hyp Hypertension was present if a patient received antihypertensive treatment before admission, or if hypertension was diagnosed during hospital stay by repeated detection of blood pressure = 160/95 mmHg.
ihd Ischemic heart disease was present if a patient had a history of IHD, or had IHD diagnosed during the hospital stay.
apo Previous stroke was recorded if the patient had previously experienced a stroke.
odd Information concerning other disabling disease was obtained on admission and included disabling diseases other than previous stroke (e.g., amputa-tion, multiple sclerosis, severe dementia, heart failure, latent or persistent respiratory insufficiency). Thus, various diseases were not registered sep-arately, and the influence of specific diseases was not evaluated.
alco Alcohol was coded if a patient was drinking on a daily basis. Ex-alcohol consumers were coded as non-alcohol consumers.
dm Patients with known diabetes before stroke and patients with diabetes diagnosed after stroke onset either during the hospital stay or because admission plasma glucose was ¿0.11 mmol/L, in accordance with the World Health Organization diagnostic criteria for diabetes.
smoke Smoking was coded if a patient smoked any kind of tobacco on a daily basis. Ex-smokers were coded as non-smokers.
af Atrial fibrillation was diagnosed if the Electrocardiogram obtained on ad-mission revealed AF.
hemo A CT scan determined stroke type as hemorrhage (bleeding) or infarct (an artery is blocked by some obstruction, e.g. a blood clot or cholesterol deposit).
sss The initial neurological stroke severity was assessed with theScandinavian Stroke Scale (SSS)at the time of the acute admission. The SSS evaluates level of consciousness; eye movement; power in arm, hand, and leg; orien-tation; aphasia; facial paresis; and gait on a total score from 0 (worst) to 58 (best)1.
cla Intermittent claudication was present if a patient had a history of inter-mittent claudication, or had interinter-mittent claudication diagnosed during the hospital stay. Intermittent claudication is a cramping sensation in the
1The scale is available athttp://www.strokecenter.org/Trials/scales/scandinavian.html.
5.2 Copenhagen Stroke Study Database 93
legs that is present during exercise or walking and occurs as a result of decreased oxygen supply2.
temp Body temperature on acute admission was recorded with a Diatek model 9000 infrared aural thermometer (Diatek); this device registers tympanic membrane temperature, which correlates well with core body temperature.
Body temperature was coded astemperature <37.0◦ C or≥37.0◦ C.
The long-term follow-up data on mortality and date of death were obtained from the Danish Central Registry of Persons, where date of death for all residents in Denmark is recorded through a unique 10-digit identification code containing information on birth date. Another experienced neurologist (L.P. Kammers-gaard) who was blinded to data obtained on admission prospectively recorded the follow-up data. Follow-up was performed during the year 1999 with Decem-ber 29, 1999, as censoring date. Furthermore, the database was updated with survival information during the year 2004 with November 3, 2003, as censoring date. Six patients had immigrated to another country and were lost on follow-up (reducing the sample size to 993).
The distribution of the categorial risk factors, and the mean values (standard deviations) of the continuous factors (of the 993 patients included) are shown in Table5.5and Table5.6respectively. All risk factors, except forageandsex, have missing values. 552 patients had no missing values in any of the selected variables.
Table 5.7 and Figure5.2 show the distribution of survival times. In Table5.8 we compute the percentage of subjects failing (dead) within a given time from admission. We note that the survival times are distributed over a time interval spanning more than 11 years, and just 19 subjects, or less than 2%, were dead on arrival. However, more than 10% died within the first week, more than 22% within 3 months, and almost 1/3 of all patients were dead within a year, indicating very high short-term mortality rate. However, as indicated in Table 5.7, the 25% quartile is 153 days or 5 months, while the median survival time is 1259 days or 4 years! As the histogram also illustrates, the mortality is very high in the days and weeks after stroke onset, but once a patient has survived this critical stage, the chances of surviving on longer terms increase rapidly.
The strength of this study is that it is prospective and community-based
includ-2This cramping usually occurs in the calf, but may also occur in the feet. When intermittent claudication is discussed it is measured by the number of “blocks” (e.g. 1 or 2 blocks) one can walk comfortably. It often indicates severe atherosclerosis. One of the hallmarks of this clinical entity is that it occurs intermittently. It disappears after a brief rest and the patient can start walking again until the pain recurs. Intermittent claudication is often a symptom of severe atherosclerotic disease of the peripheral vascular system.
Variable Yes No sex (men/women) 438 (44.1%) 555
hyp 306 (32.9%) 625
ihd 189 (20.6%) 728
apo 195 (20.7%) 748
odd 205 (21.5%) 747
alco 261 (31.5%) 569
dm 148 (15.6%) 795
smoke 364 (44.4%) 455
af 162 (16.5%) 820
cla 107 (12.2%) 773
infarct hemorrhage
hemo 61 (7.6%) 744
<37.0◦ C ≥37.0◦ C
temp 502 (58.3%) 358
Table 5.5: Distribution of categorical COST variables.
Variable Mean Std.
age 74.3 11.0 sss 38.0 17.5
Table 5.6: Distribution of continuous COST variables.
]subjets with t= 0 Min Mean Max 25% 50% (median) 75%
19 0 1587 4262 153 1259 2828
Table 5.7: Distribution of survival times (in days) in the COST data set.
1 week 2 weeks 3 weeks 1 month 3 months 6 months 1 year
10.5 13.3 15.2 16.7 22.1 26.7 32.0
Table 5.8: Percentage of subjects failing (dead) within a given time from ad-mission in the COST data set.
5.2 Copenhagen Stroke Study Database 95
0 500 1000 1500 2000 2500 3000 3500 4000 4500
0 50 100 150 200 250 300 350
Distribution of survival times (in days) in the COST data set
Number of subjects
Survival time in days
Figure 5.2: Histogram of survival times (in days) in the COST data set.
ing all patients in a well-defined community hospitalized with stroke regardless age, stroke severity, or other complicating diseases. Moreover, the stroke admit-tance rate in the area is high and close to the incidence reported in population-based studies. A limitation is that patients who die at home are not included and this may underestimate mortality. However, the small number of patients with minor strokes not being admitted to hospital may counterbalance it. Fi-nally, because we have a sizeable study population and a lengthy follow-up, we consider bias to be of no major importance for the main conclusions of this study.
Chapter 6
Comparison of Stepwise Selection and Bayesian Model Averaging Applied to Real Life Data
In our first analysis we compare stepwise selection to BMA using the real life data sets presented in the previous chapter.