For the 3-year study period, 61,013 unweighted patient visits were included, representing a national sample of about 70 million annual ED visits by non-institutionalized adults younger than 65 years of age between 2004 and 2006. In the entire sample, about 14% of the visits arrived by ambulance, 83% did not arrive by ambulance, and 4% of these visits had modes of arrival that were coded as missing or unknown.
displays the baseline patient demographic characteristics as well as unadjusted comparisons of clinical factors by mode of arrival for included visits during the 3-year study period. Patients who arrived by ambulance were more likely to be men, to be older, and to have a higher triage acuity assessment compared to patients who did not arrive by ambulance. Considering the most common reasons for visit, those who arrived by ambulance were more likely to have chest pain (9.3% vs. 5.2%, p < 0.001) and shortness of breath (4.0% vs. 1.8% p < 0.001), while those who did not arrive by ambulance were more likely to have headache (3.6% vs. 2.5%, p < 0.001), back pain (3.7% vs. 2.3%, p < 0.001), and upper extremity laceration (2.1% vs. 0.9%, p < 0.001).
Patient characteristics by mode of arrival for included visits.
In the unadjusted analysis, ED visits by patients with private insurance used an ambulance 11.5% of the time, patients without insurance used an ambulance 13.4% of the time, and patients with Medicaid used an ambulance 15.8% of the time (Medicaid vs. private p<0.001, Medicaid vs. self-pay p=0.001, self-pay vs private p=0.002).
The adjusted odds of arrival to the ED by ambulance vs. not by ambulance for a variety of patient characteristics are in . For this analysis, the post-estimation testing showed adequate goodness-of-fit. The association between insurance status and ambulance use was found to be independent of age, race, ethnicity, sex, transport to a hospital in an MSA, triage acuity, region of the country, chief complaint, and year of study. Compared to visits by patients with private insurance, those with Medicaid (aOR 1.60, 99% CI = 1.43 to 1.80) or without health insurance (aOR 1.43, 99% CI = 1.28 to 1.60) had significantly higher odds of arrival by ambulance. When compared to each other, visits by patients with Medicaid had smaller but significantly increased odds of ambulance use compared to uninsured visits (aOR 1.16, 99% CI = 1.03 to 1.30). Also included in the analysis (but not in the Table) were the visits by the unique group of patients under age 65 years with Medicare (chronically disabled or with end-stage renal disease), representing less than 5.7% of the included visits. The adjusted odds for ambulance use for this select group, compared to private insurance patient visits, was aOR 1.82 (99% CI = 1.5 to 2.2)
Adjusted odds of arriving by ambulance compared to alternate modes of transportation 2004–2006.
Male sex, older age, higher acuity classification, Northeast region, and MSA location of the ED were also each independently associated with higher adjusted odds of ambulance use for emergency care. However, these associations did not confound the magnitude or statistical significance of the association between insurance type and ambulance use. Using an enhanced model that included demographic co-variables available from the 2006 survey (median income, education level, and urban/rural character for each patient’s zip code), an analysis was performed in 2006-only ED visits. In the 2006-only analysis, compared to visits by patients with private insurance, patients with Medicaid (aOR 1.54, 99% CI = 1.27 to 1.87) and patients without insurance (aOR 1.35, 99% CI = 1.10 to 1.66) had increased odds of ambulance use for prehospital care and transportation. When the original model (without the additional covariates) was applied to the 2006 only data, the results did not materially differ compared to the combined 2004 through 2006 data.
We measured a statistically significant positive interaction between lack of insurance and metropolitan location (P value for interaction term = 0.01), with MSAs accounting for the bulk of the increased odds of ambulance use among uninsured compared to private insurance visits. No other interaction effects measured, including those between triage acuity and insurance, were found to be statistically significant (P values for interaction terms ranged from 0.2 to 0.9). The adjusted associations between insurance and ambulance use, as interacted with metropolitan location and overall acuity, are shown in and . Other than for uninsured patients in non-metro areas, for whom odds of ambulance no longer appear to differ significantly from patient encounters with private insurance, the interaction results were not materially different from the findings of the adjusted overall model. We measured higher relative odds of ambulance use for visits by patients with Medicaid and without insurance relative to patients with private insurance within low-, medium-, and high-acuity episodes.
Adjusted relative odds of arriving by ambulance compared to alternate modes of transportation 2004–2006, by acuity.
Adjusted relative odds of arriving by ambulance compared to alternate modes of transportation 2004–2006.
Six percent of the visits had incomplete insurance information documented in this sample. Complete insurance information was less likely to be found among ambulance arrivals than other modes of arrival (93% vs 94%, p = 0.005). In part because of this small but statistically significant difference, we conducted a sensitivity analysis, reclassifying patient visits for which insurance status was initially classified as “unknown” or blank to “no insurance/self-pay,” “Medicaid,” or “private insurance,” respectively. The results of the fully adjusted analysis, after incorporating the sensitivity tests, did not change the magnitude (ambulance use for Medicaid visits, aOR range 1.5 to 1.6, p < 0.001 compared to private insurance; ambulance use for self-pay visits, aOR range 1.3 to 1.4, p < 0.001), nor the significance of the findings.