One primary finding is that the increase in the number of emergency calls to the EMS Berlin was greater from 2018 to 2021 than the corresponding increase in population growth, whereas this is not specifically driven by low-acuity calls. Although we had no individual outcome data, i.e. no data on which diagnoses were later clinically confirmed on arrival at the patient or in the hospital or which patient ultimately survived the incident, our results are important evidence to provide a context for the recent media coverage [1, 2]. The increasing overall call number is underlined by official EMS operational statistics, showing that deployments increased by 8.7% from 2018 to 2021, both emergency rescues (+ 5.5%) and emergency transports (+ 37.7%) [40, 41]. Speculation about the reasons for the overall increase in utilisation is beyond the scope of this study. Although demographic ageing is a major issue in Germany, it is unlikely to be decisive here, because the population of Berlin is ageing slower than in many other parts of Germany, and its mean age remained constant at 42.6 years in the time period of the study (2018–2021) [42].

Among the low-acuity calls, some protocols and codes stand out because of their high absolute numbers (e.g. abdominal pain with no complications and responsive person after fainting).

The increasing protocol adherence is a result of MPDS quality management within the EMS, providing a nearly complete dataset for 2021. However, this raises the question of possible bias to the temporal analysis because it cannot completely be ruled out that some codes are missing not-at-random, when dispatchers aborted coding. The month-by-month analysis (Fig. 3) indicates that the calls with aborted protocols are in the majority high-acuity calls, because a rapid drop in protocol aborts within only 4 months is accompanied by a disproportionate increase in documented high-acuity cases, which is presumably a coding effect. However, there are three reasons why it is reasonable to assume that not all missing values are in reality high-acuity calls. Firstly, there is also a small increase in documented low-acuity cases within the 4 months. Secondly, internal quality management of the EMS Berlin indicates that some protocol aborts in earlier years was due to different training levels of individual dispatchers. Thirdly, some protocol aborts took place because until the beginning of 2020, call-taking without using full MPDS codes was accepted for the subgroup of calls which were transferred from other emergency call centres (e.g. public transport or police), and these cases presumably also include both low- and high-acuity cases. Taken together, it is reasonable to assume that the calls with unspecified acuity are in majority (nearly 90%) but not exclusively high-acuity calls. Only this extreme, hypothetical case would result in a very slight increase of the low-acuity proportion from 29.8 to 30%. Therefore, increasing protocol adherence did most likely not distort the low-acuity proportion to an extent that would reverse the acuity trend. The proportion of low-acuity calls remained more or less constant, it might have slightly decreased.

The regression sheds light on low-acuity predictors. Necessarily, the predictive power of the model in terms of goodness-of-fit is strongly limited, given that no data on individual pre-existing morbidity—which is decisive for whether a medical incident is per se acute—was available. Nevertheless, the regression reveals highly significant predictors. The most relevant variable is age, which should be interpreted in light of general morbidity. There is a certain bias due to the MPDS algorithms [43], because in some protocols older age leads to higher classifications in category and/or sub-determinant.

The social status index has an overall range of -1.5 (highest status) to 4 (lowest), but most neighbourhoods lie between -1 and 2. Comparing these values, the odds of low-acuity can be estimated to be 2.5% higher for the PLR with lower social status. It remains speculative whether this is caused by different availabilities of alternative health care resources, health literacy, morbidity or other effects. For example, it is possible that residents in areas with lower social status are less informed about suitable alternative health infrastructure and therefore call the EMS more often. It could also be that more non-life-threatening health incidents occur in areas with lower social status because pre-existing morbidity is higher. Future studies with additional data might further investigate this interesting correlation. Odds for a low-acuity call are 2% higher at the weekend, which could theoretically be due to less health care alternatives, but this also remains speculative.

The low-acuity odds were 12% higher for calls concerning females. Here, the different disease prevalence of women and men—especially in childbearing age, when low-code proportions are highest—should be kept in mind [44].

In this study, including the proportion of migrants did not noticeably improve the regression model. Berlin has a wide variety of migrant communities, with many people living in the city for a long time already, so the variable might be too imprecise to detect relevant differences in 112 calls.

We found great differences in the age structure of patients between districts, which is logical given their different age composition, but only slight ones concerning low-acuity proportions.

For Bavaria, Hegenberg et al. found an increase of calls exceeding population growth [16], similar to the findings for Berlin. They demonstrated higher call rates in cities compared to rural municipalities and an association between municipality size and call proportion at weekends. Schehadat et al. analysed data from Rhineland-Palatinate regarding patient transport from the scene and concluded, despite some regional differences, that population density did not significantly determine EMS utilisation. In the present study, basic differences between more vs. less densely populated areas in Berlin were investigated. We observed a slightly different mean call number, which was not statistically significant.

We included all Omega and nearly all Alpha codes among the low-acuity codes. This is debatable and other classifications are justifiable. In principle, we agree with other researchers who have stressed that MPDS categories do not map the urgency of calls in a linear, symmetrical way [7, 11]. On the other hand, several studies show an association of MPDS codes with acuity and outcome to a certain degree: Garza et al. showed increasing rates of transportation with lights and sirens from the scene with increasing determinant level, and a < 1% rate for Alpha codes [8, 12, 43]. Hettinger et al. demonstrated a correlation between MPDS codes and ED admission vs. discharge [43]. Another US study detected only few inappropriate Alpha dispatches and concluded that MPDS has a good ability to identify higher-acuity patients when protocol adherence is good [8]. It could be argued to also consider selected Bravo or Charlie codes as low-code; however, it was decided, similar to the expert panel in a study by Shah et al. [6], not to include any (except for upcodings).

Not all low-acuity calls are suitable for handling without EMS involvement. However, they can be starting points for how to avoid overtriage [14] and how to find alternatives to ambulance dispatch, like OCMS visits, ambulatory appointments, mental health joint response car [45], Sociolance [46], advance provider response units [47] or pre-hospital emergency nurse [48]—depending on individual medical need and context. The EMS Berlin has recently expanded its code list for transfer to the OCMS [49], including many Alpha codes which are frequent in our analysis. For example, abdominal pain requires ED treatment only in a minority of cases, although these are not always trivial to identify [50]. The considerable number of upcodings (B00/C00) in the dataset indicates that the transferral is not always possible, at least partly due to OCMS capacity.

Omega codes—which are widely regarded as the lowest-acuity calls [11]—make up only for a small amount of what we defined as low-acuity. This suggests that Alpha codes play a major role in scenarios to optimise EMS resources.

Given the increasing utilisation with a continuously high proportion of high-acuity cases and heterogeneous geographical distribution, the EMS should apply predictive dispatching in order to meet the challenging demand and secure quick responses. This means that available ambulances and other rescue infrastructure are allocated dynamically throughout the state of Berlin, estimating the most probable future demand in real-time.

Young people call the emergency number less frequently, but younger age is the strongest predictor for low-acuity. Therefore, this group might be suitable for target group-specific health education. An information campaign could inform them about alternative health care institutions, which are suitable for certain forms of low-acuity incidents and can improve the efficiency of care for all involved, because the patient does not need to wait for a long time in the ED, while the EMS and the ED save resources for life-threatening incidents.

Future studies should investigate the underlying morbidity of emergency calls and the predictors of cross-sectoral patient pathways after dispatch, which requires additional data. It would also be interesting to investigate the low-acuity call incidence depending on the spatial proximity of alternative medical infrastructure like hospitals and outpatient physicians. We recommend facilitating an international consensus on low-acuity MPDS codes, considering different national settings and emergency dispatch centres.

This is one of only few studies using a large individual-level secondary EMS dataset in Germany and the first to include all Berlin dispatch protocols from the recent four years. The study design benefits from MPDS implementation in Berlin as an internationally standardised system [26]. Entry errors can be considered as marginal due to the electronic documentation process. The study has good internal validity, as it provides a transparent, reproducible low-code classification.

A methodological limitation is the increasing MPDS protocol adherence (see the “Discussion” section). Predictive performance of the model is necessarily limited because no data on pre-existing individual morbidity is available. Sociodemographic indicators were available on regional, not on individual level. Therefore, each emergency call was merged with the sociodemographic parameter of the geographical area, which results in an approximation for each individual. This limits the accuracy of the regression compared to other variables like age which are available individually. We acknowledge this by referring to the geographical area when interpreting the results. We used the emergency locations as classifier for merging, which in some cases like commuters and tourists does not coincide with the place of residence. We found no direct evidence that the observed trends were specific for Berlin; however, the external validity with respect to other dispatch centres can be limited depending on different call-taking procedures and local health care context.