### Description

Forster and Lynn developed logistic regression models for predicting death in hospice patients. This can help triage hospice patients more effectively.  The authors are from George Washington University in Washington, D.C.

Parameters:

(1) gender

(2) pain

(3) no funds

(4) neoplasm

(5) Karnofsky score

(6) weakness

(7) congestive heart failure (CHF)

(8) hormonal therapy

(9) disoriented

The Karnofsky score can range from 0 to 100 (by 10's) but essentially should be in the range of 30 to 70. A patient with a higher Karnofsky score does not qualify for hospice care, while a person with a score less than 30 has a poor short term prognosis.

Parameter

Finding

Points

gender

male

1

female

2

pain

yes

1

no

0

no (lack of) funds

yes

1

no

0

neoplasm

primary lung or colon cancer

1

other

0

weakness

yes

1

no

0

CHF

yes

1

no

0

hormonal therapy

yes

1

no

0

disoriented

yes

1

no

0

where:

• The wording for "no funds" is somewhat awkward.

X for death within 3 months =

= (-0.93 * (points for gender)) + (0.61 * (points for pain)) + (1.30 * (points for no funds)) + (1.66 * (points for neoplasm)) - (0.06 * (points for Karnofsky score)) + (1.57 * (points for weakness)) + (1.23 * (points for CHF)) + (0.48 * (points for hormones)) - (1.24 * (points for disoriented)) – 0.28

Y for death within 6 months =

= (-2.33 * (points for gender)) + (1.12 * (points for pain)) - (0.39 * (points for no funds)) + (1.92 * (points for neoplasm)) - (0.04 * (points for Karnofsky score)) + (1.12 * (points for weakness)) + (1.91 * (points for CHF)) + (2.02 * (points for hormones)) - (1.43 * (points for disoriented)) + 1.43

probability of death in time period =

= 1 / (1 + EXP((-1) * (variable)))

Performance:

• The model for death in 3 months classified 89% of patients correctly.

• The model for death in 6 months classified 94% of patients correctly.

• The false positive rate for the 3 month prediction was 10% and for the 6 month prediction was 5%.

• The false negative rate for the 3 month prediction was 25% and for the 6 month prediction was 17%.