This cross-sectional observational study based on a standardised postal survey of patients who had received the DEMAND intervention. The intervention is described in detail in another paper [10]. In short, at the medical on call service 116117, trained medical stuff conducted computer-assisted structured initial assessment with the patient via telephone and gave recommendations of suitable treatment settings after appraisal of the specific situation of the patient. The DEMAND intervention was evaluated in the telephone services of eight ASHIPs located in the German federal states of Bavaria, Brandenburg, Bremen, Hesse, North Rhine-Westphalia (represented by two ASHIPs in the regions North Rhine and Westphalia-Lippe), Schleswig-Holstein and Thuringia. Compliance and patient satisfaction with the intervention have been described elsewhere [10]. DEMAND has also been implemented in selected hospitals, which has been analysed in another study [15].
Selection of participants and data collection
The selection of participants and data collection have been described in detail in a previous publication of Schäfer et al. [10]. In short, due to a strict approval process for transferring personally identifying information by the German federal governments, the start of data collection was different between regions. In North Rhine-Westphalia and Bremen, the observation time was between January and December 2020, in Brandenburg between April 2020 and March 2021, in Bavaria between June 2020 and February 2021, in Hesse between July 2020 and March 2021, in Thuringia between October 2020 and March 2021 and in Schleswig-Holstein between January and March 2021.
The respective ASHIP produced list of patients who took part in the intervention after each finished month of the observation time. Exclusion criteria were age less than 18 years or no reported valid postal address. The study centre received the contact data by electronic data transmission. A total of 10 000 patients were randomly selected. Questionnaire and patient information were sent by mail to the selected patients. The patients gave their consent to study participation by returning the completed questionnaires to the study centre. Patients were excluded retrospectively, if the letter could not be delivered due to an incorrect postal address, if the patients reported that they did not use the telephone service of the respective ASHIP, and if relatives or other household members reported that the patients were not able to fill out the questionnaire (eg, due to functional limitations) or that the patients had died.
Measurements
Recommended and used settings of emergency care were collected by patient self-report. Multiple answers were possible. Patients’ satisfaction with the computer-assisted initial assessment and the used setting were evaluated by rating their agreement to eight statements on a four point Likert scale.
Data on sociodemography, health status at the time of the emergency call, past health service use, and health literacy were also collected. Sociodemographic data involved age, sex, living arrangement, educational level of the patients, and country of birth of the study participants and their parents. The educational level was coded according to the CASMIN classification [16] in three hierarchical categories. Self-rated health, depressiveness, subjective treatment urgency, and reasons for consulting the telephone services of the ASHIP were assigned to health status. A visual analogue scale ranging from 0 (indicating the worst) to 100 (indicating the best imaginable health status at the day of the initial assessment) was used for assessing self-rated health.
Patients rated their subjective treatment urgency on a numerical rating scale ranging from 0 (indicating no urgent need for treatment) to 10 (indicating very urgent need for treatment and/or life threatening condition). The Patient Health Questionnaire PHQ-2 [17] was used to screen for symptoms of depression in the two weeks before initial assessment. It consists of two items and results in a summary score ranging between 0 (indicating no symptoms) and 8 (indicating both symptoms occurring almost every day). Patients reported their consultations reasons by open questions. JHO and AM coded these information retrospectively in the International Classification of Primary Care, Second Revision (ICPC-2) [18], which facilitates grouping by organ system and diagnosis type (eg, “symptoms/complaints”, or “infections”). The past health service use in the three months before the initial assessment (general practices, specialist practices, hospitals, and emergency services) was also collected by patient report. To collect health literacy we used the European Health Literacy Questionnaire HLS-EU-Q16 rated on a four point Likert scale. For our analyses, we dichotomised the data. The summary score was divided in three hierarchical categories: inadequate (0–8 points), problematic (9–12 points) and sufficient (13–16 points) [19, 20].
Statistical analyses
The study population was characterised by descriptive statistics. For analysing the association between recommendations, and patient characteristics we conducted multilevel, multivariable logistic regression models adjusted for random effects at the level of regions and months of observation within regions. Independent variables included sociodemographic data and the variables describing health status, past health service use, and health literacy. The respective setting recommendations were defined as dependent variables, which resulted in seven statistical test. In order to correct the Alpha-level for multiple testing, we performed Bonferroni-adjustment for these analyses and defined an Alpha-level of p ≤ 0.007 as statistically significant. All statistical analyses were based on the available data and performed using Stata 15.1.