Download ADEPT prognosis feedback

5. Exercise
Survival time and congestive heart failure
A 69 year old woman, who is a retired high school teacher, has attended an outpatient
appointment to review her test results. A month ago she presented with symptoms and signs
of congestive heart failure. She has had long-standing essential hypertension, but had been
otherwise healthy until now.
An SHO reviews the test results, the patient’s medication use and the patient’s general wellbeing. During the appointment, she asked, "Heart failure sounds serious - is it? What do I
have to look forward to?" The SHO decides to find out the average survival time for a patient
with heart failure and comes into the library for a literature search.
Source: Centre for Evidence Based Medicine
Four particular factors are central to a prognosis question – the gender and age of the
patient, the severity or stage of the target disease or condition and the existence of other
diseases or conditions (comorbidities). Three of these elements are significant to this
scenario – the age of the patient, the fact that she is a woman and the co-existence of
hypertension – all of these factors should feature in the focused question and hence the
search strategy. As for the stage of the disease – we have little to go on – it has only
recently been diagnosed but this is not uncommon in prognosis questions. We are looking
here at the course of untreated disease. Although the prognosis filter covers a wide range
of prognosis-related issues we are specifically asked to identify the “average survival
time” in this instance.
Break down the above enquiry into the following components
PATIENT/
POPULATION/
CONDITION
INTERVENTION
OUTCOME
COMPARISON
Any (3)
Risk of death
None (2)
Survival Time (2)
No intervention or
treatment (2)
Hypertension (2)
Drug therapy
Mortality (3)
Female age 69
Medication
Survival/death
Congestive Heart
Failure (2)
Treatment for
hypertension
None
69 year old woman
Survival (1)
56yr old female [??]
Quality of Life (2)
No
treatment/treatment
Aged/Female
Survival rates
Intervention
Congestive heart
failure, plus essential
hypertension
Aged Woman with
Congestive Heart
Failure and
Hypertension
or
Non-intervention
For some reason one of your responses recorded this woman as a 56 year old female.
Unfortunately this confirms all my irrational prejudices as a 33 year old (!) male about a
certain predilection for automatically subtracting over a decade from one’s age. The
medical records, however, never lie!
Now try to match the above components to the corresponding MeSH terminology
PATIENT/
POPULATION/
CONDITION
INTERVENTION
OUTCOME
COMPARISON
? Could have had the
treatment that patient
was receiving for her
hypertension?
Mortality (3)
No intervention
Survival Analysis (3)
No intervention or
treatment
Any
(covered by the
filter)
Intervention
exp Hypertension/
Incidence
Non-intervention
Hypertension (3)
Cohort studies
Aged ("all aged <65
and over>") (2)
Survival rate,
survival
Female (2)
Any
-mortality (sh)
Quality of Life
exp Heart Failure,
Congestive/
Heart failure,
Congestive (4)
Drug therapy
Prognosis (3)
or
Female and aging
<65 to 75 years>
Hypertension
While all of the above outcomes are appropriate to a prognosis question only those
specific to survival are in fact pertinent to this particular scenario. While doing this same
exercise with our group of Trent librarians only last week we also encountered a fairly
useful MeSH term “Disease progression” that doesn’t seem to have figured prominently
in most of the prognosis filters. This is defined by the MeSH browser as:
Disease Progression
The worsening of a disease over time. This concept is most often used for chronic
and incurable diseases where the stage of the disease is an important determinant
of therapy and prognosis.
Year introduced: 1995
By the way (as a side point) do you all know about the animated tutorials on MeSH now
available via the MeSH Browser?. They cover:



Searching with the MeSH Database
Combining MeSH Terms
Applying Subheadings and other features of the MeSH Database
Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=mesh
With broad subjects, you may need to use subheadings to narrow the search to a
manageable number of hits. For narrower subjects, you may find that subheadings
limit your search too much.
Which subheadings (if any) are you going to use with the condition “HEART
FAILURE, CONGESTIVE”?
Participant comments:
Complications, Diagnosis, Drug therapy, Etiology, Mortality and Physiopathology
I thought of using mortality – but this is covered with the Ovid filter. (diagnosis had not
sprung to mind until I read the next page) – this also seems to be covered with prognosis etc
in filter.
My initial search gave too many hits, so I decided to put Hypertension in too. Although final
scenario does not mention hypertension, this is mentioned with relation to the patient.
Although I also tried survival analysis, I realised that the filter covered this too.
Tried also limiting to English/sex/age – but did not make much difference.
Mortality
None because Mortality appears to be the only strictly relevant subheading and is not
sufficient on its own to identify all prognostic factors .
What about diagnosis? No – not sufficiently specific to prognosis.
While mortality is the obvious choice for a subheading the fact that this concept (albeit in
a half-full/half empty way) is already captured in the concept of “survival analysis”
means that it probably will not be required in this instance. Notwithstanding the comment
above on the lack of specificity of the subheading “diagnosis” you will recall that in the
module it describes how indexers are instructed to place materials on prognosis under the
diagnosis subheading. Another parallel strategy for the search is to use the study design
feature embodied in the one line filter “cohort-studies” – an optimal study will follow a
group of patients with hypertension and compare it with those without hypertension and
compare them for development of heart failure and subsequent mortality.
We reproduce below our own strategy along with those from some of the course
participants. Of course our strategy is not a “gold standard” merely an indicative strategy
for comparison with your own!
#
Search History
Results Display
1
congestive heart failure.mp. or exp Heart Failure, Congestive/
59245
2
hypertension.mp. or exp HYPERTENSION/
217859
3
1 and 2
4
exp SURVIVAL ANALYSIS/ or exp SURVIVAL/ or
survival.mp.
5
3 and 4
445
6
limit 5 to (female and (middle aged <45 plus years> or aging
<65 to 79 years> or "all aged <65 and over>" or "aged <80
and over>"))
211
7
exp *Heart Failure, Congestive/
35301
8
exp *HYPERTENSION/
120681
9
7 and 8 and 4
7022
275290
10 9 and 6
29
10
Using this strategy we retrieve the following relevant article:
Unique Identifier
8622246
Authors
Levy D. Larson MG. Vasan RS. Kannel WB. Ho KK.
Institution
Framingham (Mass) Heart Study, Massachusetts 01701, USA.
Title
The progression from hypertension to congestive heart failure.[see comment].
Comments
Comment in: JAMA. 1996 May 22-29;275(20):1604-6; PMID: 8622253
Source
JAMA. 275(20):1557-62, 1996 May 22-29.
Abstract
OBJECTIVES.- To study the relative and population-attributable risks of hypertension for the
development of congestive heart failure (CHF), to assess the time course of progression from
hypertension to CHF, and to identify risk factors that contribute to the development of overt heart failure
in hypertensive subjects. DESIGN.- Inception cohort study. SETTING.- General community.
PARTICIPANTS.- Original Framingham Heart Study and Framingham Offspring Study participants aged
40 to 89 years and free of CHF. To reflect more contemporary experience, the starting point of this study
was January 1, 1970. EXPOSURE MEASURES.- Hypertension (blood pressure of at least 140 mm Hg
systolic or 90 mm Hg diastolic or current use of medications for treatment of high blood pressure) and other
potential CHF risk factors were assessed at periodic clinic examinations. OUTCOME MEASURE.- The
development of CHF. RESULTS.- A total of 5143 eligible subjects contributed 72422 person-years of
observation. During up to 20.1 years of follow-up (mean, 14.1 years), there were 392 new cases of heart
failure; in 91% (357/392), hypertension antedated the development of heart failure. Adjusting for age
and heart failure risk factors in proportional hazards regression models, the hazard for developing heart
failure in hypertensive compared with normotensive subjects was about 2-fold in men and 3-fold in
women. Multivariable analyses revealed that hypertension had a high population-attributable risk for CHF,
accounting for 39% of cases in men and 59% in women. Among hypertensive subjects, myocardial
infarction, diabetes, left ventricular hypertrophy, and valvular heart disease were predictive of increased
risk for CHF in both sexes. Survival following the onset of hypertensive CHF was bleak; only 24% of men
and 31% of women survived 5 years. CONCLUSIONS.- Hypertension was the most common risk factor
for CHF, and it contributed a large proportion of heart failure cases in this population-based sample.
Preventive strategies directed toward earlier and more aggressive blood pressure control are likely to offer
the greatest promise for reducing the incidence of CHF and its associated mortality.
This tells us that only 31% of women survive 5 years with hypertensive congestive heart
failure. This doesn’t quite answer the clinician’s question which was on average survival
time. However this article is definitely relevant and if your requester was to examine the
full text you will see that the average survival time is 2.48 years.
Participant Strategies
Here are some of the strategies that you have used.
Please paste your suggested strategy here:
#
Search History
Results
1
exp Heart Failure, Congestive/
17971
2
exp HYPERTENSION/
43238
3
1 and 2
1139
4
incidence/ or exp mortality/ or mo.fs.
169277
5
exp prognosis/ or prognos$.tw.
264443
6
4 or 5
382544
7
exp cohort studies/ or cohort.tw.
242723
8
exp survival analysis/
39616
9
7 or 8
269217
10
6 and 9
114236
11
3 and 10
78
#
Search History
Results
1
exp *heart failure, congestive/
34428
2
limit 1 to (human and english language)
22062
3
exp AGED/
1346637
4
exp female/
4012571
5
2 and 3 and 4
6
exp *hypertension/
7
5 and 6
8
essential hypertension.mp.
9
5 and 8
10
incidence/ or exp mortality/ or mo.fs.
385507
11
exp prognosis/ or prognos:.mp.
471606
12
10 or 11
762949
13
exp cohort studies/ or cohort.mp.
507037
14
exp survival analysis/
51045
15
13 or 14
541177
16
12 and 15
185570
17
7 and 16
10
18
5 and 16
891
19
2 and 16
1805
20
7 or 9
124
21
18 or 19
1805
22
20 and 21
10
23
22 not 17
0
Set 17 looks to have reasonable results.
5686
119015
118
15350
13
1 Heart Failure, Congestive/mo [Mortality]
2 PROGNOSIS/
201691
3 1 and 2
760
4 exp cohort studies/
486726
5 3 and 4
#
3006
279
Search History
Results
1 exp Heart Failure, Congestive/
18118
1 Heart Failure, Congestive/mo [Mortality]
3006
2 incidence/ or exp mortality/ or mo.fs.
386505
3 exp prognosis/ or prognos$.tw.
473422
4 2 or 3
765434
5 exp cohort studies/ or cohort.tw.
508457
6 exp survival analysis/
51372
7 5 or 6
542818
8 4 and 7
186240
9 1 and 8
1206
2 exp HYPERTENSION/
43462
3 incidence/ or exp mortality/ or mo.fs.
170273
4 disease progression/
25821
5
((natural$ or disease$) adj (progress$ or course$ or
histor$)).ti,ab,sh.
6 exp prognosis/
22559
233970
7
(incidence or mortality or prognos$ or predict$ or course
or outcome$).ti,ab,sh.
636120
8 or/3-7
787700
9 exp survival analysis/
39950
10 exp cohort studies/
230028
11
(cohort$ or follow-up or compar$ or longitudinal$ or
1121484
prospective$ or multivariate or reproducib$).ti,ab,sh.
12 or/9-11
1167263
13 8 and 12
397018
14 1 and 2 and 13
224
15 limit 14 to english language
185
16 limit 15 to female
119
17 limit 16 to aging <65 to 79 years>
44
#
Search History
Results
1 hypertension.mp. or exp HYPERTENSION/
214880
*Heart Failure, Congestive/mo, co, pp, di, dt, et
2 [Mortality, Complications, Physiopathology, Diagnosis,
Drug Therapy, Etiology]
20090
3 1 and 2
1614
4
limit 3 to (english language and female and aging <65 to
79 years>)
172
5 incidence/ or exp mortality/ or mo.fs.
386964
6 exp prognosis/ or prognos$.tw.
474384
7 5 and 6
94656
8 exp cohort studies/ or cohort.tw.
509279
9 exp survival analysis/
51500
10 8 or 9
543721
11 7 and 10
40886
12 exp case-control studies/ or exp follow-up studies/
501042
13 11 or 12
517621
14 4 and 13
30
Ovid Prognosis strategy:
1. incidence/
2. or exp mortality
3. (incidence or mortality or prognos$ or predict$ or course or outcome$).ti,ab,sh.
4. mo.fs.
5.
exp prognosis
6.
exp disease progression
7.
((natural$ or disease$) adj (progress$ or course$ or histor$)).ti,ab,sh.
8.
or/1-7
9.
exp cohort studies
10. (cohort or follow-up or compar$ or longitudinal or prospective$ or multivariate
or reproducib$).ti,ab,sh.
11. exp survival analysis
12. or/9-11
1.
8 and 12
I didn’t Major or Focus MeSH disease Subject headings: too sensitive.
I combined Heart Failure, Congestive with Hypertension on the basis that Hypertension
and gender might influence disease progression.
I limited to English and Aging, on the basis that Aged was too narrow to encompass
longer term prognosis for this patient.
Bearing in mind the context, I didn’t search for freetext terms and I did limit to English
language, in order to identify a manageable number of references for a SHO to manage.
Re. the Filter - I tried to combine the best of all three suggested ‘prognosis’ strategies. I’d
found during the reading exercise that none quite hit the mark on the basis that:



exp prognosis/ found more than prognosis$ (hence Univ. of Alberta strategy
wasn’t best for me)
HIRU’s “mid-range” strategy didn’t appear to include Cohort Studies, identified
by HIRU as the single most useful term for a prognosis study.
.ti,ab,sh. found more than .tw.
Questions re. HIRU “mid-range” strategy:

Why incidence and not exp Morbidity which includes Incidence and Prevalence?

Exp Mortality includes mortality (MeSH)

Exp Prognosis/ finds more that prognos$

Why does HIRU strategy not include Exp Cohort-Studies/ which a. includes
Follow-Up Studies and b. has been identified by HIRU as single most useful term
for a prognosis study?
The above questions highlight the way in which McMaster HIRU strategies are
conducted. They are looking for optimal permutations and therefore the strategies are
computer generated not using librarian-guided intuition. This means that the alternatives
that our participant suggests are not regarded as providing an acceptable response with
regard to specificity.
Questions re. University of Alberta Advanced Strategy:

Why is follow-up in (1) separate from exp cohort studies in (5) which includes
Follow-up Studies?
For the same reason as above, the performance of the sensitivity/specificity trade-off is
regarded as less acceptable in line 5 than in line 1. The researchers prefer to report
individual terms so that when implemented the searchers can choose to drop particular
lines as and when required.
Note: Since this Unit was released the McMaster team has produced a new study looking
specifically at new terms for a prognosis filter:
BMC Med. 2004 Jun 9;2(1):23.
Developing optimal search strategies for detecting clinically sound
prognostic studies in MEDLINE: an analytic survey.
Wilczynski
NL,
Haynes
RB;
The
Hedges
Team.
Department of Clinical Epidemiology and Biostatistics, McMaster University,
Hamilton,
Ontario,
L8N
3Z5,
Canada.
[email protected]
BACKGROUND: Clinical end users of MEDLINE have a difficult time retrieving articles that are
both scientifically sound and directly relevant to clinical practice. Search filters have been
developed to assist end users in increasing the success of their searches. Many filters have been
developed for the literature on therapy and reviews but little has been done in the area of
prognosis. The objective of this study is to determine how well various methodologic textwords,
Medical Subject Headings, and their Boolean combinations retrieve methodologically sound
literature on the prognosis of health disorders in MEDLINE. METHODS: An analytic survey was
conducted, comparing hand searches of journals with retrievals from MEDLINE for candidate
search terms and combinations. Six research assistants read all issues of 161 journals for the
publishing year 2000. All articles were rated using purpose and quality indicators and categorized
into clinically relevant original studies, review articles, general papers, or case reports. The
original and review articles were then categorized as 'pass' or 'fail' for methodologic rigor in the
areas of prognosis and other clinical topics. Candidate search strategies were developed for
prognosis and run in MEDLINE - the retrievals being compared with the hand search data. The
sensitivity, specificity, precision, and accuracy of the search strategies were calculated.
RESULTS: 12% of studies classified as prognosis met basic criteria for scientific merit for testing
clinical applications. Combinations of terms reached peak sensitivities of 90%. Compared with the
best single term, multiple terms increased sensitivity for sound studies by 25.2% (absolute
increase), and increased specificity, but by a much smaller amount (1.1%) when sensitivity was
maximized. Combining terms to optimize both sensitivity and specificity achieved sensitivities
and specificities of approximately 83% for each. CONCLUSION: Empirically derived search
strategies combining indexing terms and textwords can achieve high sensitivity and specificity for
retrieving sound prognostic studies from MEDLINE.
This article is available free at:
http://www.biomedcentral.com/1741-7015/2/23
You will notice that the authors are suggesting a new one-line filter “explode
epidemiologic studies”. This would have picked out our article above as it includes the
study design filter “cohort-studies”
Table 1
Single term with the best sensitivity (keeping specificity ≥50%),
best specificity (keeping sensitivity ≥50%), and best optimization
of sensitivity and specificity (based on the lowest possible
absolute difference between sensitivity and specificity) for
detecting studies of prognosis in MEDLINE in 2000
Search term Sensitivity
Specificity
Precision
Accuracy
OVID
(%)
(%)
(%)
(%)
search*
Development Development Development Development
Validation
Validation
Validation
Validation
Diff (95%
Diff (95%
Diff (95%
Diff (95%
CI)†
CI)†
CI)†
CI)†
‡exp
64.9
78.6
1.1
78.6
73.4
8.5 (-4.6 to
21.7)
79.1
0.5 (-0.3 to
1.2)
1.4
0.3 (-0.2 to
0.7)
79.1
0.5 (-0.3 to
1.2)
epidemiologic
studies
*The search strategy is reported using Ovid's search engine syntax for MEDLINE.
The PubMed syntax is epidemiologic studies [MeSH]. †Diff = Difference,
comparing the development and validation data sets using the iterative method
of Miettinen and Nurminen for two independent binomial proportions. None of the
differences were statistically significant. ‡exp = explode, a search term that
automatically includes closely related indexing terms.
Table 2
Combination of terms with the best sensitivity (keeping specificity ≥50%), best
specificity (keeping sensitivity ≥50%), and best optimization of sensitivity and
specificity (based on abs [sensitivity-specificity] < 1%) for detecting studies of
prognosis in MEDLINE in 2000
Search
Sensitivity (%)
Strategy OVID Development
search*
Validation Diff
(95% CI)†
Specificity (%)
Development
Validation Diff
(95% CI)†
Precision (%)
Development
Validation Diff
(95% CI)†
Accuracy (%)
Development
Validation Diff
(95% CI)†
Best
Sensitivity
incidence.sh.
OR exp mortality
90.1
79.7
1.7
79.7
82.3
79.7
1.6
79.7
-7.8 (-17.9 to
2.3)
0
-0.1 (-0.5 to 0.5)
0
prognos:.tw.
52.3
94.1
3.3
93.9
OR first
episode.tw.
48.1
94.2
3.2
94.0
-4.2 (-18.6 to
10.3)
0.1 (-0.3 to 0.5)
-0.1 (-1.3 to 1.3)
0.1 (-0.4 to 0.5)
prognosis.sh.
82.9
83.7
1.9
83.7
OR
diagnosed.tw.
73.4
84.1
1.8
84.0
OR follow-up
studies.sh.
OR prognos:.tw.
OR predict:.tw.
OR course:.tw.
Best Specificity
OR cohort.tw.
Best
Optimization of
Sensitivity &
Specificity
OR cohort:.mp.
-9.5 (-21.5 to
2.5)
-0.4 (-0.2 to 1.1) -0.1 (-0.7 to 0.5)
0.3 (-0.2 to 1.0)
OR
predictor:.tw.
OR death.tw.
OR exp models,
statistical
*Search strategies are reported using Ovid's search engine syntax for MEDLINE. The PubMed syntax
is embedded in PubMed's Clinical Queries (see Discussion). †Diff = Difference, comparing the
development and validation data sets using the iterative method of Miettinen and Nurminen for two
independent binomial proportions. None of the differences were statistically significant. sh = subject
heading; exp = explode, a search term that automatically includes closely related indexing terms; :
= truncation; tw = textword (word or phrase appears in title or abstract); mp = multiple posting
(term appears in title, abstract, or MeSH heading).