J CIim

R#&abl Vol.46,No.4, pp. 379-393, 1993 0895-4356/93 $6.00 + 0.00

Rintcd in Great Britain. Ail rights reserved Copyright0 1993PcrgamonPress Ltd

THE DUKE SEVERITY OF ILLNESS CHECKLIST

(DUSOI) FOR MEASUREMENT OF SEVERITY
AND COMORBIDITY
GEORGER. PARKERSON
JR,* W. EUGENEBROADHEAD
and CIIIU-KIT J. TSE
Department of Community and Family Medicine, Box 2914, Duke University Medical Center,
Durham, NC 27710, U.S.A.

(Received in revised fom 30 September 1992)

AI&rat--The Duke Severity of Illness Checklist (DUSOI) was evaluated on 414

primary care adult patients using data collected both by medical providers at the time
of the patient visit and later by a chart auditor. Severity scores for individual diagnoses
were determined by summing the ratings for four non-disease-specific parameters:
symptom level, complications, prognosis without treatment, and expected response to
treatment. Mean diagnosis severity scores (scale O-100) among the 21 most prevalent
diagnoses varied from a low of 13.9 for menopausal syndrome to a high of 43.0 for
sprains and strains. An overall severity score was calculated by combining diagnosis
severity scores and giving highest weights to the most severe diagnoses. Provider-
generated overall severity scores (mean = 43.3) and auditor-generated overall severity
scores (mean = 38.9) were significantly correlated (coefficient of agreement = 0.59,
p < 0.0001). Diagnoses varied in their individual contribution to the overall severity
score, from 8.9% for lipid disorder to 90.0% for sprains and strains. Separate
comorbidity severity scores were calculated to measure the severity of all of each
patient’s health problems except the diagnosis under study. For example, patients with
menopausal syndrome had co-existing health problems which generated a high mean
comorbidity severity score of 43.2, while patients with sprains and strains had a low
mean comorbidity score of 4.7. The DUSOI Checklist can be used in the clinical setting
by both providers and auditors to produce quantitative severity scores (by diagnosis,
overall, and for comorbidity) which are based entirely upon clinical judgment. This
method should be useful in controlling for severity of illness in clinical studies and
indicating the outcome of medical care in terms of reduction in severity of illness
following medical interventions.

Severity of Illness Index Comorbidity Health status

INTRODUCTION from medical records. Kaplan and Feinstein

classified clinical ailments by three grades of
Measurement of severity of illness and co-
severity in their study of the effect of comorbid-
morbidity has become increasingly important
ity on diabetic outcomes [l]. Gonnella et al.
during the past 20 years, as the quality and cost
advanced the staging concept, which defines
of health care have become prime issues for
different levels of severity for specific diseases,
the public and third party payers. Most of
based primarily upon manifestations and com-
the original methodologies for measurement
plications of each disease [2-4]. Horn et al.
of these factors were developed for use in the
produced the Severity of Illness Index, a disease-
inpatient hospital setting with data abstracted
generic method which determines severity
from seven indicators: stage of the principal
*Author for correspondence. diagnosis, its interactions with other diagnoses,
379
380 GEORGE
R. PARKERKIN
JR et al.

patient rate of response to treatment, residual such as patient age and gender, diagnoses, medi-
impairment after therapy, complications, cal or surgical procedures, medications, and
dependency of the patient upon medical care, types of encounter. They are designed to quan-
and extent of non-operating room procedures tify patient population levels of severity which
required [5,6]. can be used for health services research,
More recently, Horn’s group has developed improvement of ambulatory care management,
the Computerized Severity Index, a disease- and determination of appropriate reimburse-
specific tool which uses objective clinical find- ment for medical service [18] . None of these
ings to rate severity for each ICD-9-CM measures incorporates a direct assessment of
classification code [7]. Computerized scoring severity based upon the clinical judgment of
provides a severity level score for each disease the patient’s medical provider.
and a combined score for overall severity from Most of the current methods depend upon
all comorbid conditions [7]. encounter form information and/or medical
Charlson et al. emphasized the importance record audits for their data. While these
of physician clinical judgment in the determi- methods have the advantage of relative ease of
nation of severity [8,9]. Judgments by resident use, without the necessity of burdening the
physicians were used to assess severity based medical provider with data collection, they have
upon how sick they considered each patient. the distinct disadvantage of producing second-
Residents also rated patient functional ability hand information from records which are often
and gave their prognosis for the patient’s j-year a very abbreviated version of what the medical
survival. Follow-up studies showed that predic- provider knows about the patient.
tors of l-year survival after hospitalization were The present study was undertaken to develop
functional ability, severity of illness, extent of a severity of illness measure which would be
comorbid disease and physician prognosis for completed by the medical provider at the time
survival [9]. The importance of comorbidity of the patient visit and which would produce a
also was demonstrated by Greenfield et al., in quantitative assessment of how sick the patient
their studies of the Comorbidity Index, which is, based upon the knowledge and judgment of
measures baseline comorbid severity, acute exa- the provider. A new checklist version of the
cerbations, and patient functional status [lo]. DUSOI was evaluated by comparing severity
In the primary care setting, Barsky et al. scores generated by medical providers with
studied severity by auditing ambulatory medical severity scores from medical record audit.
records to rate the severity of each diagnosis
according to the amount of disease, prognostic METHODS
threat to life, number of organs involved, dis-
ability, complications, and seriousness of treat- The study population consisted of ambu-
ment [1 11.Parkerson et al. developed the Duke latory adult patients visiting the Caswell
Severity of Illness Scale (DUSOI) as a chart Family Medical Center, a rural primary care
audit system in which severity is based upon community health center in the small town
estimates of treatability and prognosis, compli- of Yanceyville, North Carolina. The centre
cations, and symptom level for each diagnosis, provides health care for approximately 3000
and in which the diagnosis severity scores are patients of all ages, about half of whom are
combined into an overall severity score for each female, and whose racial mix is about half black
patient [12]. and half white. Patients present with a full range
Current interest in severity has been focused of health problems, but obstetrical care is not
primarily upon the development of ambulatory provided.
case-mix measures. Examples are the Ambulat- Data were collected during the 8-month
ory Visit Groups (AVGs) by Fetter et al. [13], period ending in April 1991. Consenting
the Ambulatory Severity Index (ASI) by Horn patients aged 1865 years who presented to
[14], the Products of Ambulatory Care (PACs) the clinic for their usual health care, and
by Tenan et al. [15], the Ambulatory Patient who demonstrated adequate literacy by com-
Groups (APGs) by Averill et al. [16], and the pleting a 17-item demographic questionnaire,
Ambulatory Care Groups (ACGs) by Starfield were included in the study. Literacy was im-
et al. [17]. portant because participants were required to
These are classification systems which are complete a multi-item questionnaire for a com-
based upon different combinations of factors, ponent of the study which is reported elsewhere
 Duke Severity of Illness Checklist 381

[ 191. A convenience sample was chosen in an come for the patient in terms of disability or
attempt to provide at least 30 patients in each threat to life during the 6 months following the
of 12 categories, which combined gender, black rating encounter, if the health problem were to
and white race, and three age groups (18-33, go untreated. Prognosis without treatment is
3449, and 50-45 years). considered to be an important indicator of the
During the study period medical care was basic seriousness of the health problem, even
provided by two family physicians, two general when treatment is available and will be given.
internists, and one physician assistant. All of The fourth severity parameter, treatability, indi-
these providers participated in the study by cates the prognosis with treatment in terms of
completing the DUSOI checklist and DUSOI the provider’s perceived need for treatment
analog scale (both described below) just after and the expected response to treatment if it is
seeing patients who had agreed to participate. needed. High severity parameter ratings indicate
The first author of this report, a family phys- the presence of more symptoms, more compli-
ician not associated with the clinic, served as the cations, worse prognosis without treatment, and
principal auditor. He audited all of the medical worse expected response to treatment.
records using the same DUSOI checklist that Although the DUSOI severity parameters are
was used by the providers. Two other non-clinic themselves non-disease-specific, severity ratings
family physicians audited the records for data are made separately for each of the patient’s
which were used only in the intra- and inter- health problems. The first step in completing the
auditor analyses. Audits were performed from checklist is to record each diagnosis or health
photocopies of medical records which had problem. The order in which the problems are
patient identification removed, except for entered .on the form is not important, and any
ID numbers. Portions for audit included the number can be included. All diagnoses or health
problem list, medication list, laboratory reports, problems are listed which are active at the time
and progress notes for the visit at which the of the visit or during the preceding week, includ-
patient entered the study and for the preceding ing those chronic conditions which were not
year. managed on the day of the visit. For example,
The DUSOI is an instrument for measuring a patient with pneumonia might have the
illness severity and comorbidity in ambulatory chronic problems hypertension and obesity,
patients. The DUSOI score reflects each which were not mentioned in the progress note
patient’s “burden of illness” on the day of the on the day of the index visit, but which were
patient visit and during the preceding week, i.e. listed on the problem list and/or elsewhere in the
at a given point in time, for all recorded health progress notes prior to the index visit. The
problems. Initially the DUSOI was described as patient’s medical provider may be able to recall
a chart audit modality [12]. Since then it has this information, but the auditor will have to
been revised into a checklist format that can be obtain it by reviewing the problem list, medi-
used both by clinicians at the time of the patient cation record, and progress notes for the preced-
encounter, and by medical record auditors ing year.
retrospectively (see Appendix A). The non- Health problems are listed on the DUSOI
disease-specific parameters for judging severity checklist in the words of the provider. They may
have been refined since the initial report by include standard diagnostic terms, symptoms,
separating treatability and prognosis into separ- and/or ill-defined health problems. No standard
ate components, so that the current parameters terminology is required and no explicit diagnos-
include symptom level, complications, progno- tic criteria are necessary. Definition and labeling
sis without treatment, and treatability. of health problems is based upon the implicit
The symptom level parameter indicates the judgment of the clinician who sees the patient.
symptomatic state of the patient on the day of When the chart audit approach is used, the
the severity rating and during the preceding auditor accepts the diagnostic labeling made in
week. The complications parameter shows the the record by the provider.
level of complications present during that same For the purpose of data description and
period of time, with a complication defined as a analysis, any standard diagnostic classification
health problem which is secondary to another system can be utilized. In the present study,
health problem, but which does not warrant the health problems identified by the providers
listing as a separate problem in the rater’s were coded by a data technician using the
judgment. Prognosis indicates the expected out- International Classification of Health Problems
382 GEORGE R. PARKERSON
JR et al.

in Primary Care (ICHPPC) [20]. This system A DUSOI comorbidity severity score can
provides 371 rubies for the problems seen most be computed by using the same equation as
commonly in the primary care setting [20]. that for overall severity except that the score
After listing all health problems, the next step of the diagnosis of principal interest is omitted
in completing the DUSOI checklist is to check from the calculation. For example, when a
the one of the five response options which best patient or group of patients is being studied
represents the patient’s clinical status, for each for depression, the diagnosis severity score for
of the four severity parameters for each health depression would be considered separately from
problem listed. There are no specific criteria the comorbidity severity score, which would be
which can be used by the auditor or provider for computed for each patient from the scores of all
estimating the levels of severity for each par- the diagnoses except depression.
ameter. Ratings are based entirely upon the The providers in the present study also com-
implicit clinical judgment of the rater. pleted a DUSOI analog scale which consists of
Although manual scoring is possible, comput- a line 100 mm in length with zero for lowest
erized scoring was used in the present study. overall severity and 100 for highest overall
The DUSOI diagnosis severity score for each severity. (See Appendix B.) The provider placed
diagnosis is computed by summing the four a mark along the line to indicate a rating for the
parameter scores (each of which can range patient’s overall severity of illness during the
from 0 to 4), dividing by 16, and multiplying by preceding week, and the score was determined
100 to produce a score on the scale of 0 for by measuring the distance from the zero end of
lowest to 100 for highest severity. the line to the mark.
The DUSOI overall severity score for each During the study period one subgroup of
patient is computed on a scale of O-100 by patients made return visits to the clinic and
using an equation which gives full weight to the completed the questionnaire packet at both
diagnosis or health problem with the highest visits. Data from the initial visit were used in the
diagnosis severity score, and which gives pro- analyses involving the entire study group, and
gressively diminishing weights to diagnoses with data from both visits were used in the test-retest
lower scores. This method allows entry of an analyses.
infinite number of diagnoses for each patient, Intraclass correlation coefficients (ICC) de-
while drastically limiting increments to the over- rived from a two-way mixed effects analysis of
all severity score by diagnoses which are causing variance model assuming fixed raters’ effects
the least burden of illness. The overall severity [21] were used to measure intra- and inter-rater
equation is as follows: reliability on the chart-audited data and also
to measure the agreement between audited and
DUSOI=DXl +(loo;~l) provider-generated DUSOI scores. Excellent
agreement is indicated by coefficients greater
than 0.75; fair agreement, by coefficients

where
X
(
iDX2+;DX3+-..++“”
>
between 0.40 and 0.75; and poor agreement, by
coefficients less than 0.40 [22].
Spearman rank-order correlation coefficients
were used to demonstrate the associations
DUSOI = Overall severity
between DUSOI scores and sociodemographic
DXl = Highest diagnosis severity score
variables and also between test and retest
DX2 = Second highest diagnosis severity
DUSOI scores. The Student t-test was used
score
to test for differences in continuous variables
DX3 = Third highest diagnosis severity
such as age, and the chi-square was used to test
score
for differences in categorical values.
DXn = Lowest diagnosis severity score
n = Number of diagnoses.
RESULTS
This process generates for each patient a
series of diagnosis severity scores to quantitate Study population
severity of each of their health problems, and Of the 561 patients who were asked to partici-
also an overall score which is a quantitative pate in the study, 534 (95.2%) consented, and
measure of severity resulting from all their 414 (73.8%) were included in the analyses. The
illnesses. 120 consenting patients who were excluded from
 Duke Severity of Illness Checklist 383

the analyses included 50 who were not s&i- also had diabetes; 8 had tobacco abuse; 16 had
ciently literate to complete the questionnaire, a lipid disorder; and the others had a variety
37 who gave incomplete demographic data, of co-existing illnesses. Likewise, only 1 of
28 who were too sick to participate, and 5 the 38 diabetic patients had diabetes alone.
for whom the providers did not complete the Twenty-four of the diabetics also had hyper-
DUSOI checklist. Most patients were evaluated tension; 2 had tobacco abuse; 7 had a lipid
by one of three providers, i.e. 39.6% by one of disorder; etc.
the family physicians, 26.3% by the physician
assistant, and 21.7% by one of the general Severity of illness scores
internists. The remaining 12.4% were seen by The DUSOI diagnosis severity of illness
the other two physicians. scores for the 21 most prevalent health problems
The 414 study patients had a mean age of are shown in Table 2, arranged in descending
40.5 f 13.1 SD years, with 33.6% aged 18-33 rank-order by level of severity. Mean scores
years, 39.1% aged 34-49 years, and 27.3% ranged from a high of 43.0 f 11.8 SD for
aged 50-65 years. Women constituted 58.7% sprains and strains to a low of 13.9 f 10.7 SD
of the group; black patients, 47.1%; and for menopausal syndrome. Also severity scores
white patients, 52.9%. Of the group 56.6% were varied widely within each diagnostic category,
married; 22.8% never married; 9.5% divorced; e.g. ranges of 75.0 for hypertension, 81.2 for
6.6% separated; and 4.6% widowed. diabetes, and 93.8 for alcohol abuse.
Distribution of patients by educational level Each of the diagnosis DUSOI scores was
showed 32.5% who had less than a high school calculated as the mean of the four severity
education, 35.9% who were high school gradu- parameter ratings checked off by the provider.
ates without further training, and 31.6% who The relative contribution of each of these four
had education beyond high school. Fifty-seven parameters to the total diagnosis score is shown
percent had full-time jobs, 9.3% had part-time in Table 3. Symptom level contributed the most
jobs, and 6.6% were looking for jobs; 10.8% toward severity for acute illnesses, e.g. 60.6%
were keeping house; 2.9% were going to school; for acute upper respiratory infection, and least
3.7% were retired; 8.1% were disabled; and for chronic conditions, e.g. 6.7% for hyper-
1.7% were in miscellaneous categories. tension. Complications contributed the most
Data from a subgroup of 54 patients who for chronic conditions, such as alcohol abuse
made return visits to the clinic were used with 14.8%, and the least for acute illnesses,
for test-retest analyses. These patients had a several of which had a zero contribution, e.g.
mean age of 43.8 f 14.2 years, and were 64.8% headache and vaginitis, indicating no compli-
womenand 31.5% blacks. They included 51.8% cations. Ratings for prognosis without treat-
who were married, 72.2% with at least a high ment (expected disability or death) contributed
school education, and 35.9% who were working most toward severity scores for chronic diseases,
full-time. such as ischemic heart disease (42.2%), and least
for acute illness such as vaginitis (2.6%).
Prevalence of health problems Treatability (expected poor response to treat-
Providers reported 803 health problems for ment) was the most important determinant of
the 414 patients, for a mean of 1.94 + 1.15 SD severity for chronic conditions, e.g. lipid dis-
per patient. Only 47.6% of the patients had order (84.5%), and least important for acute
one problem; 26.8% had two; 14.0% had three; conditions, e.g. acute upper respiratory infec-
8.2% had four; 2.4% had five; and 1.0% had tion (26.1%). Overall, for the 803 health prob-
six problems. The five diagnoses with the lems, treatability was the most important
highest prevalence were hypertension (28.0%), contributor to diagnosis severity scores (37.7%),
diabetes mellitus (9.2%), tobacco abuse (9.2%), and complication level was the least important
lipid disorder (7.7%), and acute bronchitis (5.2%).
(7.7%). DUSOI overall severity scores calculated
The distribution of the 21 most prevalent from weighted diagnosis severity scores are
health problems and the extent of their co- shown in Table 4 for patients with each of
morbid relationships are shown in Table 1. the 21 most prevalent health problems. These
For example, hypertension was diagnosed overall severity scores ranged from a high
in 116 patients, but only 17 of these had hyper- of 60.7 + 10.9 SD for depression, to a low of
tension alone. Twenty-four of the hypertensives 36.2 + 14.6 SD for vaginitis, with a mean overall
384 GEORGE
R. PARKERSON
JR et al.

suoguoo pue sas!nq

su!eqs pue su!eIds

atuo~puds pzsnedouab

awas!p Iu@oIoJnau Iaqlc --000

/
uoyaldaa 000000

S!I!U!8t?~ mo-o-00
i
/
*0000000
/

u!ud qX?q MO-) moooooooo

aseas!p lleaq yuays! 3!uo~q3

syqpeoapg

sasuas!p qwmo~s JaylO

1-ooo-mo-00

~N000-0--00

sysnms -/o-00-0-0-000

/
w/o 0 0 0 0 0 -00-000

/
Lla!xuy

lapJos!p p!dg

asnqe 033uqoL

sawera

uO!Sua$Iad&
 Duke Severity of Illness Checklist 385

Table 2. DUSOI diagnosis severity scores for the 21 most prevalent health problems (N = 414 patients with
803 problems)
Health problems N Mean* SDt Minimum Maximum Range
Sprains and strains 8 43.0 11.8 18.8 56.3 37.5
Other neurological diseases 9 41.0 13.3 18.8 56.3 37.5
Headache 9 z: 15.0 18.8 62.5 43.7
Bronchitis, acute 32 9.6 25.0 75.0 50.0
Obesity 29 4O:l 12.7 12.5 68.8 56.3
Low back pain 13 38.0 14.1 18.8 62.5 43.7
Sinusitis, acute or chronic 16 37.5 6.5 25.0 50.0 25.0
Chronic ischemic heart disease 14 37.1 20.0 0 68.8 68.8
Anxiety 22 36.9 19.3 0 75.0 75.0
Other stomach diseases 16 35.5 13.8 6.3 62.5 56.2
Diabetes mellitus 38 35.4 21.1 6.3 87.5 81.2
Alcohol abuse 10 33.8 36.8 0 93.8 93.8
Bruises and contusions 8 32.8 13.3 18.8 56.3 37.5
Depression 9 32.6 10.7 12.5 43.8 31.3
Osteoarthritis 15 31.3 15.1 6.3 62.5 56.2
Upper respiratory infection, acute 21 31.0 13.9 18.8 81.3 62.5
Tobacco abuse 38 30.8 12.6 18.8 75.0 56.2
Hypertension 116 26.8 16.8 75.0 75.0
Vaginitis 10 24.4 7.5 145 25.0 12.5
Lipid disorder 32 14.1 8.2 6:3 37.5 31.2
Menopausal syndrome 9 13.9 10.7 6.3 37.5 31.2
All health problems 803 31.3 17.2 0 93.8 93.8
*Duke Severity of Illness (DUSOI) scores for individual health problems, not including comorbid health
problems. Scale: 0 = lowest severity, 108 = highest severity.
tSD = Standard deviation.

score for all health problems of 43.3 + 18.6 p = 0.003). They were not correlated signifi-
SD. Overall severity scores showed weak cantly with race or marital status.
positive correlations with older age (r = 0.26, Also shown in the table are the weighted
p = 0.0001) and female gender (r = 0.12, DUSOI diagnosis scores used in calculating
p = O.Ol), and weak negative correlations with the overall severity scores, and the percentage
full-time employment (r = -0.20, p = 0.0001) contribution of each diagnosis score to overall
and higher levels of education (r = -0.15, severity. For example, the sickest patients were

Table 3. Percentage contribution to DUSOI diagnosis severity score by the four severity parameter ratings
for the 21 most prevalent health problems (N = 414 patients with 803 problems)
Severity parameters
Prognosis
Symptom Complication without
level level treatment Treatability
Health problems N W) (W W) W)
Sprains and strains 8 43.5 1.9 20.0
Other neurological diseases ; 44.8 8.4 18.6 z
Headache 41.4 0 24.2 3414
Bronchitis, acute 32 47.5 2.5 21.4 28.6
Obesity 29 24.2 12.3 15.1 48.4
Low back pain 13 49.4 2.5 16.5 31.6
Sinusitis, acute or chronic 16 51.0 0 16.7 32.3
Chronic ischemic heart disease 21.8 3.5 42.2 32.5
Anxiety :: 38.4 9.3 19.3 33.0
Other stomach diseases :: 46.2 ;*: 18.6 31.8
Diabetes mellitus 13.1 38.6 38.6
Alcohol abuse 10 25.9 14:8 25.9 33.4
Bruises and contusions 8 52.3 0 21.5 26.2
Depression 9 23.4 2.1 36.2 38.3
Osteoarthritis 15 42.7 2.6 20.0 34.7
Upper respiratory infection, acute 21 60.6 2.8 10.5 26.1
Tobacco abuse 1:: 22.9 9.1 4.9 63.1
Hypertension 6.7 6.0 42.2 45.1
Vaginitis 10 53.8 0 2.6 43.6
Lipid disorder 32 7.1 7.1 84.5
Menopausal syndrome 3:.: 0 14.9 50.0
All health problems 33:5 5.2 23.6 37.7

CE 4614-E
386 GEORGE
R. PARKER~ON
JR ef al.

Table 4. DUSOI overall severity scores for patients with the 21 most prevalent health problems
(N = 414 patients with 803 problems)
Individual health
Overall problem contribution
severity scores* to overall severity
Health problems N Mean SD Weighted score? %$
Depression 9 60.7 10.9 10.9 18.0
Diabetes mellitus 38 56.9 16.8 23.5 41.3
Other neurological diseases 9 56.6 17.4 27. I 47.9
Obesity 29 54.7 15.3 26.2 47.9
Other stomach diseases 16 53.6 16.3 22.1 41.2
Anxiety 22 53.0 17.8 28.9 54.5
Chronic ischemic heart disease 14 52.4 20.1 26.5 50.6
Headache 9 51.9 15.9 26.5 51.1
Alcohol abuse 10 50.8 33.7 26.4 52.0
Osteoarthritis 15 49.8 16.9 16.6 33.3
Bronchitis, acute 32 49.1 13.1 33.3 67.8
Hypertension 116 48.9 18.0 17.2 35.2
Tobacco abuse 38 47.9 16.7 18.8 39.3
Low back pain 13 47.8 10.1 34.1 71.3
Menopausal syndrome 9 46.2 10.7 3.2 6.9
Sprains and strains 8 46.0 6.7 41.4 90.0
Sinusitis, acute or chronic 16 44.2 9.3 34.4 77.8
Lipid disorder 32 44.0 16.6 3.9 8.9
Bruises and contusions 8 43.6 11.8 26.3 60.3
Upper respiratory infection, acute 21 38.7 14.6 26.6 68.7
Vaginitis 10 36.2 14.6 16.3 45.0
*Duke Severity of Illness (DUSOI) overall severity scores, including all problems for each patient.
Scale: 0 = lowest severity, 100 = highest severity.
tweighted severity score of the individual health problem used in calculating the overall severity
score.
$Percentage contribution of the individual health problem weighted score to the overall severity
score.

those whose diagnoses included depression. heart disease (40.9 & 19.7 SD) and obesity
However, none of those patients had depression (39.1 + 21.3 SD) were higher than mean comor-
alone, and most of their overall severity was bidity scores for bruises and contusions
caused by their other health problems. The (21.2 + 21.9 SD) and acute upper respiratory
weighted diagnosis score for depression was infection (16.4 f 18.0 SD). For all 803 health
only 10.9, and contributed only 18.0% of the problems in the study population the mean
overall severity score. This was true even though comorbidity score was 32.0 + 24.3 SD.
the mean unweighted DUSOI diagnosis score The group of 54 patients who were studied
for depression alone was 32.6 + 10.7 SD, as longitudinally had a provider-generated mean
shown in Table 2. That depression did not DUSOI overall severity score of 47.9 & 15.8 SD
receive full weight in the calculation of overall at their initial visit and 47.1 & 16.0 SD at their
severity indicates that the other individual return visit, with a test-retest correlation of 0.65
comorbid diagnoses were rated as being more (p = 0.0001). The time interval was 75.5 f 52.4
severe than depression in these 9 study patients. SD days, with minimum of 7 days and a maxi-
On the other hand, the 38 diabetic patients had mum of 205 days. Overall, this test-retest group
a mean overall severity score of 56.9 + 16.8 SD, had a greater burden of illness than the other
and 41.3% of the overall severity was due to patients in the total study group, as indicated
diabetes. not only by their higher overall severity scores
DUSOI comorbidity scores were calculated, (47.9 compared with 42.7; t = 1.93, p = 0.0536),
and for patients with the 21 most prevalent but also by their higher mean number of health
health problems, mean comorbidity scores problems per patient (2.43 at Time 1 and 2.46
ranged from a high of 54.8 + 12.7 SD for at Time 2, compared with 1.87 for the other
depression, to a low of 4.7 + 13.3 SD for sprains patients at Time 1; t = 2.74, p = 0.008). How-
and strains. Patients with chronic health con- ever, both groups had similar diagnostic pro-
ditions generally had higher comorbidity scores files, i.e. they shared 8 of their 10 most prevalent
than patients with acute illnesses. For example, health problems at the time of the initial office
mean comorbidity scores for chronic ischemic visit.
 Duke Severity of Illness Checklist 387

When the health problems of the test-retest by providers. There was no statistically signifl-
group at Time 1 were compared with those at cant difference (t = 1.61,~ = 0.11) between the
Time 2, it was found that chronic problems, mean number of diagnoses per patient by audit
such as hypertension and diabetes, were preva- (2.07 f 1.26 SD) and the number by providers
lent at both visits, and that acute problems such (1.94 + 1.15 SD). Likewise, the distribution of
as headaches and sinusitis were less likely to health problems per patient was not statistically
be prevalent at the return visit. Review of the different (&i-square = 6.58 with 6 df, p = 0.36).
medical records of this group revealed that For example, the auditor reported 42.0% of
69.8% were seen at the return visit for follow-up patients having only one diagnosis compared
of chronic or recurrent conditions from the first with the provider report of 47.6%, and 30.2%
visit, and that 30.2% were seen for acute prob- of patients with two diagnoses compared with
lems which appeared to be unrelated to the 26.8%.
acute problems on the initial visit. Diagnostic lists generated by the auditor and
the providers were similar. For example, 121
Comparison of provider and auditor scores diagnoses of hypertension were reported by the
Reliability analyses for the data which were auditor, compared with 116 by the providers.
collected by retrospective chart audit using the Of the patients with hypertension, 16 had hyper-
DUSOI checklist showed that the intra-rater tension alone by audit, compared with 17
reliability coefficient for the principal auditor by providers, and 26 had co-existing diabetes
(No. 1) was 0.76 (F = 7.48) at a mean interval instead of the 24 reported by providers. For
of 29.7 & 3.0 days for 67 charts with a mean diabetes, 37 cases were reported by audit instead
DUSOI overall severity score of 39.1 &-17.0 SD of 38, two of which had diabetes alone instead
at Time 1 and 42.2 & 16.8 SD at Time 2. Intra- of the one reported by providers.
rater reliability for auditor No. 2 was 0.67 Assessment of the agreement between
(I; = 5.05) at 31.7 f 4.2 days for 59 records. auditor and provider DUSOI scores showed a
Inter-rater reliability between auditor No. 1 coefficient of agreement of 0.59 (p < 0.0001)
and auditor No. 2 was 0.47 (F = 2.74) for 59 between the audited overall severity scores
records; between, No. 1 and No. 3 was 0.58 (mean = 38.9 f 17.5 SD) and the provider over-
(F = 3.72) for 58 records; and between No. 2 all severity scores (mean = 43.3 f 18.6 SD).
and No. 3 was 0.63 (F = 4.44) for 58 records. Coefficients of agreement between audited
The inter-rater reliability among the three scores and individual provider scores for the
auditors was 0.56 (F = 6.56) for 58 records. The three providers who evaluated 87.6% of the
p-value for all of these reliability coefficients patients were 0.67, 0.55, and 0.53 (p <O.OOOl
was 0.000 1. The mean time required for auditing for all coefficients). Coefficients were calculated
all 414 records by the principal auditor was also by diagnosis for those five health problems
2.2 f 1.6 SD minutes per medical record, with listed by both the auditor and providers at a
a minimum of 1 minute or less and a maximum frequency of 20 or more. As shown in Table 5,
of 14 minutes. coefficients of agreement for diagnosis severity
Audit checklist data compared with provider- scores were statistically significant except for
generated checklist data for the 414 study lipid disorder, where the coefficient was
patients showed 859 health problems recorded only 0.17. Others ranged from the very weak
by the auditor, compared with the 803 reported coefficient of 0.19 for hypertension to the mod-

Table 5. Agreement of provider and auditor-generated DUSOI diagnosis and overall severity scores (N = 414
patients with 803 problems)
DUSOI diagnosis scores$ DUSOI overall severity scores$
Health problems? N Provider Auditor ICC5 Provider Auditor ICC5
Hypertension 106 27.1 f 16.7 14.4 f 7.2 0.19**** 48.0 f 18.1 37.5 f 17.7 0.52****
Diabetes mellitus 35 35.0 f 21.9 25.2 f 18.3 0.70**** 57.7 f 17.1 46.5 f 16.7 0.61****
Tobacco abuse 25 31.3 + 14.0 20.3 f 8.3 0.47**** 45.9 z 18.0 40.0 ; 16.0
Lipid disorder 24 15.1 f 7.8 14.6 f 7.1 0.17**** 46.7 f 15.0 44.0 & 15.5
Bronchitis, acute 24 39.8 f 9.9 38.0 f 7.1 0.35. 47.8 f 13.8 46.5 f 10.7 0.37’
tHealth problems which were listed both by the providers and the auditor, and which had a frequency of 220.
Patients with multiple diagnosis are included in the counts for each of their diagnoses.
$Mean f standard deviation.
$ICC = Intraclass correlation coefficients between provider and auditor scores.
‘p < 0.05; **p < 0.01; ??
**p < 0.001; ?? ***p < 0.0001.
388 GEORGE
R. PARKERS~N
JR et al.

erately strong coefficient of 0.70 for diabetes. patient health problems which are rated for
Coefficients for overall severity scores varied severity. The result of this attempt to quantify
from 0.37 for patients with bronchitis to 0.75 qualitative data is a severity instrument which
for tobacco abuse, and all were statistically attains only a modest level of reliability by
significant. psychometric standards. However, while most
Test-retest analyses for audited DUSOI of the coefficients of agreement for intra-rater,
overall severity scores showed a Spearman inter-rater, and provider-auditor severity scores
rank-order correlation coefficient of 0.59 (p = do not fall into the “excellent” agreement
0.0001). Comparisons between audited scores category of greater than 0.75 suggested by
and providers’ scores were made to show the Fleiss [22], they are within the “fair” agreement
effect of having an auditor generate a score on category of 0.40.75. Given the complexity
one occasion and the provider at a different of human health problems and their effect
time, a strategy which might be useful in certain on people, the many variations in diagnostic
longitudinal clinical studies. For example, labeling, and the unknowns of prognosis in
the provider might rate severity at the initial terms of natural history of illness and expected
visit and an auditor, at the follow-up visit. response to therapy, the degree of reliability
The correlation was 0.44 (p = 0.001) between attained by the DUSOI in the present study is
audited DUSOI at Time 1 and provider DUSOI noteworthy.
at Time 2, and the correlation was 0.42 There is no question that reliability can be
(p = 0.002) between provider DUSOI at Time 1 improved by refining the DUSOI with specific
and audited DUSOI at Time 2. diagnostic and severity criteria, and perhaps
Also, comparisons were made between the this should be done for studies of a limited
auditor and providers with regard to their use number of health problems. In the meanwhile
of the four severity parameters for generating this simple instrument may be a reasonable tool
severity scores. The distribution of the four for measuring severity from the perspective of
severity parameter ratings was similar between the clinician who is active in the practice setting.
auditor and providers, in that the symptom level The present study has demonstrated that the
ratings constituted 30.7% of the individual DUSOI Checklist can be used in the primary
diagnosis DUSOI severity scores by audit, care office both by clinical providers at the
compared with 33.5% by providers. Compari- time of the patient visit, and by auditors retro-
son of auditor with provider use of the other spectively, to produce measures of severity for
parameters was 6.4 vs 5.2% for complication both individual and aggregated health prob-
level; 14.7 vs 23.6% for prognosis; and 48.2 vs lems. All of a person’s health problems can be
37.7% for treatability. included in the process and contribute to the
Patient morbidity also was assessed by the diagnosis, comorbidity, and overall severity
providers using an analog scale to indicate scores, whether or not those problems have been
overall severity of illness. The analog overall defined as specific diseases.
severity scores for the 414 patients (mean: Until medical knowledge advances to the
37.0 + 24.2) showed a coefficient of agreement point at which detailed algorithms based on
of 0.61 (p < 0.0001) with the DUSOI provider sound scientific data have been developed for
overall severity scores (mean: 43.3 f 18.6), the course and management of most health
and a coefficient of 0.42 (p < 0.0001) with problems, medicine will remain heavily depen-
the DUSOI audit checklist scores (mean: dent on implicit clinical judgment to determine
38.9 + 17.5). severity of illness. This dependence upon
clinical judgment is important, particularly
DISCUSSION when severity is being assessed for individual
patients to measure severity as an outcome of
This study has attempted to quantify implicit their medical care. Perhaps clinical judgment is
judgments of medical providers in assessing the less essential when severity is assessed as an
severity of illness of their patients. Although indicator of the intensity of medical care for
their judgments were focused on four explicit determination of health policy or reimburse-
non-disease-specific parameters of severity, ment for clinical services, as in the ambulatory
there were no explicit criteria for clinical ratings case-mix measures.
of severity within each parameter, and there Determination of “how sick the patient is”
were no explicit criteria for diagnosing the is one of the most important decisions for the
 Duke Severity of Illness Checklist 389

clinician in providing health care for the patient. heart disease (21.8%). Conversely, prognosis
This is a judgment that often requires an enor- ratings contributed more to the severity score
mous amount of data, clinical experience, for chronic ischemic heart disease (42.2%) than
basic knowledge of medicine, and information for sprains and strains (20.0%). The result for
on how the particular patient behaves in patients at the time of the office encounter was
response to his or her health problems. Is the that these two very different health problems
patient symptomatic, and if so how severe are resulted in a high and almost equal burden of
the symptoms? Have complications developed, illness. The DUSOI is designed to quantify the
and if so, of what magnitude? Is the condition total burden of illness for the patient at one
life-threatening and/or disabling, and if so, to point in time. These scores are consistent with
what extent? Is treatment indicated, and if so, that approach.
what is the expected response to treatment? Also, this study demonstrates the clinical
These basic questions, which must be answered phenomenon of wide variation of severity
in the clinical assessment of severity, provide a within diagnostic categories. Diagnostic labels
solid basis for conceptualization of the four themselves give only partial information regard-
severity parameters of the DUSOI. ing severity. For example, DUSOI diagnosis
The severity parameters for symptoms, prog- severity scores ranged from a minimum of 6.3 to
nosis, and treatability were used frequently by a maximum of 87.5 units for diabetic patients
the providers in this study to determine severity. and from 0 to 75.0 units for hypertensives. This
On the other hand, they used the complication variability is a reality which clinicians must
parameter for determination of only 5.2% of consider in patient management everyday, and
the total severity of diagnosis severity scores. which has been important in the development of
This may reflect the DUSOI characteristic of disease-specific severity measures such as those
limiting evaluation to a l-week period of time, using disease staging [24].
which might exclude complications which arose Wide variations in the contribution of a par-
in the past few months. Also, it may result from ticular health problem to overall severity also
this study population’s characteristic of gener- are demonstrated in this study. For example,
ally low prevalence of severe illnesses. Unless diabetes contributed only 41.3% of overall
studying the DUSOI in patients with severe severity for the 38 diabetics in this study group
illness also shows rare use of the complications when measured by the DUSOI formula, which
parameter, this low utilization of complications gives full weight to the diagnosis with the
does not provide sufficient grounds for remov- highest diagnosis severity score, and progress-
ing this parameter from the DUSOI, primarily ively diminishing weights to the less severe
because of the known clinical importance of diagnoses. The DUSOI approach of quantitat-
complications in very sick patients. In patient ive combination of diagnosis severity scores by
populations with a higher case mix of severe weights makes clinical sense in that the patient’s
illness the complication parameter has been overall burden of illness at a given point in time
shown to be an important parameter for instru- is usually predominated by the health problem
ments which have been developed in those of greatest severity at that time. For example,
settings [ 1-1. the severity of an acute injury, such as multiple
The data of this study indicate that acute trauma from an automobile accident, can over-
illnesses which may respond well to therapy can shadow the severity of a well-controlled serious
generate diagnosis severity scores equal to or chronic health problem, such as diabetes, even
higher than some of the chronic illnesses which when severity parameters such as prognosis are
are usually considered more serious. This is considered. As illustrated with data from the
explained by the higher contribution of symp- present study, the diagnosis severity scores
tom level parameter ratings than prognosis for sprains and strains contributed 90.0% of
ratings for the acute illnesses at the time the mean overall severity score of 46.0 for
of the patient’s office visit. For example, in patients with sprains and strains, in contrast to
Table 2 the diagnosis severity score for sprains a contribution of 26% by the diagnosis severity
and strains (43.0) is actually higher than that for scores for chronic ischemic heart disease to its
chronic ischemic heart disease (37.1). As seen patients’ mean overall score of 52.4, as shown
in Table 3, symptom level ratings contributed in Table 4.
more to the diagnosis severity score for sprains Both individual diagnosis severity scores
and strains (43.5%) than for chronic ischemic and comorbidity severity scores are important,
390 GEORGE
R. PARKERKIN
JR et al.

especially when a particular illness is being essentially the same at both visits, with a fairly
studied, and the comorbidity of other illnesses is high correlation coefficient of 0.65 (p c 0.0001).
being controlled for. For example, Coulehan Although most of the return visits were for
et al. [23] and Broadhead et al. [24] used audited chronic health problems which may change little
DUSOI scores in separate studies of depression in severity short-term, one might expect that the
to control for the severity of their patients’ severity would decrease significantly for many
illnesses other than depression. They calculated illnesses several weeks after an initial medical
their comorbidity severity score by the same visit in response to treatment, natural history of
method described in this paper, i.e. by excluding illness, and/or the phenomenon of mathematical
the diagnosis severity score for depression from regression toward the mean. A much larger
the scoring equation to obtain a depression-free group of patients would be needed to study
comorbidity score. sensitivity of the DUSOI to change in severity
The comparative data between severity over time, because of the wide variety of
scores generated by patient providers and the health problems in this type of population,
medical record auditor in the present study the wide range of patients’ severity scores, the
gave support for the validity of the DUSOI. diverse reasons for their return visits, and
Provider and auditor lists of health problems the wide range of time between their initial
were very much alike, and the coefficients of and follow-up encounters. All of these factors
agreement between provider and auditor overall decrease the precision of the present follow-up
severity scores were fairly high and statistically data, and increase the necessity for a larger
significant (ICC = 0.59, p < 0.0001). However, sample size.
in some instances there were considerable The present study is very limited in its ability
differences between provider and auditor diag- to validate the DUSOI. There is no gold stan-
nosis severity scores, such as that between dard severity instrument with which to compare
the provider diagnosis severity scores for the measure, and the collection of objective
hypertension (mean = 27.1 +_ 16.7 SD) and the pathologic and physiologic data reflecting the
auditor scores (mean = 14.4 + 7.2 SD), shown severity of multiple illnesses was not practical
in Table 5. Perhaps, part of this difference can as part of the study. Further research is needed
be attributed to inadequacies in traditional with larger numbers of patients having each
medical record keeping, where symptomatology health problem, more objective severity data,
and complications are recorded more routinely use of same-day ratings by more than one
than prognosis. While the provider and the provider, and long-term follow-up to determine
patient were literally face-to-face at the time of validity of the DUSOI in terms of the poss-
severity assessment, the auditor had to glean ible predictive effect of its scores upon future
information from the provider’s progress notes severity of illness.
retrospectively. The most valid assessment in Also, more research is needed on the validity
this situation may be that of the provider, rather of the DUSOI analog scale, the scores of which
than the auditor, because the provider may have showed a fairly high and statistically significant
much more information about the patient agreement (ICC = 0.61,~ < 0.0001) with overall
encounter than was recorded in the medical severity scores calculated from individual
record. There are serious implications here for diagnosis severity scores. While the analog
medical record keeping, particularly if records method provides a quick assessment of total
are to be used to measure severity of illness as burden of illness, it lacks the advantage of
an outcome of medical care. Records would be providing separate severity scores for each of
improved if providers included in their routine the patient’s health problems. Even so, in
progress notes their patients’ prognosis and the present study the DUSOI analog scores
expected response to treatment, in addition to provided a form of internal validation for the
symptom and complication levels. DUSOI overall severity scores in that they
The test-retest analyses in this study are required a somewhat different cognitive process
interesting because longitudinal data are very by the clinician raters.
important in studying the sensitivity of a The DUSOI provides a methodology with
measure to changes in severity over time. In the fairly high reliability and with sufficient clinical
group of 54 patients who consented to partici- face validity to justify further investigation of its
pate in this study at the time of their initial use in medical outcomes assessment. Further
and follow-up visits, the mean scores were research is needed to support its reliability and
 Duke Severity of Illness Checklist 391

validity and to evaluate its ability to measure lo* Greenfield S, Aronow HU, Elashoff RM ef ol. Flaws
in mortality data. JAMA 1988; 260: 2253-2255.
changes in severity over time. The DUSOI 11. Barsky AJ, Wyshak G, Klerman GL. Medical and
should prove useful as a method of controlling psychiatric determinants of outpatient medical utilii-
for severity in studies of functional health and 12 ation. Med Care 1986: 24: 548-560.
Parkerson GR Jr, Michener JL, Wu LR et al. Associ-
quality of life, controlling for comorbidity in * ations among family support, family stress, and per-
studies of specific illnesses, and indicating the sonal functional health status. J Clln Eu&mlol 1989:
outcome of medical care in terms of whether or 42: 217-229.
Fetter RB, Averill R, Lichtenstein JL et al. Ambulat-
not severity of illness has been reduced by 13’ ory Visit Groups: A framework for measuring pro-
medical intervention. ductivity in ambulatory care. He&h Ben Res 1984; 19:
415-437.
Acknowledgements~ompletion of the DUSOI Check- 14* Horn SD, Buckle JM, Carver CM. Ambulatory Sever-
lists at the time of patient visits was performed by W. E. ity Index: Development of an ambulatory case mix
Broadhead, M.D., Ph.D., James W. R. Harding, III, system. J Ambulatory Care Manage 1988; 11: 53-62.
M.D., M.P.H., Janet Jexsik, P.A.-C., Todd Shapley-Quinn, 15. Tenan H, Fillmore H, Caress B et al. PACs: Classify-
M.D., and Bret C. Williams, M.D., M.P.H. Chart audits for ing ambulatory care patients and services for clinical
reliability analyses were performed by Anthony Geraci, financial management. J Ambulatory Care Manage
M.D. and Jonathan Sheline, M.D., M.S. Funding was 1988; 11: 36-53.
provided by Glaxo, Inc., Research Triangle Park, NC, and 16. Averill R, Goldfield N, McGuire T et al. Design
the Department of Community and Family Medicine. Duke and evaluation of a prospective payment system for
University Medical Center. ambulatorv care. MCFA Contract 17-C-99 369/l-021.
Wallingfo;d, CX‘ 3-M Health Information Systems;
1990.
REFERENCES 17. Startield B, Weiner J, Mumford L et al. Ambulatory
Care Groups: A categorization of diagnoses for
1. Kaplan MH, Feinstein AR. The importance of classi- research and management. Health Serv Res 1991; 25:
fying initial co-morbidity in evaluating the outcome of 990-1015.
diabetes mellitus. J Chron Dla 1974; 27: 387404. 18. Weiner JP. Ambulatory case-mix methodologies:
2. Gonnella JS, Goran MJ. Quality of patient care-A Application to Primary Care Research. Ceof Proc
measurement of change: the staging concept. Med Prlmuy Care Reaearehr Theory and Methods. U.S.
care 1975; 13: 467473. Department of Health and Human Services, Public
3. Gonnella JS, Louis DZ, McCord JJ. The staging Health Service, Agency for Health Care Policy and
concept--an approach to the assessment of outcome of Research; September 1991.
ambulatorv care. Med Care 1976: 14: 13-21. 19. Parkerson GR Jr, Broadhead WE, Tse C-KJ. Quality
4. Gonnella JS, Hornbrook MC, Louis DZ. Staging of of life and functional health of primary care patients.
disease, a case-mix measurement. JAMA 1984; 251: J cull Epldssnlol 1992; 45: 1303-1313.
637644. 20. Intematlonal CIamlBeatlou of Health Problems ia Prl-
5. Horn SD. Measuring severity of illness: Comparisons mary Care. Chicago: American Hospital Association;
across institutions. Am J Public Health 1983; 73: 1975.
25-31. 21. Shrout PE, Fleiss JL. Intraclass correlation: uses
6. Horn SD, Buckley G, Sharkey PD et al. Interhospital in assessing rater reliability. Psycho1 Bull 1979; 86:
dilferences in severity of illness. N Engl J Med 1985; 420428.
313: 20-24. 22. Fleiss JL. Statisticd Methods for Rates and Pro-
I. Horn SD, Sharkey PD, Buckle JM ef al. The relation- portiom, 2nd edn. New York: Wiley; 1981.
ship between severity of illness and hospital length of 23. Coulehan JL, Schulberg HC, Block MR et al. Medi-
stay and mortality. Med Care 1991; 29: 305-317. cal comorbidity of major depressive disorder in a
8. Charlson ME, Sax FL, MacKenzie CR et al. Assessing nrimarv medical practice. Arch Intern Med 1990; 150:
illness severity: Does clinical judgement work? J Chron 2363-2367. -
Dls 1986; 39: 439452. 24. Broadhead WE, ClappChanning NE, Finch JN
9. Pompei P, Charlson ME, Douglas RG Jr. Clinical et al. Effects of medical illness and somatic svmntoms
assessments as predictors of one year survival after on treatment of depression in a family medicine
hospitalization: Implications for prognostic stratifica- residency practice. Gen Heap Psycblatry 1989; 11:
tion. J Clln Epldemlol 1988; 41: 275-284. 194-200.

(Appendices overleaf)
392 GEORGE
It. PARKERSON
JR et al.

APPENDIX A

Duke Severity of Illness (DUSOI) Checklist’ ID Number:

Provider: Date of Encounter:

If Audit, Date: Auditor: Patient’s Name:

Diagnosis or Health Problem:

None Questionable Mild Mederate Major

I. Symptoms (past week): Cl6 01 02 03 04

2. Complications (past week): Cl6 01 02 cl3 04

Disability
3. Prognosis (next 6 months, None Mild Moderate Major Threat to Life

without treatment): 06 Cl1 cl2 cl3 04

Need for Treatment Expected Response to Treatment

No Questionable If Yes - Good Questionable Poor

I

4. Treatability: 00 01 02 03 04

Diagnosis or Health Problem:

None Questionable Mild Moderate Major

I. Symptoms (past week): Cl6 01 02 03 04

2. Complications (past week): Cl6 01 02 ? ?3 ? ?4

Disability
3. Prognosis (next 6 months, None Mild Moderate Major Threat to Life

without treatment): Cl6 01 02 cl3 04

Need for Treatment Expected Response to Treatment

No Questionable If Yes - Good Questionable Poor

I

4. Treatability: cl0 01 Cl2 03 04

Diagnosis or Health Problem: _

None Questionable Mild Moderate Major

1. Symptoms (past week): 06 01 02 03 04

2. Complications (past week): 06 01 cl2 03 cl4

Disability
3. Prognosis (next 6 months, None Mild Moderate Major Threat to Life

without treatment): 176 01 cl2 03 cl4

Need for Treatment Expected Response to Treatment

-

No Questionable If Yes - Good Questionable Poor

III

4. Treatability: cl0 01 02 03 04

USE ADDITIONAL SHEETS FOR OTHER DIAGNOSES

*Copyright 1990 Department of Community and Family Medicine, Duke University Medical Center, Durham, NC, U.S.A.
 Duke Severity of Illness Checklist 393

APPENDIX B

Patient’s Name: ID Number:

Provider: Date of Encounter:

Duke Severity of Illness Analog Scale (DUSOI-A)”

(Patient’s overall severity of illness as assessed by the provider)

Instructions:

Please mark with an X the appropriate place along the line below to indicate how you would rate this patient’s overall
severity of illness during the pest week.

Lowest severity applies to someone whose total set of diagnoses results in the fewest symptoms and complications,
the least disability and threat to life, the least need for treatment, and the best expected response to treatment if needed.

Highest severity applies to someone whose total set of diagnoses results in the most symptoms and complications,
the most disability and greatest threat to life, the most need for treatment, and the worst expected response to
treatment.

LOWEST HIGHEST
SEVERITY SEVERITY
(Please mark one X along the line)

How confident are you that your rating of this patient’s severity of illness is accurate? Please circle the appropriate
category.

Not at all Not Very Very Absolutely

Confident Confident Confident Confident

0 1 2 3

*Copyright 1990 Department of Community and Family Medicine, Duke University Medical Center, Durham, NC, U.S.A.