(Stroke. 2001;32:1487.)
© 2001 American Heart Association, Inc.
Original Contributions |
Possible Effect of DRGs on the Classification of Stroke
Implications for Epidemiological Surveillance
From the New England Research Institutes (C.A.D., H.A.F.), Watertown, Mass; the Department of Community Health (K.L.L.), Brown University School of Medicine, Providence, RI; and the Division of Cardiology, Memorial Hospital of Rhode Island, Pawtucket, and the Department of Medicine (R.A.C.), Brown University School of Medicine, Providence, RI.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background and Purpose—Accurate data on the distribution of stroke subtypes are essential for understanding the forces driving recent morbidity and mortality trends. The introduction of diagnosis-related groups (DRGs) in the 1980s may have affected the distribution of stroke subtypes as defined by International Classification of Diseases, Ninth Revision (ICD-9), discharge diagnosis codes.
Methods—The Pawtucket Heart Health Program cardiovascular surveillance data were used to examine trends in stroke classification for 1980 to 1991 in relation to the introduction of DRGs in 2 communities in Massachusetts and Rhode Island, where DRGs were implemented 2 years apart. Included were all hospital discharges for residents aged 35 to 74 with a primary ICD-9 diagnosis of 431 to 432, 434, or 436 to 437 (N=1386 in Rhode Island, N=1839 in Massachusetts).
Results—In each state,
concurrently with the introduction of DRGs, the proportion of strokes
classified as cerebral occlusion (ICD-9 434.0 to 434.9) increased, and
the proportion classified as acute but ill-defined (ICD-9 436.0 to
436.9) decreased. Before DRGs, 30.0% of strokes in Rhode Island and
26.6% in Massachusetts were classified as cerebral occlusion, whereas
51.8% in Rhode Island and 51.7% in Massachusetts were classified as
acute ill defined. After DRGs were instituted, the proportions of
cerebral occlusion and acute, ill-defined stroke, respectively, were
70.9% and 8.5% in Rhode Island and 74.1% and 7.7% in Massachusetts
(
2, all
P<0.001). The proportions of
strokes classified as intracerebral hemorrhage
or transient cerebral ischemia remained
constant.
Conclusions—The implementation of DRGs may have influenced coding of strokes to the ICD-9. Findings highlight the limitations of hospital discharge data for evaluating stroke subtypes and demonstrate the need for community-based surveillance for monitoring specific trends in stroke.
Key Words: cerebrovascular disorders • diagnostic-related groups • population surveillance • stroke
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
National data and studies within communities have consistently demonstrated declining mortality from stroke in the United States beginning in the 1960s.1 2 3 4 5 Although mortality has continued to decline, there is evidence that morbidity rates may have stabilized or even increased beginning in the 1980s.1 2 5 6 7 8 9 Although the patterns of stroke over time have been consistent across studies, the forces driving these trends are complex and remain a topic of debate, with the relative contributions of primary and secondary prevention, advances in diagnostic and treatment technologies, and changes in diagnosis and classification unclear.7 9 10
Reliable data on the distribution of stroke subtypes provide essential information for determining the forces driving these trends. However, most available data rely on hospital discharge diagnoses, which may be influenced by trends in hospital reimbursement, coding, and classification, as well as by changes in the natural history of a disease.11 12 13 In particular, the introduction of diagnosis-related groups (DRGs) in the 1980s may have influenced the distribution of stroke subtypes as defined by International Classification of Diseases, Ninth Revision (ICD-9),14 discharge codes.10 13 15 16 The Pawtucket Heart Health Program (PHHP) morbidity and mortality surveillance system provided a unique opportunity to examine the impact of DRGs on the classification of stroke subtypes by using hospital discharge information as well as validated stroke rates for 2 cities in states where DRGs were introduced 2 years apart.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The Pawtucket Heart Health Program
The PHHP was a community demonstration project conducted between 1980 and 1992 to evaluate whether a comprehensive program of community-based health education produced favorable changes in risk factor levels and the occurrence of coronary heart disease and stroke.17 18 The study included 2 communities in southeastern New England, Pawtucket, Rhode Island, the intervention community, and a comparison city in Massachusetts that had similar demographic characteristics at baseline. The communities were selected on the basis of size and population stability. Pawtucket has a population of
72 000, whereas the
comparison city is a community of 100 000
(Table 1
). The identity of the comparison
community is not given in accordance with an agreement made with the
local government at the study’s inception. Detailed descriptions of
the study design have been reported
previously.17 18 19
Between 1980 and 1992, the PHHP conducted surveillance for
cardiovascular disease in each study community. DRGs
were introduced in Rhode Island on October 1, 1983 and in Massachusetts
on October 1, 1985.
|
Case Identification and Validation
Potential cases were identified from hospital
discharge tapes from the 7 hospitals in Rhode Island and Massachusetts
that serve the study communities. On the basis of data from the state
hospital associations, these hospitals were estimated to provide 96%
coverage of care for residents of the study communities. The
analysis includes all discharges for a resident of a study
community, aged 35 to 74 years, with a primary ICD-9 discharge
diagnosis code of 431 (intracerebral
hemorrhage), 432 (other unspecified cerebral
hemorrhage), 434 (occlusion of cerebral arteries), 436 (acute,
ill-defined cerebrovascular disease), or 437 (other ill-defined
cerebrovascular disease).
For each eligible case, portions of the medical record were copied, including the discharge summary, CT and/or MRI scans, neurology consultations, and autopsy reports. The study physician then reviewed all available information to determine an outcome of definite stroke, probable stroke, no ascertainable stroke, or inadequate information. This classification was based on whether (1) the clinical description was consistent with a new, localized neurological defect involving the hemispheres, brain stem, and/or the cerebellum and (2) whether there was evidence for intracerebral hemorrhage with or without intraventricular extension and with or without subarachnoid extension. The following criteria were considered exclusions from the categories of definite or probable stroke: (1) exclusively subarachnoid hemorrhage; (2) cerebral infarction related to rheumatic mitral stenosis or infective endocarditis, or stroke in the presence of prosthetic cardiac valves; (3) exclusively transient ischemic attack; and (4) evidence from the medical history that the hospitalization was for a previous stroke.
Population Estimates for Calculation of Event
Rates
Midyear population estimates by sex, age, and city
were prepared for each of the years 1980 to 1991 by using standard
demographic techniques.20
The cohort-component method was used to take single-year age groups
created from the 1980 US Census and project them forward 1 year at
a time by using national survival rates provided by the US Bureau of
the Census. Births were added for each year by using information
obtained from the Departments of Vital Statistics reports of Rhode
Island and Massachusetts. Finally, the population estimates were
corrected for immigration and outmigration by using estimates based on
the 1980 census data and corrected to the 1990 census.
Statistical Methods
The dependent variable in each analysis
was an event rate per 10 000 population per year, calculated for a
single calendar quarter and a single city. Because previous
analyses have demonstrated that the age and sex distributions
of the population remained stable across the decade studied, data were
aggregated by age and sex. The rate was constructed by counting the
number of recorded events in the appropriate time period and city,
multiplying by 4 to obtain an annualized count, and dividing by a
separately determined estimate of population (in 10 000s) in that city
for that time period. The unit of analysis was thus a group
defined by time and location, and the total sample size for each
specific event rate for the 12-year study was 48 (quarters) for each
city. In secondary analyses, the data were stratified by age
(35 to 64 years and 65 to 74 years), and the strata were
analyzed separately.
In exploratory analysis, the outcome variables with higher event rates showed proportionally higher fluctuations from quarter to quarter, similar to the variability of a Poisson-distributed random deviate. The rates were therefore square-root transformed for analysis to homogenize the residual variance across subgroups. After analysis, adjusted mean square-root rates were converted back to natural units for presentation. The transformed rates were analyzed separately for each city by segmented linear regression with independent variables representing time and period (pre/post DRG). The pre-DRG periods in Rhode Island and Massachusetts were defined as January 1, 1980 to September 30, 1983 and January 1, 1980 to September 30, 1985, respectively.
To test the null hypothesis that the time trend in a particular stroke subtype and city did not change with the advent of DRGs, we compared 2 regression models. In the first model, a single time trend was applied across the entire study period (46 degrees of freedom). In the second model, the slope and intercept were allowed to differ between pre- and post-DRG periods (44 degrees of freedom). The 2 models were compared by Fisher’s F test (df=2,44). A statistically significant result indicated that the trends before and after DRGs were different.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
A total of 2124 discharges met study inclusion criteria, with a primary ICD-9 discharge code of 431, 432, 434, or 436 to 437. Record review was completed for 97% of these cases, of which 80% were confirmed definite or probable strokes. Table 1
displays the numbers of cases in the pre- and
post-DRG periods for each state. There was no evidence for a
differential trend in rates of total stroke (ICD-9 431 to 432, 434, 436
to 437) in the pre- and post-DRG periods in either state
(Figure 1). This pattern was similar for rates of
validated stroke.
|
Although the total stroke rates remained constant over the
period, the distribution of stroke categories shifted, such that the
rates of stroke coded as cerebral artery occlusion (ICD-9 434)
increased while the rates of ill-defined cerebrovascular disease (ICD-9
436) decreased in both states
(Table 2). Prior to DRGs, 30% of strokes in Rhode
Island and 26.6% of strokes in Massachusetts were classified as
cerebral artery occlusion (ICD-9 code 434). After the introduction of
DRGs, the proportion of cases classified as cerebral occlusion
increased to 70.9% in Rhode Island and to 74.1% in Massachusetts. In
contrast, the proportion of strokes classified as ill defined declined
in both states, from 51.8% to 8.5% in Rhode Island and from 51.7% to
7.7% in Massachusetts. The proportions of stroke classified as
cerebral hemorrhage (ICD-9 431 or 432) remained fairly constant
in the pre- and post-DRG periods.
Figure 2 displays the rates of cerebral artery
occlusion (ICD-9 434) and of ill-defined cerebrovascular disease (ICD-9
436) across the study period. In Massachusetts, the trends in rates
were significantly different in the pre- and post-DRG periods
(P<0.001). A similar pattern
was observed in Rhode Island, although the difference in trends did not
reach statistical significance.
|
|
Given that DRGs were instituted as part of the Medicare
reimbursement system, any influence of DRGs on coding practices would
be expected to be most apparent among patients aged 65 or more years.
To test this hypothesis, cases were stratified according to age 35 to
64 years or 65 to 74 years, and the rates of cerebral occlusion and of
ill-defined cerebrovascular disease were examined separately in each
stratum
(Figure 3). Patterns were similar to those for all ages
combined, with an increase in rates for cerebral occlusion and a
reduction in the rates for ill-defined cerebrovascular disease. Among
persons aged 65 to 74 years, the differences in trends were
statistically significant for the pre- versus post-DRG periods in both
states (P<0.02). In the
younger age group, the shift was similar, although differences reached
statistical significance only in Massachusetts.
|
We have also documented an increase in the use of CT scans and MRI during the 1980s in the study communities.5 To test whether temporal trends in diagnostic technology may have influenced the specificity of stroke diagnosis during this period, analyses were repeated with adjustment for the proportion of cases with diagnostic imaging documented in the medical record. Results of these analyses did not differ substantially from those presented.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Previous reports have highlighted the limitations of using discharge diagnoses to assess trends in types of stroke.13 21 22 23 These limitations may be further exacerbated by the influence of reimbursement systems on discharge coding.15 The DRG prospective payment system changed the application of ICD-9, designed to capture data on disease characteristics, from its original use to claims payment.24 We had the unique opportunity to examine temporal trends in the classification of stroke in relation to the introduction of DRGs and to replicate the analysis in 2 states in which prospective payment was introduced 2 years apart.
Although ICD-9 coding and assignment of DRGs are theoretically separate processes, these data suggest that the implementation of DRGs may have influenced the coding of stroke subtypes, with an increase in classification as the more specific category of cerebral artery occlusion and a concomitant decline in the use of the category "ill-defined" stroke. These results are consistent with data from the National Hospital Discharge Survey, which also show an increase in the rate of occlusion of the cerebral arteries (ICD-9 434) and a decrease in the rate of acute, ill-defined cerebrovascular disease in the mid-1980s, when DRGs were introduced.16
Reimbursement for DRGs is determined by assigning a standard cost weight to each DRG and multiplying this weight by a standard cost, which is adjusted for institutional characteristics.25 The higher the cost weight for a DRG, the greater is the hospital reimbursement. Simborg26 introduced the concept of "DRG creep" to describe the potential for the prospective payment system to prompt a systematic shift in discharge coding toward diagnoses with increased severity and to those with higher cost weights. However, the codes examined in this analysis are all included under a single prospective payment category, DRG 14.27 Thus, financial incentives, or DRG creep, do not appear to explain the observed classification shift.
The introduction of DRGs may have prompted a generalized shift toward more accurate and specific discharge coding as hospitals sought to maximize reimbursement.12 15 Previous studies have demonstrated that a high proportion of stroke cases coded as unspecified actually have an etiology identified in the discharge summary.13 21 22 In a review of stroke registries, Yip et al28 found that the higher the relative incidence of stroke with undetermined subtype, the lower the rate of stroke classified as large-artery atherosclerotic type. Goldstein et al21 reported that nearly half of the cases assigned a code for acute, ill-defined cerebrovascular disease had a discharge summary that included an established etiology. Similarly, studies by Ellekjaer et al22 and by Leibson et al13 have reported that discharge diagnoses tend to overestimate the proportion of strokes classified as unspecified relative to that obtained when a neurologist reviews the chart. In the Rochester Stroke Registry for the period 1980 to 1989, 29% of all confirmed strokes were assigned ICD-9 code 436, "acute but ill-defined cerebrovascular disease." However, a neurologists’ review indicated that only 10% of these cases lacked sufficient information to be assigned to a specific type.13 Thus, the observed classification shift in the PHHP database may be the result of increased coding specificity and accuracy prompted by the introduction of DRGs.
The results of age-stratified analyses suggest that the advent of DRGs brought about a general increase in the specificity of discharge coding for stroke regardless of Medicare status. The trend toward increased use of the more specific ICD-9 434 codes and decreased use of code 436 occurred even among patients aged 35 to 64, particularly in Massachusetts. Given that medical records staff are unlikely to apply separate coding practices according to age or Medicare status, educational activities and training prompted by DRGs may have influenced all discharge coding. This concept is consistent with prior analyses demonstrating potential DRG-related shifts in discharge coding for ischemic heart disease in this population, regardless of Medicare status.12
During this period, the use of CT scans and MRI for stroke patients increased dramatically in the study communities and nationally.5 29 30 Thus, it is possible that a shift toward more specific coding of strokes resulted from the availability of improved diagnostics. However, the proliferation of these diagnostic technologies and the increase in the proportion of cases with neurology consults began in the pre-DRG period, and the rate of increase showed signs of slowing in the mid-1980s.5 Both states are likely to have been influenced similarly by advances in diagnostics, yet the "bump" in classification in the 2 states was offset by 2 years, corresponding to the lag in the introduction of DRGs. Finally, analyses that adjusted for the prevalence of CT or MRI did not alter our results.
The present study did not ascertain information on strokes in person over the age of 74. Thus, given the increase in stroke rates with age, many stroke patients are likely to have been excluded. However, because we were able to demonstrate that the observed change in coding was consistent across age groups, the lack of data from older age groups is not likely to have affected the validity of our conclusions.
Increased information regarding the pathophysiology of stroke subtypes would be beneficial for prevention strategies and would provide useful information regarding the forces driving recent morbidity and mortality trends. Although shifts within the broader category of stroke are not likely to influence overall statistics for total stroke, the observed shifts highlight the limitations of hospital discharge data for documenting trends in stroke subtypes. In the present study, an examination of ICD-9 code 434 alone would suggest an increasing rate of stroke resulting from cerebral occlusion. However, the concomitant decline in the rate of unspecified stroke together with the relatively stable total stroke rates suggests that a classification shift rather than a shift in the natural history underlies the trend. These results add to the body of knowledge illustrating the limitations of hospital discharge data for monitoring stroke trends and highlight the need for standardized stroke surveillance databases.
|
Footnotes |
|---|
Reprint requests to Carol A. Derby, PhD, New England Research Institutes, 9 Galen St, Watertown, MA 02472.
This work was supported by grant Nos. HL55448 and HL23629 from the National Heart, Lung, and Blood Institute of the National Institutes of Health, US Department of Health and Human Services, Bethesda, Md.
Received November 6, 2000; revision received March 21, 2001; accepted April 9, 2001.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
-
Cooper
R, Sempos C, Hsieh S, Kovar MG. Slowdown in the decline of stroke
mortality in the United States, 1978–1986.
Stroke. 1990;21:1274–1279.
[Abstract/Free Full Text] -
Broderick JP,
Phillips SJ, Whisnant JP, O’Fallon WM, Bergstralh EJ. Incidence
rates of stroke in the eighties: the end of the decline in stroke?
Stroke. 1989;20:577–582.
[Abstract/Free Full Text] -
Klag MJ,
Whelton PK, Seidler AJ. Decline in US stroke mortality: demographic
trends and antihypertensive treatment.
Stroke. 1989;20:14–21.
[Abstract/Free Full Text] - Whelton PK. Declining mortality from hypertension and stroke. South Med J. 1982;75:33–38.[Medline] [Order article via Infotrieve]
-
Derby CA, Lapane
KL, Feldman HA, Carleton RA. Trends in validated cases of fatal
and nonfatal stroke, stroke classification, and risk factors in
southeastern New England, 1980 to 1991: data from the Pawtucket Heart
Health Program. Stroke. 2000;31:875–881.
[Abstract/Free Full Text] - Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. The Sixth Report of The Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Rockville, Md: US Dept of Health and Human Services, Public Health Service, National Institutes of Health, National Heart, Lung, and Blood Institute. 1997. NIH publication No. 98–4080.
- McGovern PG, Burke GL, Sprafka JM, Xue S, Folsom AR, Blackburn H. Trends in mortality, morbidity, and risk factor levels for stroke from 1960 through 1990: the Minnesota Heart Survey. JAMA. 1992;268:753–759.[Abstract]
-
Wolf PA,
D’Agostino RB, O’Neal A, Stykowski P, Kase CS, Belanger AJ, Kannel
WB. Secular trends in stroke incidence and mortality: the Framingham
Study. Stroke. 1992;23:1551–1555.
[Abstract/Free Full Text] -
Modan B, Wagener
DK. Some epidemiological aspects of stroke: mortality/morbidity trends,
age, sex, race, socioeconomic status.
Stroke. 1992;23:1230–1236.
[Abstract/Free Full Text] -
Shahar E,
McGovern PG, Sprafka M, Pankow JS, Doliszny KM, Luepker RV, Blackburn
H. Improved survival of stroke patients during the 1980s: the Minnesota
Stroke Survey. Stroke. 1995;26:1–6.
[Abstract/Free Full Text] - Romano PS, Mark DH. Bias in the coding of hospital discharge data and its implications for quality assessment. Med Care. 1994;32:81–90.[Medline] [Order article via Infotrieve]
-
Assaf AR, Lapane
KL, McKenney JL, Carleton RA. Possible influence of the prospective
payment system on the assignment of discharge diagnoses for
coronary heart disease. N
Engl J Med. 1993;329:931–935.
[Abstract/Free Full Text] - Leibson CL, Naessens JM, Brown RD, Whisnant JP. Accuracy of hospital discharge abstracts for identifying stroke. Stroke. 1994;25:2348–2355.[Abstract]
- World Health Organization. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, Based on the Ninth Revision Conference, 1975. Geneva, Switzerland: World Health Organization; 1977.
-
Cohen BB,
Pokras R, Meads MS, Krushat WM. How will diagnosis-related groups
affect epidemiologic research? Am J
Epidemiol. 1987;126:1–9.
[Free Full Text] - Centers for Disease Control. Hospital discharge rates for cerebrovascular disease–United States, 1970–1986. MMWR. 1989;38:194–201.[Medline] [Order article via Infotrieve]
- Carleton RA, Lasater TM, Assaf AR, Lefebvre RS, McKinlay SM. The Pawtucket Heart Health Program, I: an experiment in population-based disease prevention. RI Med J. 1987;70:533–538.
-
Carleton RA,
Lasater TM, Assaf AR, Feldman HA, McKinlay S, and the Pawtucket
Heart Health Program Writing Group. The Pawtucket Heart Health Program:
community changes in cardiovascular risk factors and
projected disease risk. Am J
Public Health. 1995;85:777–785.
[Abstract/Free Full Text] -
McKinlay SM,
Carleton RA, McKenney JL, et al. A new approach to surveillance for
acute myocardial infarction: reproducibility and cost efficiency.
Int J Epidemiol. 1989;18:67–75.
[Abstract/Free Full Text] - Raymondo JC. How to estimate population. Am Demogr. 1989;11:46–49.
-
Goldstein LB.
Accuracy of ICD-9-CM coding for the identification of patients with
acute ischemic stroke: effect of modifier codes.
Stroke. 1998;29:1602–1604.
[Abstract/Free Full Text] -
Ellekjaer H,
Holmen J, Kruger O, Terent A. Identification of incident stroke in
Norway: hospital discharge data compared with a population-based stroke
register. Stroke. 1999;30:56–60.
[Abstract/Free Full Text] -
Benesch C, Witter
DM, Wilder AL, Duncan PW, Samsa GP, Matchar DB. Inaccuracy of the
International Classification of Diseases (ICD-9-CM) in identifying the
diagnosis of ischemic cerebrovascular disease.
Neurology. 1997;49:660–664.
[Abstract/Free Full Text] - Hsia DC, Krushat WM, Fagan AB, Tebbutt JA, Kusserow RP. Accuracy of diagnostic coding for Medicare patients under the prospective-payment system. N Engl J Med. 1988;318:352–355.[Abstract]
- Lagoe RJ, Lauko SJ. Stroke hospitalization under prospective payments: analysis of diagnosis related group 14. Arch Phys Med Rehabil. 1985;66:773–776.[Medline] [Order article via Infotrieve]
- Simborg DW. DRG creep: a new hospital-acquired disease. N Engl J Med. 1981;304:1602–1604.[Medline] [Order article via Infotrieve]
- Yale School of Organization and Management. The New ICD-9-CM Diagnosis-Related Group Classification Scheme, Definitions of Diagnosis-Related Groups. New Haven, Conn: Health Systems Management Group; MDC 01. 1981.
-
Yip P, Jeng J,
Lee T, Chang Y, Huang Z, Ng S, Chen R. Subtypes f ischemic
stroke: a hospital-based registry in Taiwan (SCAN-IV).
Stroke. 1997;28:2507–2512.
[Abstract/Free Full Text] - Kovar MG, Pokras R, Collins JG. Trends in medical care and survival from stroke. Ann Epidemiol. 1993;3:466–470.[Medline] [Order article via Infotrieve]
- Graves EJ. CAT Scan Use in Short-Stay Non-Federal Hospitals: United States, 1979–1982. Hyattsville, Md: US Dept of Health and Human Services, Public Health Service, National Center for Health Statistics; 1984.
This article has been cited by other articles:
 |
|
 |
|
  P. Piriyawat, M. Smajsova, M. A. Smith, S. Pallegar, A. Al-Wabil, N. M. Garcia, J. M. Risser, L. A. Moye, and L. B. Morgenstern Comparison of Active and Passive Surveillance for Cerebrovascular Disease: The Brain Attack Surveillance in Corpus Christi (BASIC) Project Am. J. Epidemiol., December 1, 2002; 156(11): 1062 - 1069. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||