisons of Ontario DAD data trends show that the changes in Ontario
data have been greatest for the programs (Medicine and Surgery) where the complexity
methodology is applied, and greatest for comorbidity and grade list diagnoses.
Ontario In-patient Complexity has Likely Increased
It is likely that the actual complexity of acute care in-patients in Ontario hospitals has
increased since 1995–1996. Ontario hospitals had funding reductions in the mid-90’s,
followed by restructuring by the Health Services Restructuring Commission, which led to
decreases in available beds. At the same time increases in ambulatory procedures have
removed some of the least complex surgical patients from in-patient care, leaving higher
average complexity of the patients remaining as in-patients.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
36
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The following graph shows the changes in Ontario acute care activity from 1995–1996
to 2000–2001.
Percent Change in Ontario Acute Care Activity from 1995–1996 to 2000–2001
25%
Day Surg
20%
Avg. LOS
15%
Discharges
Disch per 1,000
10%
5%
0%
-5%
-10%
-15%
-20%
1995–1996
1996–1997
1997–1998
1998–1999
1999–2000
2000–2001
The decrease in acute care discharges and discharges per 1,000 population and the
increase in day surgery cases, and the recently increasing length of stay, all support
the hypothesis that Ontario acute care hospital beds are increasingly used by more
complex patients.
Actual Increased Patient Complexity Unlikely to Account Completely
for Data Changes
However, the magnitude of the changes in the Ontario data (180% increase in Medicine
patients with life threatening illness, 156% increase in post-admit comorbidities for
Surgical patients, 23% reduction in in-patients with no complexity) are far too great to be
explained purely by shifts from in-patient to ambulatory care.
Shift to Ambulatory Care Occurred in Other Provinces Too
The following table shows that the shift from in-patient care to ambulatory procedures
was not confined to Ontario. On average, the other provinces included in the DAD data
decreased their in-patient case volume and increased their ambulatory procedure volume
at almost the same rates as in Ontario.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
37
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
5-Year Change in In-patient and Ambulatory Procedure (SDS) Case Volumes for
Ontario and Rest of Canada15
Rest of Canada
(excl. Alberta)
Ontario
Province
Fiscal Year
Inpatient
1996–1997
1997–1998
1998–1999
1999–2000
2000–2001
% Change
1996–1997
1997–1998
1998–1999
1999–2000
2000–2001
% Change
1,220,944
1,176,853
1,161,740
1,150,646
1,134,728
-7.1%
1,032,810
1,021,081
1,021,088
997,315
954,538
-7.6%
SDS
960,622
1,028,880
1,043,470
1,092,415
1,135,556
18.2%
577,114
607,157
640,902
665,938
676,920
17.3%
Total
2,181,566
2,205,733
2,205,210
2,243,061
2,270,284
4.1%
1,609,924
1,628,238
1,661,990
1,663,253
1,631,458
1.3%
% SDS
44.0%
46.6%
47.3%
48.7%
50.0%
13.6%
35.8%
37.3%
38.6%
40.0%
41.5%
15.7%
We were unable to identify other hospital system or population health changes that would
explain the changes in the Ontario (and New Brunswick) DAD data, but that would have
much less impact in the other provinces. This left the hypothesis that the Ontario DAD
data changes were more likely due to changes in coding and reporting practices and not
changes in the acute care patient population.
Analysis of Individual Ontario Hospital Data
To test this hypothesis, a series of analyses were conducted using individual hospital data.
If there was variation in the DAD data between Ontario hospitals, such that a subset were
driving the large changes seen at the aggregate level, and there were no apparent sudden
changes in program mix or population served for these hospitals, then the change could be
attributed to upcoding.
Focus on Data Impacting Complexity
The analyses focused on the data elements with the greatest impact on complexity
assignment (and the associated weighted cases): diagnosis typing (comorbid diagnoses)
and grade list diagnoses.
15
Alberta data were excluded because of the incomplete reporting of ambulatory procedure data in the CIHI DAD.
Quebec discharge data is not included in the DAD.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
38
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Ontario Diagnosis Typing
The first set analyses of the 6-year Ontario hospital dataset examined diagnosis typing and
the average number of diagnoses reported per case. The Ontario hospital data were
separated by hospital type:
Hospital Peer Groups
• Small—less than 2,000 in-patient RIW weighted cases in 2001–2002
• Medium—more than 2,000 in-patient weighted cases but less than 15,000
• Large—more than 15,000 in-patient weighted cases
• Teaching—Member of the Ontario Council of Teaching Hospitals
Reporting of Type 1 and 2 Diagnoses
The following table shows that the increases in rates of reporting of Type 1 and 2
diagnoses were greatest for the teaching and large community hospitals.
Change from 1996–1997 to 2000–2001 in Ontario Diagnoses per In-patient Case by
Diagnosis Type and Hospital Type
Diagnoses per
Case
Hospital
Type
Small
Medium
Large
Teaching
Type 1
25%
44%
72%
91%
Type 2
65%
98%
130%
195%
Type 1 & 2
27%
49%
78%
107%
Type 3
Total
-6%
14%
-2%
27%
-9%
40%
-5%
58%
Increase May Reflect Greater Emphasis on Measurement Using DAD Data
The large urban hospitals in Ontario have, since the late 90’s, been developing enhanced
decision support functions and attempting to move to greater emphasis on data driven
and performance based decision-making. These hospitals have also been more able to
recruit health records coders and analysts than have hospitals in smaller, more rural
communities. These may be contributing factors to their greater increase in
comprehensiveness of diagnosis reporting in their DAD submissions.
Variation within Teaching Peer Group
However, even within a peer group the change in coding and reporting of comorbid
diagnoses is not uniform. While the teaching hospital peer group had an average 195%
increase in Type 2 diagnoses per in-patient case, there were three individual Ontario
teaching hospitals with increases less than 30%, and four had increases more than 200%.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
39
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
For Type 1 diagnoses per in-patient case, one teaching hospital had a drop of 4%, while at
the other end of the spectrum one hospital had an increase of 164%.
Change from 1996–1997 to 2000–2001 in Type 1 and 2 Diagnoses per In-patient
Case by Individual Ontario Teaching Hospital
Teaching
Hospital
A
B
C
D
E
F
G
H
I
J
K
Total
Type 1 Dx Type 2 Dx
per Case
per Case
147%
373%
62%
109%
-4%
25%
164%
280%
92%
165%
10%
27%
36%
15%
83%
215%
46%
36%
106%
176%
91%
346%
91%
195%
The different rates of increases in reporting of Type 1 and 2 diagnoses in the teaching
hospitals haven’t led to more comparable data for 2001–2002 (i.e. the increases haven’t
been greatest for hospitals that started with low reporting levels). For 2001–2002, there is
a three-fold range, from lowest to highest, in the average number of Type 1 and 2
diagnoses per in-patient case for the Ontario teaching hospitals for medical and surgical
patients. The coefficient of variation (C.V.)16 for Type 1 and 2 diagnoses per case for the
teaching hospitals is 40%. The following table summarizes the range (maximum minus
minimum value) and coefficient of variation for Type 1 and 2 diagnoses per in-patient
case for 1996–1997 and for 2001–2002 Ontario data.
16
The coefficient of variation is a measure of relative variation calculated as standard deviation divided by mean, and then
multiplied times 100 to generate a percentage.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
40
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
1996 and 2001 Range and Coefficient of Variation for Type 1 and 2 Diagnoses
per Case by Ontario Peer Group
Hospital Peer
Group
1996/1997
Range
C.V.
2001/2002
Range
C.V.
Small
2.50
45%
1.47
48%
Medium
1.25
33%
1.38
39%
Large
0.79
19%
1.78
41%
Teaching
0.85
19%
1.96
40%
For all but the smallest hospitals17 there was an increase in the range and coefficient
of variation for Type 1 and 2 diagnoses per in-patient case from fiscal year 1996–1997 to
2001–2002. While in 1996–1997 the smallest ranges and lowest C.V. values were for the
large and teaching hospitals, by 2001–2002 they had the largest ranges and the C.V.
values had doubled.
The hospitals with the lowest rates are not necessarily the hospitals perceived to have
the least complex patient population, and the hospitals with the highest rates are not
necessarily the hospitals perceived to have the most complex patient population. The
following table shows the Type 1 and 2 diagnoses per in-patient case for each of the
medical and surgical programs for individual Ontario teaching hospitals in 2001–2002.
Conclusion that Variation Has Increased and Comparability Reduced
Our conclusion from the review of the comorbid diagnosis reporting in Ontario is that
while there has been increased coding and reporting of Type 1 and 2 diagnoses in Ontario
since 1996–1997, the variation in reporting rates between hospitals has increased, and the
comparability of this data between hospitals has been reduced.
Review of this and similar Ontario data by the MOHLTC and the JPPC has heightened
concern in Ontario about comparability of the DAD data and the reliability of
performance measures based on the complexity measures contained in the data.
17
The large range for small hospitals for 1996–1997 was because one small hospital reported virtually no
Type 1 or 2 diagnoses.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
41
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Ontario Grade List Diagnosis Analyses
Increased Grade List Reporting will Increase Complexity
If the changes in the Ontario DAD data were at least partially a result of attempts by
hospitals to maximize their RIW weighted cases for funding purposes, we would expect to
see large increases in coding and reporting of grade list diagnoses, since only grade list
diagnoses can impact complexity assignment. Previous tables have shown that the overall
increase in grade list diagnoses per in-patient case in Ontario hospitals was 73% from
1996–1997 to 2001–2002. However, this increase was not uniform across the peer groups;
it was much greater in teaching and large community hospitals.
Percent Change in Reported Grade List Diagnoses per In-patient Cases
from 1996–1997 to 2000–2001 by Ontario Hospital Peer Group
97%
71%
47%
28%
Small
Medium
Large
Teach
Within the peer groups, the range and C.V. for grade list diagnoses per in-patient case
also increased from 1996–1997 to 2001–2002.
1996 and 2001 Range and Coefficient of Variation for Grade List Diagnoses
per Case by Ontario Peer Group
Hospital Peer
Group
1996–1997
Range
C.V.
2001–2002
Range
C.V.
Small
1.50
49%
1.47
50%
Medium
0.89
34%
1.38
33%
Large
0.75
26%
1.78
30%
Teaching
0.74
21%
1.96
39%
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
42
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The change in grade list diagnosis reporting rates for individual Ontario teaching hospitals
is shown in the following chart. Four hospitals had increases in grade list diagnoses of 25%
or less, while four hospitals had increases of more than 100% (with one more than 200%).
Percent Change in Reported Grade List Dx per In-patient Case for Individual Ontario
Teaching Hospitals from 1996 to 2001
205%
147%
136%
108%
98%
69%
69%
25%
A
D
J
E
B
K
H
I
19%
G
15%
F
0%
C
As with the reporting of comorbidities, the varying rates of increase in reporting of grade
list diagnoses have not resulted in greater consistency of rates across teaching hospitals.
The range in rates is greater than would be expected for a set of hospitals with a focus on
tertiary and quaternary care.
2001–2002 Grade List Diagnoses per Medical/Surgical In-patient Case for Individual
Ontario Teaching Hospitals
2.76
2.64
2.25
2.05
1.86
1.58
1.43
1.20
1.14
1.03
0.81
A
D
J
K
E
H
F
I
C
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
G
B
43
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
While the overall increase in grade list diagnoses per in-patient case was 73%, the
increase was much greater for some individual grade list diagnoses. The following table
shows the grade list diagnoses with the greatest percent increase in volumes reported as
Type 1 and 2 from 1996–1997 to 2001–2002 in Ontario.
Type 1 and 2 Grade List Diagnoses with Greatest Percent Increase in Reported
Ontario Volume from 1996 to 200118
Type 1 & 2 Grade List Diagnoses
ICD-9 Diagnosis Code and Description
7197
2888
2867
2738
7876
2768
2869
2767
2761
7823
2851
7990
5180
2760
2754
2766
2752
7883
Difficulty In Walking
Other Diseases White Blood Cells
Acquired Coagulation Fact Defic
Oth Disorder Plasma Protein Met
Incontinence of Feces
Hypopotassemia
Oth/Unspec Coagulation Defects
Hyperpotassemia
Hyposmolality/Hyponatremia
Edema
Acute Posthemorrhagic Anemia
Asphyxia
Pulmonary Collapse
Hyperosmolality/Hypernatremia
Disorders Of Calcium Metabolism
Fluid Overload
Disorders Magnesium Metabolism
Incontinence Of Urine
1996
135
81
165
493
217
2 774
1 050
1 300
3 038
1 251
3 919
794
2 743
570
1 738
595
368
1 164
2001
2 468
1 338
2 051
6 031
2 046
21 770
7 293
8 527
19 193
7 903
23 998
4 211
14 444
2 977
8 272
2 816
1 588
4 824
Case
%
Increase Increase
2 333
1728%
1 257
1552%
1 886
1143%
5 538
1123%
1 829
843%
18 996
685%
6 243
595%
7 227
556%
16 155
532%
6 652
532%
20 079
512%
3 417
430%
11 701
427%
2 407
422%
6 534
376%
2 221
373%
1 220
332%
3 660
314%
Difficulty in Walking
Much of the 1,728% increase in reporting of the “Difficulty in Walking” grade list
diagnosis was from three community general hospitals where the diagnosis was routinely
used for many joint replacement and stroke patients. The number of cases in these three
Ontario hospitals with this diagnosis exceeded the total number of cases with the
diagnosis reported in all of the non-Ontario DAD hospitals combined.
18
Only those diagnoses with at least 1,000 cases reported as Type 1 and 2 in 2001–2002 are shown in the table.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
44
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Grade List Diagnosis Change by ICD Chapter
The following table shows the change in the number of reported Type 1 and 2 grade list
diagnoses by International Classification of Disease chapter.
Change in Reported Ontario Type 1 and 2 Grade List Diagnoses by ICD Chapter
1996
2001
139,379
246,971
107,592
%
Increase
77%
Signs and Symptoms
48,890
119,190
70,300
144%
Endocrine
23,716
96,859
73,143
308%
Blood
25,783
86,459
60,676
235%
Respiratory
40,892
81,050
40,158
98%
Genitourinary
43,004
74,347
31,343
73%
Injury and Poisoning
44,826
67,649
22,823
51%
Digestive
42,075
61,976
19,901
47%
Mental
17,683
33,449
15,766
89%
9,498
16,724
7,226
76%
Infectious
11,575
16,337
4,762
41%
Nervous
8,620
16,040
7,420
86%
Neoplasms
9,114
12,636
3,522
39%
Musculoskeletal
3,266
7,373
4,107
126%
Congenital
1,641
2,253
612
37%
Perinatal Conditions
1,932
1,994
62
3%
359
1,423
1,064
296%
472,253
942,730
470,477
100%
ICD Chapter
Circulatory
Skin
V-Code
Total
Increase
Greatest Increases for Endocrine and Blood Diagnoses
In percentage terms, the greatest increases in reported grade list diagnoses were for
Endocrine (308%) and Blood (235%). Many of the individual diagnoses previously
identified as having the greatest percent increase fall in the Endocrine and Blood
chapters of the ICD.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
45
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Requirement for “Significant Influence”
Most Endocrine and Blood diagnoses require laboratory test results to confirm their
presence. However, an abnormal test result is not sufficient by itself to justify coding these
diagnoses as a Type 1 or 2 diagnosis. To be a Type 1 or 2 diagnosis the diagnosis must
have “a significant influence on the patient’s length of stay or significantly influences the
management/treatment of the patient while in hospital”.19
Some Hospitals Identify Comorbidities Based on Lab Tests Only
During the CIHI re-abstraction project, some Ontario hospitals participating in the study
were found to frequently have Type 1 and 2 diagnoses recorded, apparently based only on
laboratory test results. There was no documentation on the medical record that the
presence of the diagnosis had any impact on either the length of stay or the treatment of
the patient. The fact that the greatest percent increase in grade list diagnoses in Ontario
occurs for diagnoses that are based on laboratory test results suggests that some hospitals
are reporting diagnoses as Type 1 or 2 without actually assessing the significance of impact
on the patient stay.
Automatic Recoding of All Type 3 Diagnoses to Type 1
Non-application of the test of significance of impact was clearly the case for one Ontario
hospital that re-submitted their DAD data at the end of the fiscal year. When the resubmitted data were compared to the originally submitted data, the only change was that
every instance (12,632 cases) of a Type 3 diagnosis (secondary) for a patient to which the
complexity methodology would apply had been changed to a Type 1 diagnosis. These
changes were clearly made automatically, without any assessment of the validity of the
diagnosis as a Type 1.
Analyses of Endocrine Diagnosis Reporting
We focused our next analyses on the Endocrine grade list diagnoses, to see whether the
increases in reporting were consistent across all types of hospitals. Just as the overall
increase in reported grade list diagnoses is much greater for the larger Ontario hospitals,
we found that the increase in Endocrine grade list diagnoses was also greater in the largest
hospitals. The table on the following page shows the results of the analysis by peer group.
There was a 453% increase in Type 1 and 2 Endocrine grade list diagnoses reported by
teaching hospitals, but only an 82% increase for the small hospitals.
Within the teaching hospitals, the increase from 1996–1997 to 2001–2002 in reported
Endocrine grade list diagnoses ranged from 2% to 2,151%.
19
In January 2003 the CIHI diagnosis typing guideline was modified to further reinforce these requirements.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
46
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Increase in Reported Type 1 and 2 Endocrine Grade List Diagnoses
by Ontario Hospital Peer Group20
Hospital Peer
Group
Type 1 and 2 Endocrine Grade List Diagnoses
1996–1997 2001–2002
Small
Change % Change
999
1,818
819
82%
Medium
4,060
9,726
5,666
140%
Large
8,785
35,845
27,060
308%
Teaching
8,192
45,300
37,108
453%
22,036
92,689
70,653
321%
Total
Change in Reported Type 1 and 2 Endocrine Grade List Diagnoses
from 1996–1997 to 2001–2002 for Individual Ontario Teaching Hospitals
Hospital
20
1996
2001
Change
% Change
A
781
17,578
16,797
2151%
B
120
451
331
276%
C
650
663
13
2%
D
893
7,188
6,295
705%
E
390
2,799
2,409
618%
F
897
1,353
456
51%
G
372
440
68
18%
H
754
1,646
892
118%
I
709
1,089
380
54%
J
1,343
6,426
5,083
378%
K
1,283
5,667
4,384
342%
Total
8,192
45,300
37,108
453%
The total Endocrine grade list diagnosis increase does not match the increase shown in the previous table because a small
number of cases from hospitals not assigned to peer groups have been excluded here.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
47
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
For the Ontario teaching hospital data for fiscal year 2001–2002, 21% of all medical/
surgical in-patient cases had an Endocrine grade list diagnosis recorded as a Type 1 or
Type 2 diagnosis. The rate for individual hospitals ranged from 4% (there were four
hospitals at 5% or lower) to 54%. The reported 54% rate for one hospital would mean (if
the data were accurately coded) that the majority of that hospital’s in-patient medical and
surgical in-patient cases had endocrine grade list diagnoses that either:
•
significantly affected the treatment received, or
•
required treatment beyond maintenance of the pre-existing condition, or
•
increased the length of stay by at least 24 hours.
Although the accuracy of this hospital’s data could only be confirmed through
re-abstraction, the extremely high rate has resulted in considerable variation.
Percent of 2001–2002 Medical/Surgical In-patient Cases with Type 1 or 2 Endocrine
Grade List Diagnosis for Individual Ontario Teaching Hospitals
54%
31%
25%
11%
A
4%
5%
B
C
5%
D
E
F
G
9%
H
19%
20%
J
K
5%
I
Ontario Analysis Conclusions
The initial analysis of the Ontario DAD data suggested that Ontario acute care hospitals
have faced larger increases in patient complexity and relative cost, and much improved
LOS performance, compared to most hospitals in the other provinces and territories who
submit data to the DAD. The result has been that on average, the Ontario DAD data
describe an acute hospital patient population that looks significantly different from the
patients admitted to hospital elsewhere in Canada.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
48
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
There are no obvious hospital or health system differences that would explain the
apparent differences in the patient populations. The differences between the Ontario
patient population and patients in other provinces are greatest for medical and surgical
patients (to whom the CIHI complexity methodology applies) and much less for mental
health and birthing patients (to whom the complexity methodology is not applicable). It
appears that the differences are much more likely due to differences in data coding and
reporting practices (particularly as they relate to the complexity methodology) than to
true difference in the patients admitted to acute care.
The review of the Ontario DAD data has shown that the large changes in rates of
reporting for comorbidities and grade list diagnoses, and the resulting higher patient
complexity, are not uniform across peer groups. Larger hospitals have reported greater
changes in their DAD data.
Within peer groups, the DAD data changes have not been uniform. There is wide
variation between apparently similar hospitals in rates of reporting of comorbidities.
For some Ontario hospitals, the recent change in data has been very substantial
resulting from coding practice rather than actual changes in patient characteristics
and/or care requirements.
This coding variation contributed to by a small number of Ontario hospitals may have
compromised the comparability of the DAD data at the individual hospital level. This was
the reason behind the recent decision of the Ontario MOHLTC and JPPC not to use
weighted cases based on CIHI complexity levels for hospital funding purposes.
While potentially compromised at the individual hospital level, the Ontario DAD data,
even with the coding variation, still demonstrate the effectiveness of the complexity
methodology in differentiating groups of patients. The following table shows the
2001–2002 average length of stay and percent in-hospital mortality for Ontario medical/
surgical in-patient cases by complexity level. Both the average length of stay and the
actual in-hospital mortality are still progressively greater for the higher complexity levels.
2001 Ontario Medical/Surgical Average LOS and
In-hospital Mortality by Complexity Level
Complexity
Average
LOS
Percentage
In-hospital
Mortality
1
4.2
1.7%
2
8.1
5.4%
3
10.7
9.7%
4
21.4
22.8%
All
6.8
4.8%
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
49
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The current concern with lack of comparability of DAD data between hospitals, because
of unwarranted variation in patient complexity levels, is not necessarily only because of
flaws in the underlying complexity methodology, but also because of difficulty in
monitoring adherence to, and enforcing compliance with, CIHI coding and reporting
standards. In parallel to this analysis, CIHI perceived a need to clarify its Diagnosis Typing
Coding Standards. In January 2003, a clarification of the diagnosis typing standards was
circulated to Canadian hospitals. In addition, the continued relevance and utility of
diagnosis typing, as a practice will be examined in the re-development scheduled for
CMG, RIW and Complexity Overlay resulting from the national adoption of ICD-10-CA
and CCI. In the meantime, hospitals and their coders have been advised to continue this
practice using the clarified standard. This will allow data to be collected and used to
validate whether diagnosis typing ought to continue as a practice in the re-developed
CMG and RIW products.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
50
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Sensitivity to Complexity Methodology
Changes
CIHI Developed Revised Grade List Grouper as Alternative to Ontario
Abandonment of Plx
Concurrent with this project, the Ontario JPPC was considering whether to continue to
use complexity-based RIW weighted cases as a hospital activity measure for acute care
funding in Ontario, or to use weighted cases developed by the MOHLTC that did not
depend on complexity assignment. The dataset prepared to support this project was also
used by CIHI to assess the impact of revising the complexity diagnosis grade list. These
revisions were performed as a means to retroactively minimize the impact of the significant
increases in the reporting of selected grade list diagnoses affecting complexity assignment.
The hope was that a revised CMG grouper, using a complexity methodology based on a
revised grade list, could be used by CIHI to generate RIW weighted case measurements
that could in turn be used by the JPPC and MOHLTC for funding purposes. This would
allow a modified complexity methodology to be maintained instead of completely
abandoning the complexity methodology for funding purposes in Ontario.
Revised Grade List Grouper as Diagnostic Tool
While CIHI’s revised grade list CMG grouper was not ultimately accepted for use for
funding in Ontario, we can use it as a diagnostic tool to assess the degree of reliance of
hospitals in Ontario and elsewhere on grade list diagnoses that are most subject to coding
variations. The larger the reduction in RIW weighted cases for a hospital when the
revised grade list grouper was used, the more that hospital’s weighted cases had previously
been driven by coding of these questionable diagnoses.
Grade List Revisions
CIHI fully acknowledges that this methodology described here to revise the grade list was
crude. It was the best available method given the time constraints faced in CIHI’s
discussions with Ontario. However, conceptually the methodology has merit and is being
examined more fully as an alternative to retain Complexity Overlay for use by other
Canadian provinces. We emphasize strongly, that clinical review of grade list changes and
statistical verification of the significance of comorbidities on length of stay will be
considered as part of a more comprehensive approach used to retain Complexity Overlay.
This method employing clinical review and statistical tests of significance will be
examined as CIHI seeks to retaining Complexity Overlay for use until it can be redeveloped using ICD-10-CA and CCI activity data.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
51
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The 440 grade list diagnoses were originally selected because they helped to differentiate
between non-complex and complex patients in the same CMG. The grade list diagnoses
were selected based on clinical review and statistical analyses conducted in the early
1990’s. The intent of the grade list revision was to review the grade list diagnoses to assess
whether they are still reliable to identify complex patients, and to remove the diagnoses
no longer considered being reliable for that purpose.
For example, ICD code 276.8 Hypopotassemia is a grade list diagnosis that increased in
reported volume by more than 600% in Ontario hospitals from 1996–1997 to 2001–2002.
In some Ontario teaching hospitals more than 10% of patients have this diagnosis as a
comorbidity, while in others it is less than 1%.
Percent of Medical/Surgical Patients with Hypopotassemia Coded as Type 1 or 2
Dx in Ontario Teaching Hospitals in 2001–2002
12%
11%
11%
9%
5%
2%
2%
2%
1%
1%
0%
0%
The review of the grade list diagnoses was intended to assess whether a diagnosis like
Hypopotassemia should remain on the complexity grade list.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
52
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Grade List Review Process
The approach for the review is shown in the flow chart below.
Approach for Ontario Review of Grade List Diagnoses
Review Reported
Grade List
Diagnoses (Dx)
Clinical Review to
Confirm/Modify
Dx to Remove
Remove Dx
> 200% Increase
Since 1996
Remove Dx with
Variation within
Peer Groups in
Surgical Rate
Remove Dx with
Variation within
Peer Groups in
Medical Rate
Modify CMG Grouper
to Exclude Identified
Diagnoses
(Assessment of
variation based on
inter-quartile range
over median, above
75th percentile)
Re-Group
Ontario
Data
Three initial criteria were used to flag grade list diagnoses to be considered for
removal from the grade list:
1. Any diagnosis with an increase in reported volume (as a comorbidity) greater
than 200% between 1996–1997 and 2001–2002. The diagnoses with the greatest
percent increase tend to be the same diagnoses identified as often upcoded in the
re-abstraction studies.
2. Any diagnosis with large variation in rate per surgical case across Ontario hospitals.
Large variation was considered to be a ratio of the inter-quartile range to the median
rate per surgical case above the 75th percentile ratio of all of the diagnoses.
3. Any diagnosis with large variation in rate per medical case across Ontario hospitals.
Large variation was considered to be a ratio of the inter-quartile range to the median
rate per medical case above the 75th percentile ratio of all of the diagnoses.
Review of Grade List by CIHI Classification Specialists
The grade list, with the flagged diagnoses, was then provided to CIHI classification
specialists who confirmed or modified the flagged diagnoses based on their assessment of:
•
appropriateness of reporting of the diagnosis in re-abstracted data;
•
prior identification of the diagnosis as problematic by coders;
•
certainty of clinical criteria for the diagnosis; and
•
potential program differences across hospitals that would justify large
inter-hospital variation (e.g. renal failure diagnoses concentrated in hospitals
with large dialysis programs).
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
53
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The classification specialists confirmed some of the grade list diagnoses marked for
removal, rejected some of the flagged diagnoses (thereby retaining them on the
grade list), and added flags for removal to some diagnoses not identified using the
three statistical criteria.
The revised grade list was then used to re-group the data, generating a new dataset with
fewer higher complexity cases and lower RIW weighted case volumes.
Impact on 2001–2002 Ontario Data
The approach to revising the grade list was applied to the 942,730 counts of the
440 grade list diagnoses in the Ontario 2001–2002 Ontario DAD in-patient database.
Of the 440 grade list diagnoses:
•
44 diagnoses were flagged because of an increase in volume greater than 200%
•
57 diagnoses were flagged because of large surgical variation
•
45 diagnoses were flagged because of large medical variation
•
35 of the flagged diagnoses were retained on the grade list due to the clinical advice
from the classification specialists
•
85 additional diagnoses were removed from the grade list solely due to clinical advice
from the classification specialists
166 Diagnoses Removed from Grade List
The net impact was to eliminate 166 of the grade list diagnoses and to reduce the
2001–2002 Ontario grade list diagnosis count to 354,638 (38% of the original volume of
grade list diagnoses). A CMG grouper was then modified to use only the remaining grade
list diagnoses to identify complex cases, and a revised total RIW weighted case number
calculated for each Ontario hospital.
The impact on RIW weighted cases of using the revised grade list grouper is shown in the
following table.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
54
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Impact of Revised Grade List Grouper on Ontario Hospital 2000–2001
RIW Weighted Cases
Impact on RIW Weighted Cases
Medical/
Surgical
RIW
Total
RIW
Average Change
-8.9%
-7.2%
Median Change
-6.2%
-4.9%
Maximum Change
-19.1%
-15.5%
Minimum Change
0.8%
0.6%
# Hospitals > 10% Change
26
15
# Hospitals < 2% Change
14
18
Teaching Median Change
7.4%
-6.1%
Large Median Change
7.9%
-5.8%
Medium Median Change
5.7%
-4.7%
Small Median Change
5.2%
-4.6%
Average Reduction of 8.9% of Medical/Surgical Weighted Cases
The average reduction in medical and surgical weighted cases was 8.9%, but the range
was from a reduction of 19.1% in medical and surgical weighted cases to an increase of
0.8%21 in medical and surgical weighted cases. Because the complexity methodology is
only applicable to medical and surgical activity the impact on the total hospital weighted
cases (including mental health and birthing) is less.
Greatest Impact on Large and Teaching Hospitals
The teaching and large Ontario hospitals had the largest median reductions in RIW
weighted cases from application of the revised grade list grouper.
21
An increase in weighted cases can occur using the revised grade list where a case that was a Typical patient with the
original grade list becomes an Outlier with the revised grade list and is then assigned a higher RIW.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
55
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Use of Revised Grade List Grouper to Assess All DAD Data
The next step was to re-group the remainder of the DAD data (the non-Ontario portion)
and compare the impact on non-Ontario hospital RIW weighted cases with the results for
the Ontario hospitals.
Change in RIW Weighted Cases as Surrogate Upcoding Measure
The assumption was that the change in medical/surgical RIW weighted cases from
application of the revised grade list grouper was a valid measure of the degree of reliance
of a hospital on questionable comorbidities. If some non-Ontario hospitals were also
upcoding to the same extent as Ontario hospitals, we would expect to see large impacts on
their medical and surgical RIW weighted cases by using the revised grade list grouper. The
magnitude of the change in RIW weighted cases could considered a surrogate upcoding
index.
The following graph shows the 2000–2001 average grade list diagnoses per 100 medical
and surgical cases by province.
2000–2001 Grade List Diagnoses per 100 Medical/Surgical In-patient Cases
Y.T.
Nun.
N.W.T.
B.C.
Alta.
Sask.
Man.
Ont.
N.B.
N.S.
P.E.I.
N.L.
0
10
20
30
40
50
60
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
70
80
56
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Impact of Revisions to Grade List on Diagnoses per 100 Medical/Surgical
Patients by Province
Y.T.
Nun.
Retained
N.W.T.
Removed
B.C.
Alta.
Sask.
Man.
Ont.
N.B.
N.S.
P.E.I.
N.L.
0
10
20
30
40
50
60
70
80
The impact was greatest in Ontario, where more than 60% of the occurrences of grade list
diagnoses were removed, and in New Brunswick, where 58% of the diagnoses were removed.
Impact of Revisions to Grade List on Volume of Grade List Diagnoses
(percentage removed) by Province
Y.T.
Nun.
N.W.T.
B.C.
Alta.
Sask.
Man.
Ont.
N.B.
N.S.
P.E.I.
N.L.
40%
45%
50%
55%
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
60%
65%
57
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
The resulting impact on RIW weighted cases, by province is shown below.
Reduction in 2000–2001 Weighted Cases Due to Revised Grade List, by Province
Province
N.L.
P.E.I.
N.S.
N.B.
Man.
Sask.
Alta.
B.C.
N.W.T.
Nun.
Y.T.
Ont.
% Chg. In
% Chg. In
Med/Surg
Total RIW
RIW
-1.3%
-1.1%
-2.5%
-2.1%
-2.8%
-2.3%
-4.2%
-3.5%
-3.3%
-2.6%
-3.0%
-2.5%
-4.0%
-3.2%
-2.9%
-2.3%
-1.4%
-1.0%
-1.8%
-1.1%
-3.7%
-2.9%
-8.9%
-7.2%
Ontario Reduction More Than Double Any Other Province
The average impact on Ontario is more than double the impact on any other province.
The greatest impact outside Ontario is on New Brunswick and Alberta, and the least
impact outside Ontario is on Newfoundland and Labrador. The relative impact on
Ontario is clearly shown in the chart below.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
58
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Percent Reduction in Medical/Surgical Weighted Cases Due to Revised Grade List
Ont.
Y.T.
Nun.
N.W.T.
B.C.
Alta.
Sask.
Man.
N.B.
N.S.
P.E.I.
N.L.
-10%
-9%
-8%
-7%
-6%
-5%
-4%
-3%
-2%
-1%
0%
The distribution of the impact on medical and surgical RIW weighted cases for individual
hospitals in Ontario is greater than for the rest of the DAD.
Range of Reduction in Medical/Surgical Weighted Cases for Individual Hospitals,
with Application of Revised Grade List
0.0%
-1.0%
th
75 Percentile
-2.0%
Other
-3.0%
-4.0%
th
25 Percentile
75th Percentile
-5.0%
-6.0%
Ontario
-7.0%
-8.0%
-9.0%
th
-10.0%
25 Percentile
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
59
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Only one quarter of non-Ontario hospitals have a reduction in medical and surgical RIW
weighted cases of more than 3.5%. One quarter of Ontario hospitals have a reduction in
medical and surgical RIW weighted cases of more than 8.9%.
Reduction in Non-Ontario Medical/Surgical Weighted Cases
Due to Revised Grade List by CIHI Peer Group
Peer Group
0–49 beds
50–99 beds
100–199 beds
200–399 beds
400 + beds
Teaching
Paediatrics
Total
% Chg. In
% Chg. In
Med/Surg
Total RIW
RIW
-2.7%
-2.4%
-2.6%
-2.2%
-2.7%
-2.2%
-3.3%
-2.7%
-3.2%
-2.5%
-3.8%
-3.0%
-3.9%
-3.0%
-3.3%
-2.7%
The greatest impacts on RIW weighted cases for the non-Ontario hospitals are in large
and teaching hospitals. This result is similar to the Ontario result, but the magnitude of
the impact is much less.
The distribution of individual hospitals by reduction in their medical and surgical RIW
weighted cases is shown below.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
60
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Distribution of Reduction in Weighted Cases Due to Revised Grade List for
Individual Hospitals by Province
Province
< 1 % 1 to 3% 3 to 5% 5 to 7% 7 to 10%
Alta.
9
49
38
11
4
B.C.
12
49
27
3
1
3
4
7
Man.
N.B.
2
9
N.L.
19
15
N.S.
10
19
Nun.
6
1
2
1
Ont.
4
19
23
4
3
32
10
P.E.I.
18
Y.T.
Total
7
> 10%
2
1
1
1
N.W.T.
Sask.
% of
Hospitals
> 7%
Range of Reduction in Med/Surg Weighted cases
38
5
27
26
3
1
75
202
120
65
36
28
Total
113
5%
92
1%
7
0%
26
4%
34
0%
36
0%
1
0%
4
0%
137
39%
7
0%
68
4%
1
0%
526
12%
In Ontario, 39% of hospitals have a greater than 7% reduction in medical and surgical
RIW weighted cases when the revised grade list grouper is used. For all other provinces,
no more than 5% of hospital have a reduction in excess of 7%.
The difference is even more striking when the size of the hospitals is taken into account.
In Ontario, it is the larger hospitals that are impacted the most by the revised grade list
grouper (and that have been most reliant on the eliminated diagnoses). 80% of the total
RIW weighted cases in Ontario are in hospitals that have a reduction in medical and
surgical weighted cases greater than 7%.
In all other provinces and territories except Saskatchewan, 1% or less of all weighted cases
are in hospitals with a reduction in medical and surgical weighted cases greater than 7%.
In Saskatchewan, only 3% of the total provincial weighted cases are in hospitals with a
reduction greater than 7%.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
61
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Medical/Surgical Weighted Cases by Province by Range of Reduction in Weighted
Cases for Individual Hospitals
Range of Reduction in Med/Surg Weighted cases
Province
<1%
1 to 3%
3 to 5%
5 to 7%
7 to 10%
> 10%
Total
% of Wtd.
Cases in
Hospitals
> 7%
Alta.
6,658
65,656
234,724
50,111
2,267
3,137
362,552
1%
B.C.
9,285
195,742
215,671
1,483
1,978
-
424,157
0%
Man.
-
28,348
80,899
-
-
-
109,247
0%
N.B.
301
25,503
55,878
34,940
1,026
-
117,649
1%
N.L.
13,703
27,830
-
-
-
-
41,534
0%
N.S.
5,993
87,290
33,828
2,416
-
-
129,527
0%
Nun.
-
-
-
-
-
688
0%
N.W.T.
688
495
2,584
185
2,678
57,172
101,004
P.E.I.
-
12,805
5,895
-
-
Sask.
6,108
68,663
48,549
6,575
3,389
Y.T.
-
-
2,010
-
-
Ont.
Total
45,222
572,281
778,642
360,841
456,366
1,566,376
1,575,035
-
3,264
0%
2,615,012
80%
-
18,699
0%
-
133,283
3%
-
2,010
0%
3,957,623
53%
526,940
530,077
The distribution of DAD hospitals for which the revised grade list grouper would reduce
medical and surgical weighted cases by more than 7% is:
•
53 hospitals in Ontario;
•
6 hospitals in Alberta;
•
3 hospitals in Saskatchewan;
•
1 hospital in British Columbia; and
•
1 hospital in New Brunswick.
The highly impacted hospitals in Ontario tend to be large, while the highly impacted
hospitals elsewhere, tend to be small.
Coding Variation Largest in Ontario
The reliance on the diagnoses removed from the grade list to generate RIW weighted
cases is far greater in Ontario hospitals than in the DAD hospitals from elsewhere in
Canada. This result suggests that the potential upcoding identified in Ontario hospitals
is confined to that province.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
62
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Conclusions from Analysis Results
Review of Results
The comparisons of DAD data by province for the 1996–1997 fiscal year show that prior
to the introduction of the CIHI complexity methodology there was provincial variation in
the data submitted by acute care hospitals to the DAD. This variation included:
•
Differences in comprehensiveness of reporting of ALC days. Less than 1% of acute
hospital days were reported as ALC in New Brunswick, while almost 10% were ALC
in Ontario.
•
Differences in number of comorbid diagnoses reported on in-patient records. Alberta
hospitals reported more than twice as many comorbid diagnoses than hospitals in
Newfoundland and Labrador and Prince Edward Island.
•
Differences in number of grade list diagnoses reported on in-patient records. Alberta
hospitals reported more than twice as many comorbid diagnoses than hospitals in
Newfoundland and Labrador, Prince Edward Island, and New Brunswick.
These differences impacted the relative reported acute care patient complexity, length of
stay performance, and in-hospital mortality, with the hospitals in the provinces with the
lowest reporting rates appearing to be the worst performers.
There Has Always Been Variation in DAD Data
While the introduction of the complexity methodology, coupled with increased use of
RIW weighted cases based on complexity for hospital funding, is associated with increased
variation in DAD data, there was already variation in the data prior to that. Some of the
variation likely reflects true differences in acute care hospital roles and funding, but some
was also due to provincial differences in:
•
interpretation of coding and reporting guidelines;
•
availability of health records and decision support staff;
•
presence of a data driven approach to hospital system planning and funding; and
•
documentation practices that vary by facility within provinces.
Between 1996–1997 and 2000–2001 the Ontario RIW per in-patient day rose by 11%.
Only New Brunswick (6%) and Saskatchewan (4%) rose by more than 2%.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
63
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Greatest Change in Diagnosis Reporting in New Brunswick and Ontario
The average number of comorbid diagnoses per in-patient case increased by 90% in
New Brunswick and 60% in Ontario. The increase was less than 10% in most other
provinces. A similar pattern was seen for reporting of grade list diagnoses, with an increase
from 1996–1997 to 2000–2001 of approximately 100% in New Brunswick and Ontario,
and an increase less than 20% in British Columbia, Alberta, Saskatchewan, Manitoba,
Nova Scotia, and Newfoundland and Labrador.
As a result of the increased reporting of comorbid and grade list diagnoses,
New Brunswick and Ontario hospitals had a 100% increase in complexity level 4
(life threatening illness) patients. In spite of the 100% increase in the number of
patients reported as being at risk of death, New Brunswick hospitals had less than a 5%
increase in actual in-hospital deaths, and Ontario hospitals had a decrease of 2%.
Ontario is a Data Outlier
In 2000–2001 Ontario hospitals had the best LOS performance compared to the CIHI
expected LOS (which is adjusted to reflect reported patient complexity). The average
number of grade list diagnoses per in-patient case in Ontario was three times the number
in Newfoundland and Labrador, and 50% higher than the rates for any other provinces or
territories. Despite similar rates of change in reported diagnoses in New Brunswick and
Ontario, because New Brunswick started at a much lower level than Ontario, only the
Ontario data were substantially different from the other DAD data in 2000–2001.
Ontario Changes Were Not Uniform Across Hospitals
When the changes in DAD data in Ontario were examined in more detail we found that
the changes (particularly with respect to reporting comorbid diagnoses and grade list
diagnoses) were not uniform across hospital types or individual hospitals. Large and
teaching hospitals had the greatest changes, but even within these groups there was much
variation between individual facilities. There were individual hospitals with such dramatic
changes (e.g. 205% increase in reported grade list diagnoses per case, 2,151% increase in
reported endocrine comorbidities per case) that the changes in data could not be
reflective of true changes in their patient populations.
Focus Must Be Diagnosis Typing and Reporting of Comorbidities
Combining the results of the analyses of Ontario hospital-specific data with the findings of
the re-abstraction studies points leads to a focus on diagnosis typing and variation in
identification of diagnoses as comorbid conditions. Some Ontario hospitals appear to have
reported diagnoses as comorbid conditions without assessing the significance of the
presence of the diagnoses as required by CIHI diagnosis typing guidelines.
Although larger increases in capture and reporting of diagnostic data in hospitals in
provinces that were previously under-reporting would improve the comparability of DAD
data, this does not seem to be what has happened, with the possible exception of New
Brunswick. The apparent gaps in rates of reporting of diagnostic data between provinces
continue to exist, and in the case of Ontario, have been exacerbated.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
64
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
“Undercoding” Will Also Compromise Utility of DAD Data
While the emphasis of this analysis has been on increases in reporting comorbidities and
the resulting variation, “undercoding” should also be a concern. In provinces where there
has not been a history of using DAD data for planning or funding, or where there are
shortages of health records professionals, not all relevant diagnostic information may be
captured in the DAD data. This compromises the comparability and utility of
performance measurements based on DAD data as much as upcoding.
Lessons Learned
The value of the CIHI DAD is its utility to support comparisons of activity and
performance between hospitals. This value is predicated on the assumption that the DAD
data as collected, coded, and reported in a consistent manner by all of the participating
hospitals. CIHI recognizes this and these analyses, the recent re-abstraction studies, and
feedback from stakeholders have all helped CIHI identify steps that must be taken to
address variation in coding practices. Some of the actions initiated by CIHI include:
•
Established a national CMG re-development committee with an aggressive timetable
for recommendations.
•
Issued clarifications of the diagnosis typing standard to reduce variations in coding.
•
Published review of CIHI Quality Assurance Practices (November 2002).
•
Published results of the application of the Data Quality Framework to the DAD.
Some key lessons for CIHI from the findings of review of variation in the DAD are:
•
Monitoring of adherence to CIHI diagnosis typing guidelines for DAD submissions
must be enhanced.
•
CIHI will evaluate and assess the relative benefits and risks of continuing to rely on a
diagnosis typing approach that contains an element of subjective judgement as
opposed to a pre-determined black and white, data driven diagnosis typing system (as
is used to determine complications and comorbidities in the United States DRG
system). This was a consistent message delivered in CIHI’s Spring 2003 consultations
nationally. Participants in these provincial meetings expressed a desire to re-examine
the continued utility of this subjective coding practice.
•
There should be regular reporting to hospitals of their level of adherence to CIHI
DAD data reporting guidelines.
•
There should be regular review of the CMG grouping methodology to ensure that
variables previously identified as important to support differentiation of patient groups
retain their value over time as coding and reporting practices change.
•
Special attention should be paid to data elements and calculated variables used to
support funding allocations.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
65
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Monitoring and Reporting Variation
in the DAD
The analyses presented in this report provide examples of potential indicators that could
be used to monitor variation between hospital in coding and reporting of DAD data. Most
of these indicators have been used to compare year over year change in data or have been
applied to large population-based datasets where differences between the populations
would be expected to be small.
A greater challenge for CIHI is the identification and development of indicators that can be
used to assess the data from individual hospitals and that might be implemented on a “real
time” basis, so that potential upcoding or undercoding can be identified as it happens.
Variation as a Sentinel Event
Sentinel events are a type of adverse event that are indicative of underlying systemic
concerns. In health care, sentinel events are unexpected occurrences involving death or
serious physical or psychological injury, or risk thereof. Sentinel events signal the need for
immediate investigation and response. For CIHI, any significant coding variation up or
down is considered to be a sentinel event.
Sentinel Indicators
A sentinel indicator is very similar to a screening tool for use in data processing by CIHI.
It can be an indicator that automatically triggers investigation and follow-up in response.
Ideally CIHI could create indicators to identify individual abstracts submitted to CIHI
DAD that indicate that the submitting organization is not following CIHI coding
protocols. Additionally there could be indicators that would be applied to a pool of data
(not individual records) and be based on an assessment of patterns in the data.
Indicators for Individual Records and for Pools of Data
An indicator that applies to an individual record could be used in real time to flag records
that require correction or that should not be accepted into the DAD. Indicators based on
assessing patterns in a pool of data could be used with periodic data submissions if the
volume of records was sufficient, or could be retrospectively used with annual submissions.
No Black and White Indicators of Upcoding Applicable to Individual Records
CIHI experience with edit tests applied to individual patient records has been that unless
the test generates a rejection of the record (and not just a warning) the offending record
will not necessarily be corrected. For an edit test to flag a record for rejection there must
be certainty that the record is incorrect. With the current diagnosis typing guidelines
there are very few, if any, diagnoses that could never be coded as a Type 1 or 2 diagnosis.
We were unable to identify any indicators that could be applied to individual records.
Some combinations of diagnoses were considered to be very rare but not impossible.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
66
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Use of Indicators to Apply to Pools of Data
It is much more feasible for CIHI to develop and apply indicators that can be used to
assess the likelihood of upcoding in a pool of data. The indicators could be used to:
•
Provide a warning back to the submitting institution that their data appear unusual
and show evidence of upcoding/undercoding. As well, a copy of any reports/findings
sent to individual hospitals would need to be shared with provincial Ministries as well
to allow follow-up at a provincial level.
•
Support communication with Ministries of Health regarding the apparent quality and
comparability of the data submitted by individual hospitals.
•
Support development of a published comparison of data quality across hospitals. This
could then be used by researchers and hospitals to assess the utility of the DAD data
from individual hospitals for comparisons of performance and activity.
•
Identify institutions where audit or re-abstraction of their data is recommended.
•
Identify institutions whose data should be excluded from DAD for the purposes of
calculating average values (e.g. expected LOS, Typical RIW).22 However, hospitals
would be given the prior opportunity to investigate, correct as required, and report on
questionable data prior to their exclusion.
In all cases the indicators can only lead to suspicion of “upcoding/undercoding”;
confirmation of upcoding or undercoding would require on-site chart review
(i.e. re-abstraction).
In addition to facility-based indicators, extreme values for population-based indicators
monitored and reported by CIHI may provide evidence of undercoding or upcoding by the
hospitals that provide care to the population.
Potential Indicators
A potential indicator that can be used as was done in this analysis is changes in grade list
diagnosis volumes or distribution of patients by complexity level. Additional indicators
based on changes in data patterns compared with prior submissions from the same
organization include:
•
Percent change in weighted cases per in-patient day greater than 5%, with no
addition or discontinuance of a major program.
•
Percent change in volume of Type 1 or 2 diagnoses per medical or surgical
in-patient case.
•
Percent change in volume of Type 1 or 2 diagnoses per mental health in-patient case.
•
Percent change in actual LOS versus ELOS performance.
22
Exclusion of data from the DAD would be problematic for population-based analyses and indicators where
comprehensive capture of the activity for a population is required.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
67
CODING VARIATIONS IN CIHI DISCHARGE ABSTRACT DATABASE DATA
Potential indicators based on comparisons with peer organizations include:
•
Frequency of categorization of diagnosis as comorbid condition within specific CMG
(e.g. “difficulty in walking” as comorbid condition for stroke or joint replacement CMG).
•
Actual LOS less than 80% of ELOS.
•
Type 1 diagnoses per in-patient medical case above 90th percentile for peer group.
•
Type 1 diagnoses per in-patient surgical case above 90th percentile for peer group.
•
Type 2 diagnoses per in-patient medical case above 90th percentile for peer group.
•
Type 2 diagnoses per in-patient surgical case above 90th percentile for peer group.
•
Ratio of case volume in non-complex vs. complex CMG significantly different from
peers. Examples of potential CMG pairs to be assessed would include:
−
simple vs. complicated appendectomy;
−
c-section with complicating diagnosis vs. c-section with no
complicating diagnosis;
−
VBAC with complicating diagnosis vs. VBAC with no complicating diagnosis;
−
vaginal delivery with complicating diagnosis vs. vaginal delivery with no
complicating diagnosis;
−
neonates, with similar birth weight, with and without problem diagnoses; and
−
mental health CMG with Axis 3 diagnoses vs. without Axis 3.
More sophisticated tools to assess the comparability of CIHI data, referred to as “coding
indices” have been developed23 to compare the actual distribution of a hospital’s cases by
complexity with the expected distribution, taking into account variables such as:
•
hospital size and teaching status;
•
hospital case mix, procedure mix, and program mix;
•
actual in-hospital mortality; and
•
hospital location (e.g. urban vs. rural).
These coding indices were developed to be used with DAD data using the CIHI
complexity overlay but the same principles could be used to develop a CIHI coding index.
One advantage of a coding index is that it can be used to create a continuous measure
showing the degree to which a hospital has reported more (or less) resource intensity than
would be expected, rather than just a dichotomous upcoding measure.
23
Examples are the Ontario Ministry of Health Hospital Coding Index and the Thiinc Consulting RIW Weighted
Complexity Index.
C AN AD I AN I N S T I T U T E F O R H E AL T H I N F O R M AT I O N
Better Health Information for Better Health
Une meilleure information sur la santé pour une meilleure santé
www.cihi.ca
www.icis.ca
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement