Table 3 Model performance of three regression methods for mapping the PHQ-8 to the EQ-5D-3L utility scores

1

BETAMIX M1a

1 component without truncation

Probability mass at full health

0.0904

0.2014

0.2621

18

18

18

2

BETAMIX M1b

2 components without truncation

Probability mass at full health

0.0130

0.1868

0.2381

15

11

13

3

BETAMIX M1c

2 components with truncation

Probability mass at full health

−0.0043

0.1839

0.2370

12

9

10.5

4

BETAMIX M1d

2 components with truncation

Probability mass at full health and truncation point

−0.0024

0.1861

0.2390

13

13

13

5

BETAMIX M2a

1 component without truncation

Probability mass at full health

0.0866

0.1962

0.2607

17

17

17

6

BETAMIX M2b

2 components without truncation

Probability mass at full health

0.0101

0.1825

0.2349

8

7

7.5

7

BETAMIX M2c

2 components with truncation

Probability mass at full health

−0.0038

0.1806

0.2341

6

5

5.5

8

BETAMIX M2d

2 components with truncation

Probability mass at full health and truncation point

−0.0013

0.1813

0.2355

7

8

7.5

9

BETAMIX M3a

1 component without truncation

Probability mass at full health

0.0659

0.1864

0.2504

14

16

15

10

BETAMIX M3b

2 components without truncation

Probability mass at full health

0.0119

0.1800

0.2321

5

1

3

11

BETAMIX M3c

2 components with truncation

Probability mass at full health

0.0020

0.1774

0.2328

2

3

2.5

12

BETAMIX M3d

2 components with truncation

Probability mass at full health and truncation point

0.0057

0.1765

0.2326

1

2

1.5

13

OLS M1

0.0000

0.1837

0.2374

11

10

10.5

14

OLS M2

0.0000

0.1798

0.2347

4

6

5

15

OLS M3

0.0000

0.1784

0.2331

3

4

3.5

16

TOBIT M1

−0.0263

0.1870

0.2413

16

15

15.5

17

TOBIT M2

−0.0264

0.1834

0.2389

9

12

10.5

18

TOBIT M3

−0.0264

0.1836

0.2390

10

14

12