Can current EMS dispatch protocols identify layperson-reported sentinel conditions?

Managed care organizations are seeking opportunities to capitate for emergency medical services (EMS). These and others pressures are motivating EMS planners to find innovative ways to redeploy existing EMS resources. A successful redeployment of resources implies a mechanism for more carefully matching EMS resources to need than currently exists.

To determine whether the methods dispatchers currently use to assign nature codes (NCs) and severity codes (SCs) also can distinguish between patients with important sentinel conditions and those without.

This was a six-month prospective study (June to November 1997). Portland Fire Bureau (PFB) paramedic units documented dispatcher-assigned NCs and SCs and the presence or absence of study-established sentinel findings. The PFB paramedics also verified or corrected dispatcher-assigned NCs and SCs using dispatch algorithms identical to those in use at this urban dispatch center. Cross-tabulation tables (SPSS version 6.1) with chi-square statistics were established to illustrate the relationship between SC strata within specific NCs and the presence or absence of sentinel findings.

One thousand two hundred eighty-five usable cases fell into 25 unique NCs. The designation SC 1 (emergent) was assigned by the dispatcher 307 (24%) times, SC 3 (urgent) was assigned 907 (71%) times, and SC 9 (neither emergent nor urgent) was assigned 26 (2%) times. The SC was missing 45 (3%) times. The PFB records were matched to 1,040 (82%) dispatch records. Sentinel conditions were identified in 411 (40%) of these cases. Eight (32%) of 25 NCs were stratified into two or more levels of dispatcher-determined SCs. One cross-tabulation table for each of these eight NCs was developed to display the relationship between SC strata and the presence or absence of a sentinel condition. Five tables produced statistically significant chi-square tables (p < 0.05). None achieved the study-specified level of 95% sensitivity.

Current dispatcher-assigned NCs and SCs do not appear adequate to detect callers with study-developed sentinel criteria.