Skip to main content

Table 2 Summary statistics comparing the univariate #2 and multivariate #1 approaches, overlap cases, and the pool of controls

From: A proposed case-control framework to probabilistically classify individual deaths as expected or excess during extreme hot weather events

Variable

Description

Uni. #2 probable excess deaths

Multi. #1 probable excess deaths

p-value between uni. #2 and multi. #1 results

Overlap between uni. #2 and multi. #1

Overlap between all combinations

Pool of controls

N

Total number

114

114

 

72

30

11632

Age (%)

Age at death (%)

      

  < 75 years

83.3

52.6

<0.001

79.2

96.7

37.4

  > = 75 years

16.7

47.4

 

20.8

3.3

62.6

Location of death (%)

Location of death (%)

      

 In hospital

37.7

37.7

0.288

34.7

0

51.9

 Residential institution

17.5

27.2

 

19.4

0

30.9

 Home

36.0

28.9

 

37.5

80.0

14.5

 Other

8.8

6.1

 

8.3

20.0

2.7

Neighborhood deprivation quintile (%)

Neighborhood deprivation quintile (%)

      

 1 (least deprived)

6.1

0

0.049

0

0

20.7

 2

4.4

0

 

0

0

20.7

 3

20.2

14.0

 

15.3

20.0

19.2

 4

21.9

33.3

 

26.4

13.3

19.1

 5 (most deprived)

47.4

52.6

 

58.3

66.7

20.2

Population density quintile (%)

Population density quintile (%)

      

 0–181 persons/km2

1.7

0

0.829

0

0

20.1

 182–526 persons/km2

8.8

1.8

 

2.8

3.3

19.5

 527–681 persons/km2

10.5

18.4

 

8.3

16.7

19.9

 682–2356 persons/km2

28.9

30.7

 

33.3

20.0

19.8

 2357–2573 persons/km2

50.0

49.1

 

55.6

60.0

20.7

Greenness

Mean NDVI measurement

0.252

0.290

0.008

0.256

0.254

0.329

Selected underlying causes of death (%)

Underlying cause of death (%)

      

 Cardiovascular (ICD-10 starting with I)

17.5

28.1

NA

22.2

36.7

27.1

 Respiratory (ICD-10 starting with J)

6.1

11.4

 

8.3

6.7

10.0

 Cancer (ICD-10 starting with C)

36.0

25.4

 

30.6

10.0

30.7

 External (ICD-10 starting with X)

14.9

8.8

 

13.9

26.7

3.1

  1. Identified cases from the univariate and multivariate combinations were compared with respect to the five variables used in the analyses, as well as with respect to selected underlying causes of death. The overlap cases identified in both combinations (N = 72) and by all combinations (N = 30) were compared with the pool of controls