Table 2.
Topic modeling for 2020 hospital admissions comparing 10 topics for 2 COVID-19 patient encounter subcohorts: those with unhealthy opioid use and those with no unhealthy opioid use (N=4110). Subcohorts were identified using the Substance Misuse and Referral to Treatment Artificial Intelligence (SMART-AI) digital classifier for opioid misuse [26]. The topic numbers are labels and do not reflect a ranking of topics.
| Topic | Concepts | ||
| Unhealthy opioid use (n=102) | |||
|
|
1 | Cardiopulmonary illnesses and social determinants of health | |
|
|
2 | Opioid misuse comorbidities | |
|
|
3 | Renal and cardiac pathologies, HIV-related terms | |
|
|
4 | Neurological comorbidities with altered mental status | |
|
|
5 | Neurological workups with cardiac disturbances | |
|
|
6 | Critically ill/intensive care unit patients | |
|
|
7 | Risk for overdose with cardiopulmonary and respiratory distress | |
|
|
8 | Chronic opioid misuse with respiratory distress | |
|
|
9 | Opioid overdose patients | |
|
|
10 | Chronic illness and traumatic injury | |
| No unhealthy opioid use (n=4008) | |||
|
|
1 | Classic COVID-19 hospitalization with severity | |
|
|
2 | Sepsis-related, less clearly COVID-19 related | |
|
|
3 | COVID-19 orders/procedures, moderate to severe neurological orders | |
|
|
4 | Long-term intensive care unit patients | |
|
|
5 | Chronic obstructive pulmonary disease | |
|
|
6 | Interventions for acute respiratory failure | |
|
|
7 | Very neurologically focused, less COVID-19 related | |
|
|
8 | Chronic conditions associated with severe COVID-19 | |
|
|
9 | Less severe COVID-19 symptoms and measures | |
|
|
10 | Chronic disease highly susceptible to COVID-19 | |