Introduction
Background
Machine learning
Performance evaluation
Data sources
Clinical NLP software
Methodology
Data
Asthma | Diabetes | COPD | Readmission | |
---|---|---|---|---|
Discharge summary #1 | 1 | 1 | 0 | True |
Discharge summary #2 | 1 | 0 | 1 | False |
Machine learning
HRRP cost
Modeling cost
C | Total cost of Diagnostic Related Group (DRG) |
C
np
| Total cost of DRG for new patient(s) under analysis |
ω | Risk adjustment factor |
R | Number of readmissions for current fiscal period |
P | Number of total patients in DRG for current fiscal period |
ρ
| Expected rate |
\( \hat{\rho } \)
| Predicted rate |
p
r
| Probability of needing readmission |
N | Number of new patients under consideration |
Ns
| Number of new patients to select for intervention |
p
s
| Probability of intervention success |
\( \bar{c}_{t} \)
| Average cost of patient intervention (i.e. home healthcare professional) |
Optimal patient intervention
Framework
Results
C | $10,000,000 |
C
np
| $10,000 * N |
ω | 1 |
P | 1000 |
\( \bar{c}_{t} \)
| $800 |
Assumption | R |
\( \rho \)
|
\( \hat{\rho } \)
|
\( p_{s} \)
|
---|---|---|---|---|
A | 143 | 0.14 | 0.143 | 0.90 |
B | 143 | 0.14 | 0.143 | 0.97 |
C | 205 | 0.20 | 0.205 | 0.90 |
D | 205 | 0.20 | 0.205 | 0.97 |
Assumption | MinCost | Classification | AI |
---|---|---|---|
A | $11,920 | $18,640 | $96,720 |
B | $11,040 | $18,640 | $96,720 |
C | $6720 | $18,640 | $96,720 |
D | $6160 | $18,640 | $96,720 |
Assumption | Classification (%) | AI (%) |
---|---|---|
A | − 36.05 | − 87.67 |
B | − 40.77 | − 88.58 |
C | − 63.94 | − 93.05 |
D | − 66.95 | − 93.63 |
Average | − 51.93 | − 90.07 |