Our analysis of over 1.3 million ambulance conveyances identified differences in pre-alert practice that were not attributable to case mix. We identified that pre-alert practice was affected by a combination of hospital, clinician, and patient factors. Although many pre-alerts were for key pre-alertable conditions (sepsis, stroke, STEMI, or trauma), around two-thirds of pre-alerts were for conditions that may require a higher level of clinical judgement when deciding whether to pre-alert. Within this analysis, receiving hospital and hospital turnaround status were key factors in pre-alert practice suggesting that ambulance clinician concerns about anticipated response from the receiving ED may have an influence over decision making. There was some evidence that newly qualified paramedics may pre-alert more than more experienced clinicians, although the size of effect detected was small. We found no evidence that clinicians in the final hour of their shift were more likely to pre-alert.
Despite the impact and importance of pre-alerts on patient care, we have not identified other literature exploring pre-alert practice and factors affecting pre-alert rates for general populations, although several studies have reported on pre-alert practice for specific conditions where the benefit of pre-alert is more clearly defined.
Differences in practice between ambulance services suggests local protocols and priorities have an impact on pre-alert decisions. Boyd et al identified important differences in ambulance service guidance in the UK, with differences in physiological thresholds for pre-alert even for conditions with established care pathways with services listing between 4 and 45 conditions as suitable for pre-alert.19 There was also variation in national ambulance and ED guidance regarding the criteria to determine whether a pre-alert is appropriate.2 In the US, emergency medical services (EMS) criteria for pre-alerts are likely to vary and practice appears to be dictated by requirements of local EDs.3,26 Lin et al identified statistically significant differences in EMS prehospital notifications for stroke between hospitals and regions, and concluded that disparities in EMS prenotification use occurred by state and geographic region.
Inconsistent pre-alerting within individual hospitals has been reported elsewhere, with Sheppard et al and Brown et al both identifying under-alerting of patients with suspected stroke, reporting that pre-alerts were not consistently used in suspected stroke patients, with 27% of patients who were FAST positive not pre-alerted, and 22% of patients who met the local criteria for pre-alert not being pre-alerted.6,7
Other studies have identified differences in patient factors affecting pre-notification for specific conditions. Blusztein et al identified male sex as an independent predictor of pre-notification for STEMI.27 Lin et al found that female patients were less likely to receive EMS prenotification for stroke but also identified higher likelihood of EMS prenotification for younger patients, and significant ethnic disparities in prenotification, with an adjusted odds ratio of pre-alert for black patients of 0.94 (CI0.92–0.97) compared with white patients.28 Sheppard et al did not identify any statistically different racial or sex differences in stroke pre-alerting, which could be due to the small overall sample (n = 271).7
It is unclear whether differences identified within our study are due to case mix that was not detected within the model, implicit bias in practice or different presentation of symptoms. We were unable to explore racial differences due to poor reporting of ethnicity within the ePR. However, our study supports findings of previous studies that report disparities in treatment based on non-clinical patient characteristics, and suggests that inequalities in care exist.29
Our findings also demonstrate that pre-alert decision-making is affected by clinician and contextual factors, with pre-alert decisions being affected by anticipated ambulance handover delay as well as clinical experience. Weyman et al similarly identified that clinician perceptions of personal vulnerability and organisational blame in the event of a wrong decision (e.g. waiting in a queue) are likely to influence more risk-averse decision-making.30
Limitations
The differences in pre-alert rates between the three ambulance services are likely due partly to organisational differences but also due to differences in rates of recording the pre-alert within the ePR. At site 1, pre-alert recording was mandated during the final 6-month period and therefore likely to provide an accurate estimate of true pre-alert recording rates. However, recorded pre-alerts did not increase during this period.
We are unclear whether the missing data is missing at random. Missing data may be more likely for sicker patients where the ePR is likely to be completed after patient handover. However, although pre-alert recordings may not be missing at random due to patient condition, this is unlikely to affect other results such as clinician role, receiving hospital or patient sex. Even just using data for site 1 and considering sites 2 and 3 as sensitivity analyses, still demonstrated significant variation in pre-alert rates between ambulance clinicians.
Difference in pre-alert rates between sites may be due in part to under-recording of pre-alerts, or due to differences in local protocol. Data suggests that sepsis pre-alerts are higher at site 3 than for other sites, which reflects the local protocol requiring pre-alert for any red-flag sepsis. However, it is not known whether this reflects genuinely higher pre-alerting rates or higher recording of pre-alerts.
This study was undertaken within 3 UK ambulance services and transferability may be limited for settings outside the UK. However, given known recognised variation in practice and a lack of clear protocols within other settings the level of variation identified within this study is likely to be found elsewhere. The time period for which data was collected included the second period of COVID-19 lockdown in the UK (Jan – March 2021), which reduces the potential transferability of findings. The proportion of pre-alerts due to COVID-19 or respiratory disease were likely higher within this dataset than in other years. However, this is unlikely to affect pre-alert practice for other conditions significantly.
We adjusted for case mix using the UK AACE/RCEM non-physiological criteria for a pre-alert. However, coding of this field required assumptions to be made, including the use of physiological parameters for some non-physiological criteria, since working impression codes were either not available for certain pre-alertable presentations, or the severity of the patient presentation could not be determined by the working impression alone e.g. drug overdose, trauma, haemorrhage). This coding had to be undertaken on a service-by-service basis since provided fields and working impression codes differed between services. It is possible that this may lead to differences in categorisation in the model.
Differences in proportions of patients with sex labelled as ‘transgender’ or ‘not reported’ differed by ambulance service, suggesting that some ambulance services did not use the category ‘transgender’ within their coding and this field may not be reliable.
The analysis was exploratory in nature and not confirmatory. The aim was to explore what variables might predict the use of pre-alerts, with the aim to guide future research. Although the size of effect was different within the combined and separate models (which may be expected as these were different datasets), the different models did all show that clinical variables are the key predictors, with hospital factors, anticipated handover delay, patient sex and clinician role all being predictors.