Professional Documents
Culture Documents
Background and Purpose—Surveillance is necessary to understand and meet the future demands stroke will place on
health care. Administrative data are the most accessible data source for stroke surveillance in Canada. The International
Classification of Diseases, 10th revision (ICD-10) coding system has potential improvements over ICD-9 for stroke
classification. Our purpose was to compare hospital discharge abstract coding using ICD-9 and ICD-10 for stroke and
its risk factors.
Methods—We took advantage of a switch in coding systems from ICD-9 to ICD-10 to independently review stroke patient
charts. From time periods April 2000 to March 2001, 717 charts, and from April 2002 to March 2003, 249 charts were
randomly selected for review. Using a before-and-after time period design, the accuracy of hospital coding of stroke
(part I) and stroke risk factors (part II) using ICD-9 and ICD-10 was compared. We used careful definitions of stroke
and its types based on ICD-9 using the fourth and fifth digit modifier codes.
Results—Stroke coding was equally good with ICD-9 (90% [CI95 86 to 93] correct) and ICD-10 [92% (CI95 88 to 95
correct) with ICD-10. There were some differences in coding by stroke type, notably with transient ischemic attack, but
these differences were not statistically significant. Atrial fibrillation, coronary artery disease/ischemic heart disease,
diabetes mellitus, and hypertension were coded with high sensitivity (81% to 91%) and specificity (83% to 100%).
ICD-10 was as good as ICD-9 for stroke risk factor coding.
Conclusions—Passive surveillance using administrative data are a useful tool for identifying stroke and its risk factors
using both ICD-9 and ICD-10. (Stroke. 2005;36:1776-1781.)
Key Words: epidemiology 䡲 health services research 䡲 stroke 䡲 risk factors
Downloaded from http://ahajournals.org by on September 25, 2019
Received March 12, 2005; final revision received April 10, 2005; accepted May 3, 2005.
From the Calgary Stroke Program, Departments of Clinical Neurosciences (M.D.H., R.A.K.), Medicine (M.D.H.), and Community Health Sciences
(M.D.H.), Faculty of Medicine, University of Calgary, Alberta, Canada.
Both authors contributed equally to this work.
Correspondence to Michael D. Hill, Associate Professor, Calgary Stroke Program, Department of Clinical Neurosciences, University of Calgary,
Foothills Hospital, Room 1242A, 1403 29th St NW, Calgary, Alberta, T2N 2T9, Canada. E-mail michael.hill@calgaryhealthregion.ca
© 2005 American Heart Association, Inc.
Stroke is available at http://www.strokeaha.org DOI: 10.1161/01.STR.0000174293.17959.a1
1776
Kokotailo and Hill Coding of Stroke Using ICD-9 and ICD-10 1777
office, or those who did not present to medical attention. A single our determination as the gold standard, the proportion correct of
health records technologist conducted coding of hospital discharge hospital health technologist coding of AIS, SAH, ICH, and TIA
abstracts at the university hospital. Before the study period, this using ICD-9 and ICD-10 were then calculated and compared. For
person trained with the Calgary Stroke Team, learning more about practical reasons, chart reviewers were not blinded to how the charts
stroke and its clinical diagnosis and management. Further, an had been coded by the health records technologist.
ongoing dialogue exists between the health records department and Statistical comparisons were made using Fisher’s exact test, and
the Calgary Stroke Team to resolve coding issues. At the 2 all proportions are reported using exact CIs. It was not possible to
community hospitals, coding was not centralized to a single health
assess the sensitivity and specificity for the diagnosis of stroke
records technologist.
because nonstroke diagnoses were underrepresented by design.
All patients with a discharge diagnosis (ICD-9 codes 430.x, 431.x,
Instead, we report the PPV of coding by stroke type among patients
433.x1, 434.x1, 435.x, 436, and 362.3) of stroke were acquired for
the 2000/2001 fiscal year (April to March); for the 2002/2003 fiscal with stroke as well the Kappa statistic () as a measure of the
year, ICD-10 codes I60.x, I61.x, I63.x, I64.x, H34.1, and G45.x were agreement between coder and researcher. The PPVs were arbitrarily
used to identify the patients. In ICD-9, the fourth and fifth digits categorized as “poor” (⬍70%), “good” (70% to 79%), “very good”
were used to exclude or include patients. In the intervening year, (80% to 89%), or “excellent” (ⱖ90%). was considered as having
coding switched from ICD-9 to ICD-10, and this time period substantial agreement between health technologist and researcher as
between data collection epochs allowed for learning on the new indicated by 0.61 to 0.80; almost perfect agreement: 0.81 to 1.00.3
ICD-10 system. All patients (n⫽2529) with a diagnosis of stroke in
the primary diagnostic position, implying that stroke was the most
responsible diagnosis for length of stay, comprised the sampling
Part II
frame. This approach has been shown to result in high specificity and A 10% random sample of charts, which included all available
positive predictive value (PPV).3 A research assistant trained by a diagnostic positions, was drawn from the sampling frame. The
stroke neurologist in definitions of stroke and its risk factors sample size was estimated to provide, on average, a 10% 95% CI
performed the chart review. The neurologist resolved any ambigu- width without adjustment to allow for multiple comparisons. Source
ities in diagnosis. The investigators accessed the same patient chart documents, namely physician history and physical examination
documents as health records technologists. If the chart contained notes, physician progress notes, discharge summaries, nursing notes,
multiple admissions, only those documents from the admission under and laboratory, echocardiogram, and ECG reports were reviewed
review were used. systematically for atrial fibrillation, coronary artery disease/ischemic
heart disease, diabetes mellitus, history of cerebrovascular accident,
Part I hypertension, hyperlipidemia, renal failure, and tobacco use. Codes
Major stroke types subarachnoid hemorrhage (SAH), intracerebral were assigned only to those risk factors that fulfilled the diagnostic
hemorrhage (ICH), acute ischemic stroke (AIS), and transient criteria as outlined in Table 2. Criteria were investigator derived, on
ischemic attack (TIA) were defined as described in Table 1. A the basis of current medical guidelines. Using the gold standard, the
stratified random sample of charts was drawn for review, stratified sensitivity, specificity, and percent correct of hospital health tech-
by major stroke type and by year. Sampling of AIS within the ICD-9 nologist coding of these stroke risk factors using ICD-9 and ICD-10
cohort was further oversampled compared with ICD-10 to allow for were calculated and compared.
1778 Stroke August 2005
TABLE 2. Criteria and Indicators for Chart Review Diagnosis of Stroke Risk Factors
Risk Factor Criteria Indicators
Atrial fibrillation Paroxysmal or chronic AFIB History, ECG, holter monitoring
Coronary artery disease/ischemic heart disease Angina, myocardial infarction, or coronary atherosclerosis Angioplasty, stenting, CABG, ECHO, medication
Diabetes mellitus Type 1 diabetes or type 2 diabetes Fasting blood glucose History, medication, lab results
ⱖ7.0 mmol/L (126 mg/dL)
History of cerebrovascular disease Stroke, TIA, stroke(s), or TIA(s) History, imaging reports, medication
Hyperlipidemia Total cholesterol ⬎5.0 mmol/L (193 mg/dL), LDL History, medication
cholesterol ⬎3.0 mmol/L (116 mg/dL) lipoprotein analysis
done after a 9- to 12-hour fast
Hypertension Without diabetes: systolic BP ⱖ140 mm Hg; diastolic BP History, blood pressure readings in the
ⱖ90 mm Hg. With diabetes: systolic BP ⱖ135 mm Hg; nonacute phase, ⱖ1 week from admission,
diastolic BP ⱖ85 mm Hg average of ⱖ2 readings on ⱖ2 visits after
initial presentation, medication
Renal failure Serum creatinine ⱖ100 mmol/L (1.1 mg/dL) Hemodialysis, peritoneal dialysis, medication
Tobacco use Ex-smoker (daily for ⱖ1 year) or current smoker History
AFIB indicates atrial fibrillation; CABG, coronary artery bypass grafting; LDL, low-density lipoprotein; BP, blood pressure; ECHO, echocardiography.
(no neurovascular imaging was done, or no imaging substantially increased the accuracy of ICD-9 coding for
reports were available for review) and on clinical data and ischemic stroke when including code 433 in the stroke
neurovascular imaging reports in 75.6% of charts. definition.
On the whole, ICD-9 coding was excellent with 90% (CI95, At the hospital level, within each coding system, coding
86 to 92) correct; ⫽0.86 (CI95, 0.81 to 0.91). ICD-10 was was nonsignificantly better using ICD-10 at 95% (CI95, 91 to
TABLE 3. PPV and Agreement of Stroke Type Coding by ICD Coding and Hospital Site
ICD-9 ICD-10
No. Coding Based on Chart Review No. Coding Based on Chart Review
Stroke Type
AIS 18 (123)* 16-TIA 4-weakness unspecified 4-other 2-admitted to rehabilitation nonacutely 2 2-TIA
1-SAH 1-missing chart (95-admitted for elective carotid endarterectomy)*
ICH 0 12 8-AIS 3-SAH 1-missing chart
SAH 3 2-ICH 1-admitted to rehabilitation nonacutely 1 1-ICH
TIA 1 1-AIS 6 5-AIS 1-admitted to rehabilitation non-acutely
*Carotid endarterectomy patients were identified correctly when using the fourth and fifth digits of the ICD-9 coding system. They were coded as vascular occlusion
without infarct. They are included here to emphasize what the specific use of the fourth and fifth digits are in our sample.
98) accuracy at the university site compared with 80% (CI95, Part II
68 to 90) accuracy at the community sites (P⫽0.503). A total of 137 charts from 2000/2001 (ICD-9) and 112 from
Differences were less marked with ICD-9 (90% versus 89% 2002/2003 (ICD-10) were randomly selected for review. The
accuracy; P⫽0.883). median age of patients was 73 (IQR, 65 to 81), and 49% were
TABLE 5. Sensitivity, Specificity, PPV, and Agreement of Stroke Risk Factors by ICD Coding and Hospital Site
ICD-9 ICD-10
CAD/IHD 89 (71–98) 94 (83–99) 92 (84–97) 0.83 (0.76–0.91) 86 (65–97) 97 (88–99) 94 (86–98) 0.84 (0.84–0.92)
DM 94 (69–99)3 98 (91–99) 98 (91–99) 0.92 (0.87–0.98) 87 (66–97) 100 (94)* 96 (89–99) 0.91 (0.85–0.97)
History of CVA 44 (25–65) 100 (93)* 81 (70–89) 0.51 (0.41–0.61) 48 (29–68) 93 (82–98) 78 (67–87) 0.45 (0.35–0.56)
Hyperlipidemia 70 (44–89) 98 (91–99) 92 (84–97) 0.75 (0.67–0.84) 67 (51–81) 100 (91)* 83 (73–90) 0.67 (0.57–0.77)
Hypertension 84 (71–94) 88 (71–97) 86 (76–93) 0.72 (0.63–0.81) 95 (85–99) 88 (67–97) 93 (84–97) 0.82 (0.75–0.90)
RF 50 (6–93) 99 (94–99.9) 96 (89–99) 0.55 (0.34–0.76) 66 (22–96) 100 (95)* 98 (91–99) 0.79 (0.63–0.95)
Tobacco use 0 (9)† 98 (87–99) 52 (40–63) -0.03 (⫺0.09–0.05) 25 (13–40) 100 (98)* 60 (48–71) 0.24 (0.16–0.31)
Community hospital sites
AFIB 90 (55–99) 100 (92)* 98 (90–99) 0.94 (0.87–1.0) 63 (24–91) 100 (83)* 90 (72–98) 0.71 (0.58–0.84)
CAD/IHD 77 (54–92) 83 (67–94) 81 (68–90) 0.60 (0.50–0.71) 90 (55–99) 100 (82)* 97 (82–99) 0.92 (0.87–0.97)
DM 90 (66–99) 100 (90)* 97 (93)* 0.92 (0.86–0.98) 100 (69)* 95 (73–99) 97 (82–99) 0.93 (0.87–0.98)
History of CVA 13 (3–30) 100 (87)* 53 (39–67) 0.12 (0.02–0.23) 10 (0–44) 100 (82)* 69 (49–85) 0.13 (0.02–0.23)
Hyperlipidemia 36 (11–69) 100 (78)* 88 (76–95) 0.48 (0.36–0.60) 40 (12–74) 100 (82)* 79 (60–92) 0.47 (0.35–0.59)
Hypertension 79 (63–90) 100 (78) 85 (72–93) 0.66 (0.56–0.76) 74 (51–90) 67 (22–96) 72 (52–87) 0.33 (0.21–0.45)
RF 25 (3–65) 98 (89–99) 88 (77–95) 0.31 (0.09–0.53) 50 (1–99) 96 (81–99) 93 (77–99) 0.46 (0.24–0.68)
Tobacco use 17 (4–37) 100 (89)* 66 (51–77) 0.19 (0.11–0.27) 43 (9–81) 100 (84)* 86 (68–96) 0.53 (0.46–0.60)
All sites
AFIB 81 (58–96) 99 (95–99.9) 96 (91–99) 0.85 (0.76–0.95) 98 (92–99) 96 (89–99) 98 (92–99) 0.81 (0.71–0.92)
CAD/IHD 84 (70–93) 90 (81–95) 88 (81–93) 0.73 (0.64–0.82) 98 (4–19) 95 (88–98) 98 (4–19) 0.87 (0.80–0.92)
DM 91 (76–98) 99 (94–99.9) 97 (92–99) 0.92 (0.87–0.98) 99 (93–99.9) 97 (91–99) 99 (93–99.9) 0.91 (0.86–0.97)
History of CVA 28 (16–41) 100 (95)* 69 (61–77) 0.31 (0.20–0.41) 95 (87–99) 76 (67–83) 95 (87–99) 0.37 (0.27–0.48)
Hyperlipidemia 57 (37–76) 99 (95–99.9) 91 (84–95) 0.66 (0.56–0.76) 100 (94)* 82 (74–89) 100 (94)* 0.63 (0.53–0.74)
Hypertension 82 (72–89) 92 (80–98) 85 (78–91) 0.70 (0.60–0.79) 83 (65–94) 88 (79–93) 83 (65–94) 0.69 (0.60–0.79)
RF 33 (9–65) 98 (94–99) 93 (87–96) 0.41 (0.19–0.63) 99 (94–99.9) 96 (91–99) 99 (94–99.9) 0.70 (0.51–0.88)
Tobacco use 7 (1–16) 99 (92–99.9) 58 (48–66) 0.06 (⫺0.02–0.13) 100 (94)* 67 (57–76) 100 (94)* 0.29 (0.22–0.37)
AFIB indicates atrial fibrillation; CAD/IHD, coronary artery disease/ischemic heart disease; DM, diabetes mellitus; CVA, cerebrovascular accident; RF, renal failure.
*97.5% 1-sided lower CI; †97.5% 1-sided upper CI.
1780 Stroke August 2005
female. Table 5 summarizes the sensitivity, specificity, and had not had an index stroke on that admission. In the ICD-9
PPV of stroke risk factors according to ICD system and system for coding AIS, there are several codes that are
hospital site. Overall, coding of all risk factors was found to described as acute arterial occlusion without infarct (433.00,
be similar between ICD-9 and ICD-10. Global sensitivity was 433.10, 433.20, 433.30, 434.10, and 434.90). In our study, all
lower, with ICD-10 at 58% (CI95, 52 to 63) compared with 95 cases of elective carotid endarterectomy were coded as
67% (CI95, 61 to 72); P⫽0.234. Overall specificity was equal acute arterial occlusion without infarct. As suggested by
at 97% (CI95, 96 to 98) compared with 97% (CI95, 96 to 99); Goldstein, these codes have particularly low accuracy for the
P⫽1.000. Overall accuracy was similar at 84% (CI95, 82 to diagnosis of true stroke. This problem does not occur with
87) compared with 87% (CI95, 84 to 89); P⫽0.691. Charac- ICD-10.
teristic of both schemes, atrial fibrillation, coronary artery The identification of stroke risk factors is more variable.
disease/ischemic heart disease, diabetes mellitus, and hyper- Atrial fibrillation, coronary artery disease/ischemic heart
tension coding was very good to excellent, showing a high disease, diabetes mellitus, and hypertension are identified
degree of sensitivity and specificity. Conversely, coding of a with a high degree of confidence, whereas history of cere-
history of cerebrovascular accident, hyperlipidemia, renal brovascular disease, hyperlipidemia, renal failure, and to-
failure, and tobacco use ranged from poor to excellent, having bacco use are identified to a lesser degree. The poor coding of
quite low sensitivity but high specificity. The sensitivity the latter 4 risk factors may be attributable to poor charting by
improved only slightly for these risk factors with the switch physicians and nursing staff, a lack of perceived importance
to ICD-10. These trends were observed at the hospital level. by health technologist coders, or a lack of time to “code
No differences were observed between the university and everything.” Education and understanding may help to im-
community hospital sites. prove this situation. The emergence of the electronic health
record may allow automated and better coding of such risk
Discussion factors within administrative databases. The ICD-10 system
Our data suggest that the administrative diagnoses of stroke
itself might benefit from the inclusion of more specific
and its risk factors are quite good and that no quantitative
diagnostic codes for these comorbidities to improve their true
improvements have been realized with the switch to the
diagnosis. However, the ability to reliably identify stroke risk
ICD-10 system. Two caveats associated with using such data
factors among stroke patients using administrative data is an
to make reliable conclusions are worth review. First, by
important addition to use of such data in health services
definition, they apply only to hospital-based care. Therefore,
research.
data from those individuals who do not seek medical attention
Our study has notable limitations. We were unable to blind
Downloaded from http://ahajournals.org by on September 25, 2019
5. Mayo NE, Neville D, Kirkland S, Ostbye T, Mustard CA, Reeder B, (ICD-9-CM) in identifying the diagnosis of ischemic cerebrovascular
Joffres M, Brauer G, Levy AR. Hospitalization and case-fatality rates for disease. Neurology. 1997;49:660 – 664.
stroke in Canada from 1982 through 1991. The Canadian Collaborative 10. Leibson CL, Naessens JM, Brown RD, Whisnant MD. Accuracy of
Study Group of Stroke Hospitalizations. Stroke. 1996;27:1215–1220. hospital discharge abstracts for identifying stroke. Stroke. 1994;25:
6. Ostbye T, Levy AR, Mayo NE. Hospitalization and case-fatality rates for 2348 –2355.
subarachnoid hemorrhage in Canada from 1982 through 1991. The 11. Piriyawat P, Smajsova M, Smith MA, Pallegar S, Al-Wabil A, Garcia
Canadian Collaborative Study Group of Stroke Hospitalizations. Stroke. NM, Risser JM, Moye LA, Morgenstern LB. Comparison of active and
1997;28:793–798. passive surveillance for cerebrovascular disease: the Brain Attack Sur-
7. Reker DM, Hamilton BB, Duncan PW, Yeh SC, Rosen A. Stroke: who’s veillance in Corpus Christi (BASIC) Project. Am J Epidemiol. 2002;156:
counting what? J Rehabil Res Dev. 2001;38:281–289. 1062–1069.
8. Goldstein LB. Accuracy of ICD-9-CM coding for the identification of 12. Yiannakoulias N, Svenson LW, Hill MD, Schopflocher DP, Rowe
patient with acute ischemic stroke: effect of modifier codes. Stroke. BH, James RC, Wielgosz AT, Noseworthy TW. Incident cerebrovascular
1998;29:1602–1604. disease in rural and urban Alberta. Cerebrovasc Dis. 2004;17:72–78.
9. Benesch C, Witter DM Jr, Wilder AL, Ducan PW, Samsa GP, Matchar 13. Field TS, Green TL, Roy K, Pederson J, Hill MD. Trends in stroke
DB. Inaccuracy of the International Classification of Diseases occurrence in Calgary. Can J Neurol. 2004;31:387–393.
Downloaded from http://ahajournals.org by on September 25, 2019