Table 3.
Hierarchical model of the association of factors with documentation of the Dix-Hallpike test in this emergency department dizziness population (n= 3,133 unique individuals).a
| OR, 95% CI, unless otherwise specified |
|
|---|---|
| Fixed Effects Parameters | |
| Age | 1.00 (0.99,1.02) |
| Female | 1.15 (0.74,1.79) |
| Race-ethnicity | |
| Non-Hispanic White | Ref |
| Mexican-American | 1.12 (0.68, 1.87) |
| Other | 0.65 (0.21, 2.05) |
| Episodic presentation | 2.17 (1.42, 3.33) |
| Symptoms | |
| Dizziness NOS | Ref |
| Imbalance | 2.04 (0.97, 4.31) |
| Vertigo | 4.88 (2.94, 8.11) |
| Month | 0.97 (0.95, 0.99) |
| Mid-level provider | 1.32 (0.75, 2.31) |
| Hospital | |
| 1 | Ref |
| 2 | 0.40 (0.11, 1.45) |
| 3 | 0.00 (0.00, 0.00) |
| 4 | 0.87 (0.18, 4.27) |
| 5 | 1.17 (0.27, 5.12) |
| 6 | 0.81 (0.12, 5.30) |
| Random Effects Parameter | |
| Provider, ICC b | 0.50 |
OR, odds ratio; CI, confidence interval; ICC, intraclass correlation coefficient. Mid-level provider = resident, physician assistant, or nurse practitioner
c-statistic = 0.93, indicating excellent model discrimination. The number of individuals in final model is reduced from total population of 3,522 due to exclusion of repeat visits (338) and missing data on provider (26) and race-ethnicity (25). The model included 73 unique providers who saw a median of 25 patients (IQR, 5–72)
ICC equal to 0.50 means that 50% of the variation in the probability of DHT utilization was explained at the provider level (i.e., due to physician practice differences) and the other half was due to random variation.