Salloum et al developed several models for predicting the performance status of a patient with lung cancer using claims-based measures. The authors are from Hentry Ford Health System in Detroit, Wayne State University, and Virginia Commonwealth University in Richmond.
Models developed: 6, with 3 pre-2009 and 3 post-2009
Patient selection: age >= 50 years with primary lung cancer
Outcomes: good performance status vs poor (with good defined as ECOG 0 or 1 and Karnofsky Performance Status >= 80.
Parameters for Model 2 for post-2009:
(1) age at diagnosis
(2) AJCC Stage IV disease
(3) chronic obstructive pulmonary disease (COPD)
(4) number of inpatient stays
(5) any number of outpatient visits
(6) number of Emergency Department visits
(7) any number of durable medical equipment (DME) claims
Parameter |
Finding |
Points |
AJCC Stage IV disease |
absent |
0 |
|
present |
1 |
COPD |
absent |
0 |
|
present |
1 |
outpatient visits |
none |
0 |
|
>= 1 |
1 |
DME claims |
none |
0 |
|
>= 1 |
1 |
X =
= (-0.04 * (age in years at diagnosis)) - (0.55 * (points for Stage IV disease)) - (0.88 * (points for COPD)) - (0.8 * (number of inpatient stays)) + (1.08 * (points for outpatient visits)) - (0.25 * (number of ED visits)) - (0.6 * (points for durable medical equipment)) + 4.25
probability of a good performance status by Model 2 for post-2009 data =
= 1 / (1 + EXP((-1) * X))