Table 2 Readmission risk prediction without −Model (1), and with −Model (2), text-mining.
From: Text mining of outpatient narrative notes to predict the risk of psychiatric hospitalization
Dependent variable (Yi,t+i) | ||
|---|---|---|
Model (1) | Model (2) | |
Before CTO | 2.158*** (0.283) | 2.114*** (0.266) |
During CTO | 0.912*** (0.233) | 0.873*** (0.229) |
Clinic Visit | −0.138 (0.203) | 0.219 (0.232) |
Injection | −0.843*** (0.198) | −0.708*** (0.214) |
Prescription (Rx) Change | 0.676*** (0.245) | 0.603** (0.264) |
Rx Change x Before CTO | −0.464 (0.544) | −0.515 (0.559) |
Rx Change x During CTO | −0.775 (0.475) | −0.805 (0.495) |
Hospital Discharge | 2.480*** (0.229) | 2.520*** (0.224) |
Treatment Dropout | 1.284*** (0.232) | 1.317*** (0.242) |
Unstable | −14.032 (256.247) | |
Stable | −0.500** (0.196) | |
Adverse Appearance | −0.143 (0.604) | |
Adverse Behavior | 0.227 (0.319) | |
Adverse Danger | 0.872 (0.561) | |
Adverse Impulse Control | 0.173 (0.639) | |
Adverse Insight | −0.076 (0.302) | |
Adverse Language | 0.524** (0.261) | |
Adverse Mood | 0.082 (0.225) | |
Adverse Thought Content | 0.074 (0.233) | |
Adverse Thought Process | −13.278 (39.069) | |
Appearance | −0.297 (0.194) | |
Behavior | 0.033 (0.195) | |
Danger | 0.075 (0.291) | |
Impulse Control | −0.120 (0.320) | |
Insight | 0.296 (0.223) | |
Language | 0.084 (0.226) | |
Mood | −0.097 (0.192) | |
Thought Content | −0.376* (0.198) | |
Thought Process | 0.394 (0.395) | |
Constant | −5.073*** (0.218) | −5.005*** (0.213) |
Observations | 18,382 | 18,382 |
Log Likelihood | −1,175.571 | −1,157.678 |
Akaike Information Criterion | 2,373.141 | 2,377.355 |
Bayesian Information Criterion | 2,459.152 | 2,619.748 |
