Trends and barriers of emergency medical service use in Addis Ababa; Ethiopia

A cross sectional quantitative and qualitative mixed study design was conducted in Addis Ababa. The city is the capital of Ethiopia with a population of more than 4.6 million people. The majority the population is less than 30 years of age. The common languages spoken in the city include Amharic (71.0%), Oromiffa (10.7%), Gurage (8.37%), and Tigringa (3.60%). In addition Oromiffa is the major language spoken in rural Addis Ababa and surrounding zones. Patients visiting emergency centers of selected Addis Ababa hospitals with time sensitive disease and trauma were included in the study. The city has 11 public hospitals, five of which were specifically selected for inclusion in this study (St. Paul’s hospitals millennium medical college (SPHMMC), Tikur Ambassa Specialized hospital (TASH), All African leprosy and tuberculosis rehabilitation and training (ALERT) Hospital, Tirunesh Beijing Hospital, and Minilik II hospital). The selected hospitals are tertiary care public hospitals and provide advanced trauma care. In addition, the hospitals are located in different directions and in the gateways of the city. These parameters for hospital selection allowed for a patient sample with varying socio economic, clinical and demographic backgrounds and minimized double counting.

The study period was for one month in March of 2017. The assumption was made that there are no significant differences in emergency trends and ambulance use in Addis Ababa based on month. All consecutive emergency patients who arrived directly from the scene in the study month and met inclusion criteria were included. The inclusion criteria were those patients who sustained severe trauma (labeled red or orange by South Africa Triage scale) and illnesses with time sensitive conditions (acute respiratory distress, stroke, and chest pain). The exclusion criteria included Laboring mothers, critically ill patients with no reliable surrogate, and cases of inter-facility transfer from another health care facility.

A modified standard survey instrument was used [7]. The questionnaire included demographics, previous ambulance use, how the patient accessed an ambulance, length of time to access the ambulance, care given at the scene, care given during transportation, triage category, and parameters for knowledge and perception assessment. In addition an in-depth interview was conducted on 30 purposively selected participants. Before initiating data collection, the questionnaire was pre-tested on patients and individuals who accompanied them in all of selected hospitals. Pre-testing assessed the applicability of the survey instrument and facilitated improvement of the format and wording. Analysis of data from a similar study in Gabon [8] and our pre-test results indicated that responses were acceptably similar whether interviewing the patient or a relative who was at the scene. Accordingly, a patient was selected for an interview if he/she could provide history or a companion of the patient was able to provide the full interview. Off duty nurses working in emergency rooms of the study hospitals were recruited as data collectors. Ten data collectors, two for each hospital were trained in data collection. The training was in two sessions, one day a week, for two weeks prior to the data collection time.

During the data collection, the questionnaires were checked for completeness and quality by supervisors and lead investigators. Data was collected during March 2017 and then the lead investigators cleaned, coded, entered and analyzed results using SPSS version 20. Descriptive statistics and logistic regression were done for quantitative data. An analytic model was developed to assess the effect of independent variables on the primary outcome variable (ambulance use). Independent variables included sex, age, language spoken by responder, distance from scene to hospital, car ownership, patient accompaniment, type of emergency, scene of emergency, belief of importance in ambulances, believing ambulance is faster than taxi, previous ambulance use, attitude towards quality, and educational status, all of which were assessed as part of the survey. To explore the relationship between independent variables and the outcome variable, descriptive cross tabs with chi-square analysis were performed. Variables were selected according to previous literature and the expert opinion of the authors. Variables that were found to be significant in bivariate analysis at the 0.1 level were entered into a multivariate model. A qualitative data from a total of 30 interviewees, 6 from each hospital, was translated, transcribed arranged thematically by the lead investigators.

A written ethical clearance regarding the study was obtained from St. Paul Hospital millennium medical college institutional review board (IRB). Permission was received from the participating hospitals’ emergency departments. Informed consent was directly obtained from study participants. For unconscious patients, their surrogate was asked to provide consent and for minors (defined as children less than 18 years of age); their guardians were asked to consent. Autonomy and confidentiality were preserved during the study and during dissemination of the results. Qualitative data was coded and kept safe with the lead investigators and removed from the audio data collection tool.

A large majority of respondents were male (n = 297, 69.2%) and the mean age of all patients was 33.8 years (SD, 10), with higher number of respondents were between 31 and 50 years (n = 210, 49.0%). Most of them were self-employed (n = 137, 31.9%) and majority have no car (n = 389, 90.7%). In regards to residency, most of the respondents resided in Addis for more than 25 years (n = 166, 38.7%) (Table 1).

A significantly lower proportion of patients (n = 87, 20.3%) used an ambulance for their current emergency condition; the majority of patients used a taxi or private car (n = 326, 74.2%). Patients were brought to hospital mainly accompanied by family members (n = 329, 76.7%). The most common emergency scene was the patient’s residence (222, 52.8%), followed by road-side (n = 159, 37.1%). A larger proportion of trauma patients were extricated and moved from the site of trauma by a by stander (n = 116, 58%), followed by police (n = 48, 24%). Amongst the ambulance utilized groups, the median waiting time for ambulance arrival was 21 min (IQR 17). Eighty-one (93.1%) of the ambulance transported patients were able to name one ambulance call number (Table 2).

Multivariate logistic regression analysis was done to assess the factors determining the dependent variable (Ambulance use). The factors in the model included sex, age, language spoken by responder, distance from scene to hospital, car ownership, patient accompaniment, type of emergency, scene of emergency, believing ambulance is faster than taxi, previous ambulance use, and educational status. Significant association was found in language, patient companion, and previous ambulance use (Table 3). Patient companion was categorized as police man, bystander and family /relatives. Patients who were companied by police (OR = 1.53) were positively associated with ambulance use. Language spoken by respondents was categorized as Amharic and non-Amharic speakers. Failure to speak Amharic was negatively associated with ambulance use P = 0.001. Previous ambulance use (P = 0.001) was positively associated to current ambulance use. Believing an ambulance is faster than a taxi or private car was positively associated with ambulance use, but the data were not statistically significant (P = 0.77).

Almost all of the participants (96.3%) believed ambulance was important for their emergency condition. The two common mentioned benefits from ambulance use were treatment during transport (n = 177, 42.4%) and faster transfer to hospital (n = 165, 40.4%). A higher proportion of participants believed ambulances are safer (n = 341, 79.5%) and faster (n = 298, 69.5%) than taxi or private car. The majority also believe there is high quality care in ambulances (n = 300, 69.9%). However, a significantly high proportion of participants perceived that there are not enough ambulances in Addis Ababa (n = 305, 71.1%). There is also significant discrepancy between the acceptable and actually waiting time. The median acceptable time to get the ambulance (according to respondent’s perception) was 16 min but actually perceived ambulance waiting time was 40 min. See Table 4 for more detail.

A slightly higher proportion of interviewees named at least one ambulance number (n = 229, 53.4%), and the majority of them recalled 939 as the emergency ambulance number (n = 179, 78.2%). In regards to the source of information for ambulance numbers, a majority of the respondents were informed via radio (n = 128, 56.3%), followed by television (n = 54, 23.9%). Most participants knew there was a free ambulance service in Addis Ababa (n = 311, 71.2%) but almost half of the participants did not believe they could get the ambulance by calling the toll free ambulance number 939 (n = 202, 48.2%).

During in-depth interviews regarding the barriers to use ambulance, the majority of interviewees confirmed the findings of the quantitative results. Long arrival time for ambulance called during emergencies, not enough distribution of ambulance stations, and difficulty of accessing the phone were additionally emphasized. Most responded (n = 27, 90%) that if they can get to a phone, they are sure they will get the service. They were asked if the ambulances were misused in the city, for which most participants (n = 23, 77%) opined that, as far as their knowledge goes, ambulances were being used for only intended purposes. Only a few (n = 3, 10%) reported misuse other than for the intended purpose. See Table 5 below for selected quotes from the qualitative responses.

The study was conducted only in urban public hospitals with trauma centers, making generalization to other settings difficult. It may also not represent actual community level utilization. Estimations of ambulance waiting time may have been effected by recall bias. In addition, the relatively small sample size of the study may have masked the effect of some factors contributing to ambulance utilization.