(Stroke. 1997;28:1142-1146.)
© 1997 American Heart Association, Inc.
Articles |
Epidemiology and Costs of Acute Hospital Care for Cerebrovascular Disease in Diabetic and Nondiabetic Populations
From the Department of Public Health Medicine, Bro Tâf Health Authority, Cardiff (C.J.C.); Department of Medicine, University Hospital of Wales, Cardiff (C.L.M., J.R.P.); Unit of Health Care Sciences, University of Oxford (England) (L.G.); and Department of General Practice, University of Wales College of Medicine, Cardiff (N.C.H.S.) (UK).
Correspondence to Dr John Peters, University Hospital of Wales, Heath Park, Cardiff, Wales, UK CF4 4XW. E-mail wmdcjc{at}cf.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Background and Purpose Little is known about the pattern of cerebrovascular disease (CVD) for diabetic and nondiabetic patients or about the cost of treatment for CVD in the United Kingdom. The purpose of this study was to extend previous work to describe the epidemiology and cost of acute care of CVD as a frequent comorbidity of diabetes in a UK population (408 000 people).
Methods Routine data describing inpatient care for a 4-year period were analyzed (financial years 1991/1992 to 1994/1995). These data had undergone record linkage to draw together records from the same patients. Cost estimates were determined by attributing a diagnosis-related group cost weight to each record. Mortality data from an overlapping period were supplied by the Office of Population Censuses and Surveys.
Results There were 11 196 CVD admissions (3.1% of all admissions). Of these, 7351 (66%) were primary diagnoses. These admissions were generated by 5358 patients (3904 primary diagnosis). For people with diabetes, the incidence rate was between 23 and 32.8 per 1000 per year compared with 2.4 to 3.3 per 1000 for the population as a whole, depending on the use of primary and subsidiary codes. The age-adjusted relative risk of stroke in diabetic men versus nondiabetic men was 3.70 (95% confidence interval, 3.53 to 3.88) and in women was 4.35 (95% confidence interval, 4.37 to 4.76). We describe other epidemiological relationships. The cost of CVD is between £1.1 and £1.6 million per 100 000 population—at least £0.7 million per 100 000 for CVD alone. Approximately 15% of this value is related to diabetes, and an estimated 94% of this diabetes-related expenditure is potentially avoidable.
Conclusions CVD represents a major source of expenditure for health services, and diabetes is confirmed as a major risk factor within this disease group. Differences between diabetic and nondiabetic inpatient patterns of CVD may reflect greater incidence of comorbidities in the former.
Key Words: costs and cost analysis • diabetes mellitus • epidemiology • hospitalization
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
In a study of 16 European and two Asian populations, the age-standardized incidence rates of stroke varied from 101 to 285 and from 47 to 198 per 100 000 in men and women, respectively.1 The fatality rate at 28 days varied from 15% to 49% among men and 18% to 57% among women.
The epidemiology of diabetes-related CVD is not described separately and is often grouped together with that of cardiovascular disease2 3 on the assumption that the nondiabetic risk factors—smoking, hypertension, obesity, hyperlipidemia, physical inactivity, deprivation, and heavy alcohol consumption—are common to both. Grouping the two related forms of macrovascular disease may, however, mask possible independent relationships between each and its individual risk factors.4
People with diabetes are known to be at increased risk of cerebrovascular events.5 6 The estimated relative risk reported by Sacco7 was 1.5 to 3.0, and results from the Framingham Study showed that the incidence of atherothrombotic cerebral infarcts was 2.5- to 3.5-fold greater in the diabetic population aged 45 to 74 years than in the nondiabetic population.8 Later results from the Framingham Study revised these relative risks to 1.4 for diabetic men and 1.72 for diabetic women after adjustment of their data for other known risk factors.9 Bell4 describes 12 studies showing increased relative risk ranging from 1.4 to 6 for men and 1 (in the fifth decade) to 13 (overall) for women. We have shown that in our resident population there is a crude 12-fold increased probability (95% confidence interval, 10.8 to 12.8) of being admitted to the hospital for a cerebrovascular event in the diabetic population compared with the nondiabetic population.10
In a mortality study, diabetes was said to be responsible for 16% of all stroke deaths in men and for 33% in women.11 A study that examined stroke survival did not find this sex-related relationship; the 30-day survival rate was 89% in men and 79% in women. Survival at 1 year was 79% and 64% for men and women, respectively.12
The objective of this study was to investigate the epidemiology of CVD and specifically its relationship with diabetes. We also attempted to quantify the inpatient costs of care of stroke management in the two populations.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Data and Subjects
The study used a data set consisting of 4 years of routine hospital data from a large population (408 000 people) whose demographic characteristics are known to reflect those of the United Kingdom as a whole.13 These data have undergone a number of processes that are intended to achieve three objectives: (1) to quantify all activity related to people with identified diabetes, (2) to identify all records relating to the same individual with the use of record linkage, and (3) to estimate resource use by the inclusion of a cost weight attributing each record to a DRG.
All hospital contract minimum data sets—a minimum data set required for each hospital admission in the UK NHS—for South Glamorgan residents treated anywhere in Wales were examined for a 4-year period (financial years 1991/1992 to 1994/1995). Selected primary and subsidiary ICD-9 diagnostic codes for the diabetic versus nondiabetic populations, stratified by age and sex, were used. The local prevalence of diabetes is known to be 1.36%.14
Identification of People With Diabetes
The methods of identifying people with diabetes and data
validation have been reported previously.10 Briefly, they
were identified from one of four sources: either (1) a three-digit
ICD-9 diagnostic code (ICD 250) on one of six
diagnostic fields on routine inpatient records, (2) a
diagnosis of diabetes deduced by inclusion on a diabetic outpatient
clinical database, (3) an appointment at an outpatient diabetic clinic
identified from the patient administration system, or (4) inclusion on
a local primary care diabetes database covering 80% of all primary
care practices in the district.
Record Linkage
Record linkage was performed with the use of the Oxford
Record Linkage Study algorithms.15 Data relating to
patient identification on each record were probability matched to
determine which records related to the same individual. A new
patient master index was then constructed with a unique identification
number allocated to all routine records.
Selection of Records Relating to Cerebrovascular
Disease
Records were selected when a three-character ICD-9
diagnostic code for diagnoses related to CVD was
present as a primary or subsidiary diagnosis (ICD-9 codes 430 to
438 inclusive). Each record included data on outcomes (death or
discharge) and length of stay.
Mortality From Stroke: OPCS Data
Data describing all mortality in the district from 1993 to 1995
(inclusive) were supplied by the OPCS. With the use of the same primary
diagnosis codes as noted above, deaths from CVD were identified.
Although there is a recognized underrecording of diabetes on
death certificates in the United Kingdom, these data do allow
comparison with routinely recorded in-hospital mortality from
CVD.
Determination of Inpatient Costs
An estimate of costs was applied to each diabetic and
nondiabetic inpatient admission by assigning a cost weight associated
with a DRG.16 17 These were derived from ICD-9 and OPCS-4
codes and translated by DRG grouper software (version 12) from the
National Case-Mix Office. The sum of cost weights for each DRG for all
admissions was then used to calculate absolute cost. By dividing total
inpatient revenue expenditure by total number of cost weights, we
derived a cash value per unit cost weight.
Determination of Excess Costs
The excess cost incurred by the diabetic state was
calculated for each individual admission with a primary diagnosis
(ICD-9 three-character code) relating to CVD. Excess bed days and costs
were determined by calculating expected values based on admission
rates, lengths of stay, and cost for an age-matched nondiabetic
population and subtracting these from the values observed for the
diabetic population. Bed-days included long-stay occupancy but excluded
non-health service nursing home accommodations.
Calculation of Rates and Relative Risks
The incidence rates presented represent the
number of individuals admitted for a cerebrovascular event per year.
This presents a clearer picture than considering event rates since
local practice is to admit patients with cerebrovascular events into an
acute unit before transfer to a rehabilitation or long-stay bed. Each
event will therefore normally generate two admissions. We accepted that
this definition was limited since it excluded recurrences of
events. We also assumed that the annual incidence was constant for the
4-year period, and we excluded events that occurred in the community
and that did not result in hospitalization.
We derived age-standardized relative risks by applying the rate of individuals admitted in the diabetic population to the nondiabetic population, using community-derived prevalence of diabetes to create the denominator.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Hospital Activity and Descriptive Epidemiology
During this period there were 11 196 admissions with CVD coded as a primary or subsidiary diagnosis. This represented 3.1% of all admissions over the 4-year period. Of these admissions, 7351 had a form of CVD recorded as a primary diagnosis. The 11 196 admissions were generated by 5358 individual patients, of whom 3904 individuals (73%) were admitted with CVD as a primary diagnosis. These two values suggest that there is a crude annual incidence rate of CVD requiring hospital admission (and surviving to admission) of between 2.4 per 1000 (based on only the primary diagnosis) and 3.3 per 1000 based on CVD coded within any diagnostic field. For patients with diabetes, the respective figures are 23.0 per 1000 and 32.8 per 1000.
For both diabetic and nondiabetic patients, the ratio of patients to admissions was just over 2, indicating that an individual cerebrovascular event will usually generate two admissions: the first to a general medicine ward or emergency department and the second to the stroke unit or a long-stay bed.
Fig 1
shows the proportion of patients admitted
with a diagnosis of CVD for patients with and without diabetes. Of a
total of 5358 patients admitted, 771 (14.4%) were patients with
diabetes. At all ages the proportion of patients admitted was higher
for patients with diabetes, but both populations follow a similar
trend, with few differences between the sexes and a strong age-related
effect.
|
Compared with the population as a whole, the age-adjusted
relative risk for a cerebrovascular-related admission for the diabetic
population was 3.70 (95% CI, 3.53 to 3.88; P<.0001) for
men and 4.56 (95% CI, 4.37 to 4.76; P<.0001) for women.
Fig 2 illustrates the relative risk of a patient with
diabetes being admitted with CVD compared with a patient without
diabetes by sex and age group. This shows a range for men of 19.34
(95% CI, 10.46 to 35.77; P<.0001) for the group aged 35 to
44 years to 2.38 (95% CI, 2.01 to 2.83; P<.0001) for those
75 years and older, and a range for women of 26.76 (95% CI, 10.25 to
69.32; P<.0001) for the group aged 25 to 34 years and 3.46
(95% CI, 3.02 to 3.98; P<.0001) for those 75 years and
older.
|
There is little difference in relative risk for patients with and
without diabetes for different ICD-9 groupings except for much lower
values for codes relating to hemorrhage (ICD-9 430 to 432)
(Table 1). This reflects the different etiology of
hemorrhage as opposed to thrombotic CVD and provides useful
control data. Since the absolute numbers involved are small, however,
there is little effect on the epidemiological conclusions.
|
Length of Stay
The lengths of stay for CVD are shown in Table 1
. The mean
length of stay for the late effects of CVD was 133 days. The overall
mean length of stay was less for patients with diabetes compared with
the nondiabetic population (32 days versus 38 days, respectively). This
overall comparison masks the real difference between the two
populations. When only live discharges are considered, the length of
stay for patients with diabetes was 31 days compared with 28 days for
patients without diabetes. Of the 121 admissions (15.7%) resulting in
death for the diabetic population, the average length of stay before
hospital death was 24 days, significantly less than the 58 days for the
915 (27.1%) nondiabetic population (P<.0001). All diabetic
patients died within 90 days, whereas 21 patients without diabetes
survived in the hospital for more than 1 year. The length of stay for
all these patients for administrative reasons was recorded as 365
days, and therefore in reality the mean length of stay will be slightly
greater than indicated. The median length of stay for nonsurvivors was
9 days for patients with diabetes compared with 12 for those patients
without diabetes.
Destination on Discharge
Examination of the likely destination after a patient's discharge
has important implications because some destinations involve transfer
to other hospitals for further treatment, such as rehabilitation.
Fifty-four percent of discharged patients were sent to their usual place of residence after discharge. This does not include any residence funded by the NHS. Seventeen percent of patients were transferred to other NHS hospitals. The other main category is death. Patterns were similar in the diabetic and nondiabetic populations.
Mortality In and Outside the Hospital
Of the 3904 patients with a primary diagnosis of CVD recorded
on routine hospital records, 1036 (26.5%) died in the hospital,
and of these 121 (11.7%) had diabetes. The age-adjusted relative risk
of hospital mortality was slightly less in the diabetic population
(relative risk, 1.13; 95% CI, 1.02 to 1.25; P<.05). The
mean age at death for patients with diabetes was 75 years compared with
76 years for patients without diabetes. According to the OPCS mortality
data, 1207 deaths from stroke were recorded over the 3-year period,
343 (28%) of which occurred outside the hospital.
Comparison of Hospital and OPCS Mortality Data
Using the period of overlap between the routine hospital and
OPCS mortality data sources (January 1, 1993, to March 31, 1995), we
were able to compare the recording of mortality from CVD. From
the hospital data source, there were 590 deaths with CVD recorded
as the primary diagnostic code, and of these 78 (13.2%)
were attributed to patients with diabetes. The mortality database,
however, recorded 678 inpatient deaths from CVD as the primary
cause of death, of which 39 (5.8%) had diabetes recorded as a
subsidiary cause. This represents a possible
underrecording on the hospital data source of 14.5% for CVD
overall but also suggests that of the estimated deaths attributable to
patients with diabetes on the OPCS mortality database, only 43.8% are
recorded as such.
Hospital Costs and Excess Costs
In this locality, with a resident population of 408 000 people,
the estimated cost of acute care for admissions with a primary
diagnosis of CVD was £18 million ($29 million in October 1996) over a
4-year period, at 1994/1995 pay and prices. This is approximately £4.5
million per year, £1.1 million per 100 000 population per year (Table 2), or alternatively £2446 per admission, although this
should be doubled for each individual event since in most cases each
event will generate at least two admissions. Within this estimation,
the most frequent cause of admission was "acute but ill-defined
CVD" (ICD-9 436), which accounted for 68.9% and 60.6% of
admissions and 78% and 59.9% of bed-days for the diabetic and
nondiabetic populations, respectively, as a proportion of each group's
total cerebrovascular activity. This represents a cost of £0.7
million per 100 000 population per year (64% of related
expenditure).
|
Admission of people with CVD who also had diabetes cost 15.5% of the CVD-related expenditure. Moreover, approximately 94% of this expenditure (£2.6 million of £2.75 million over 4 years) would not have been expected had the diabetic population had the same probability of admission as the nondiabetic population.10
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This retrospective study of 4 years of routine data for a UK District Health Authority uses methods that identify all coded CVD and all activity related to the care of patients with identified diabetes. Details of patient encounters with health services outside secondary (hospital) care (that is, by community services and general practice) are not included. We do, however, describe a large number of CVD-related events in the diabetic population compared with the nondiabetic population. It must be emphasized that the costs are strictly secondary care costs of treating the condition. Likewise, the epidemiological findings relate only to those people who survive a cerebrovascular event and are admitted to an acute care NHS hospital.
The patterns of cerebrovascular events show that the diabetic population has an increased risk of events compared with the nondiabetic population. In both populations there is a strong age effect, with little difference in prevalence rates between the sexes.
The long lengths of stay evident in these patients illustrate the greater level of morbidity experienced by those who do not die in the early stages of their hospital admission. People with diabetes have a shorter overall mean length of stay as a result of earlier hospital death than those without diabetes. This possibly reflects the higher incidence of multiple pathology.
Although the epidemiology and the etiology of CVD appear to be described extensively, there are few studies of resource use for this group of patients. Forbes18 reported that stroke patients were responsible for 2% of all discharges and 13% of bed-days, at a cost of approximately £7500 per discharge, which is a 50% increase over our figure of £5000. We could find no reports of previous studies that described the cost of CVD in diabetes.
CVD is costly because of the frequency with which it occurs and
the long length of stay for acute treatment and subsequently for
rehabilitation. We have estimated the cost of CVD to be £1.1 million
per 100 000 population per year in the NHS for those events with a
primary diagnosis described in Table 2 at 1994 pay and prices, almost
15% of which is for those with diabetes. This equates to approximately
3.3% of NHS hospital revenue expenditure for CVD and 0.5% for CVD in
diabetes. If admissions with a subsidiary diagnosis of a
cerebrovascular event are taken as an upper estimate of stroke-related
activity, this would be £1.6 million per 100 000 population per year.
This upper estimate may be artificially high because subsidiary
diagnoses are unlikely to predominate over noncerebrovascular primary
diagnoses in cost assignment.
This study attempts to illustrate the complexity of CVD and its impact on the NHS. We have focused on one of the major risk factors for the condition, diabetes. The prevalence of known diabetes in South Glamorgan is less than 2%, but diabetes-related CVD accounts for at least 13% of CVD activity and 14% of CVD costs. We recommend a comprehensive prospective study to fully determine the economic impact and epidemiology of this important disease group.
|
Selected Abbreviations and Acronyms |
|---|
|
|
Acknowledgments |
|---|
This study was supported by the British Diabetic Association (C. Currie and C. Morgan). We would like to thank Leicester Gill from the Unit of Healthcare Epidemiology at Oxford University for performing the record linkage and Alan Roderick for help with collating data.
Received January 23, 1997; revision received March 21, 1997; accepted March 27, 1997.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
1. Thorvaldsen P, Asplund K, Kuulasmaa K, Rajakangas A-M, Schroll M, for the WHO MONICA Project. Stroke incidence, case fatality, and mortality in the WHO MONICA Project. Stroke. 1995;26:361-367.
2. Salomaa V, Tuomilehto J. Diabetes and macrovascular diseases. In: Williams R, Papoz L, Fuller J, eds. Diabetes in Europe. London, England: John Libbey; 1994:46-55.
3. Kant I. Today's epidemics: circulatory diseases. In: Jacobsen B, Smith A, Whitehead M, eds. The Nation's Health: A Strategy for the 1990s. London, England: King Edward's Hospital Fund for London; 1991:30-44.
4. Bell DSH. Stroke in the diabetic patient. Diabetes Care. 1994;17:213-219.[Abstract]
5. Biller J, Love BB. Diabetes and stroke. Med Clin North Am. 1993:77;95-110.
6.
Mast H, Thompson JLP, Lee S, Mohr JP, Sacco RL.
Hypertension and diabetes mellitus as determinants of multiple lacunar
infarcts. Stroke. 1995;26:30-33.
7. Sacco RL. Risk factors and outcomes for ischemic stroke. Neurology. 1995;45(suppl 1):S10-S14.
8.
Kannel WB, McGee DL. Diabetes and
cardiovascular disease: the Framingham Study.
JAMA.. 1979;241:2035-2038.
9.
Wolf PA, D'Agostino RB, Belanger AJ, Kannel
WB. Probability of stroke: a risk profile from the Framingham
Study. Stroke. 1991;22:312-318.
10. Currie CJ, Williams DRR, Peters JR. Patterns of in and out-patient activity for diabetes: a district survey. Diabet Med.. 1996;13:273-280.[Medline] [Order article via Infotrieve]
11.
Tuomilehto J, Rasetnyte D, Jousilahti, P, Sarti
C, Vartiainen E. Diabetes mellitus as a risk factor for death
from stroke: prospective study of the middle-aged Finnish
population. Stroke. 1996;27:210-215.
12. Brown RD Jr, Whisnant JP, Sicks JD, O'Fallon WM, Wiebers DO. Stroke incidence, prevalence and survival: secular trends in Rochester, Minnesota, through 1989. Stroke. 1996;27:373-380.
13. South Glamorgan Health and Social Care Profile. Cardiff, Wales: South Glamorgan Health Authority; 1993.
14.
Butler C, Peters JR, Stott N. Glycated
haemoglobin and metabolic control of diabetes mellitus:
external versus locally established clinical targets.
BMJ.. 1995;310:748-788.
15. Gill L, Goldacre M, Simmons H, Bettley G, Griffith M. Computerised linking of medical records: methodological guidelines. J Epidemiol Community Health. 1993:47:316-319.
16. Bardsley M, Coles J, Jenkins L, eds. DRGs and Health Care: The Management of Casemix. London, England: King Edward's Hospital Fund for London; 1987.
17.
Söderlund N, Milne R, Gray A, Raftery J.
Differences in hospital casemix, and the relationship between casemix
and hospital costs. J Public Health Med.. 1995;17:25-32.
18. Forbes JF. Cost of stroke. Scott Med J.. 1993;38:S4-S5.[Medline] [Order article via Infotrieve]
This article has been cited by other articles:
 |
|
 |
|
  R. Luengo-Fernandez, A. M. Gray, and P. M. Rothwell Costs of Stroke Using Patient-Level Data: A Critical Review of the Literature Stroke, February 1, 2009; 40(2): e18 - e23. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C J Currie, M Jones, J Goodfellow, P McEwan, C L Morgan, C Emmas, and J R Peters Evaluation of survival and ischaemic and thromboembolic event rates in patients with non-valvar atrial fibrillation in the general population when treated and untreated with warfarin Heart, February 1, 2006; 92(2): 196 - 200. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M Jones, P McEwan, C L. Morgan, J R Peters, J Goodfellow, and C J Currie Evaluation of the pattern of treatment, level of anticoagulation control, and outcome of treatment with warfarin in patients with non-valvar atrial fibrillation: a record linkage study in a large British population Heart, April 1, 2005; 91(4): 472 - 477. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. A. Grover, L. Coupal, H. Zowall, C. M. Alexander, T. W. Weiss, and D. R.J. Gomes How Cost-Effective Is the Treatment of Dyslipidemia in Patients With Diabetes but Without Cardiovascular Disease? Diabetes Care, January 1, 2001; 24(1): 45 - 50. [Abstract] [Full Text]
|
|
|
|
|
|
R. G. Holloway, C. G. Benesch, C. R. Rahilly, and C. E. Courtright A Systematic Review of Cost-Effectiveness Research of Stroke Evaluation and Treatment Stroke, July 1, 1999; 30(7): 1340 - 1349. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||